Sensitive Periods in Behavior: Song, Language and Sight

A male white-crowned sparrow raised without hearing adult song still sings as an adult, but its song is simple and abnormal. Playing it the right song later in life does little to fix it.

The bird is not deaf and its vocal organ works. What it missed was an experience at the right age. Many behaviors, from song to grammar to fine vision, are built this way: the nervous system expects certain input during a window of development, and what arrives in that window shapes the circuit. The questions for this lesson are how sharp those windows are, what sets their edges, and how much of the evidence from animals carries over to people.

Critical period vs. sensitive period: how sharp is the window?

A critical period is a time during which a behavior is especially susceptible to, and requires, particular experience in order to develop normally. Think of these times as windows of opportunity and of vulnerability. The course draws one distinction. A critical period begins and ends abruptly, like a door that opens and then shuts. A sensitive period rises and falls gradually, like a tide. Most of the behavioral windows in this lesson turn out to be the second kind, and naming the difference keeps you from reading a graded decline as a locked door.

Windows are where the old argument of nature vs. nurture becomes an experiment. Genes build the circuit that can learn and set when it can learn; experience supplies what is learned. Neither can be removed without changing the outcome.

Experience does not start the process. The earliest behaviors come from simple circuits that run on their own. In chick embryos, leg movements continue after the sensory input to the lumbar spinal cord is removed, so they are generated within the cord. One useful scheme sorts early behaviors into three kinds. Anticipatory behaviors, such as rhythmic bursting in the spinal cord or the grasp reflex, rehearse circuits that will be used later. Adaptive behaviors are transient and serve a single stage, as hatching does. Subtractive behaviors are building blocks that are later absorbed into something more complex, as crawling gives way to walking. Learned behaviors are laid on top of this base.

Song: listen first, practice later

Songbirds give the clearest case of a window for learning. In the 1960s and 1970s Peter Marler raised young white-crowned sparrows in sound-isolated chambers and played them recorded song at chosen ages. Birds that heard song early in life later sang good copies. Birds that heard nothing developed simple, abnormal songs, and recordings played to older birds had little effect. Song learning has two phases. In the sensory phase the bird listens and memorizes. In the sensorimotor phase it practices, producing rambling sounds that it gradually shapes until they match the memory. In white-crowned sparrows the two phases can be months apart. Masakazu Konishi showed that a sparrow deafened between them sings a very poor song: it must hear itself to correct its voice. Skill learning doesn’t come easy, even for a bird.

The zebra finch, now the standard laboratory songbird, runs a faster schedule. Only males sing. A young male memorizes a tutor, usually his father, over roughly his first two months after hatching, overlaps the phases, and settles on a fixed, or crystallized, song by about three months. Timings differ between species, so a number from one does not transfer to the other.

The song circuit was mapped in adult canaries by Fernando Nottebohm, Tegner Stokes and Christiana Leonard in 1976. Lesions of the forebrain nucleus HVC, or of the nucleus RA to which it projects, disrupted song, and RA sends fibers to the brainstem motor neurons that control the syrinx, the bird’s vocal organ. That HVC to RA to brainstem route is the motor pathway. A second route, the anterior forebrain pathway, runs from HVC through area X and the thalamus to the nucleus LMAN, and from LMAN back to RA. In 1984 Sarah Bottjer, Elizabeth Miesner and Arthur Arnold found that lesions of LMAN (called the magnocellular nucleus of the anterior neostriatum in the paper) in juvenile male zebra finches disrupted song development but did not affect stable song in adults. The anterior forebrain pathway matters while song is learned.

What do its synapses need? The NMDA receptor is a glutamate receptor whose channel opens only when glutamate is bound and the membrane is already depolarized, so it reports that two cells were active together (see the lesson on ion channels as receptors). In 1996 Mark Basham, Ernest Nordeen and Kathy Nordeen infused the NMDA receptor blocker AP5 into the anterior forebrain of juvenile zebra finches before each of ten tutoring sessions. Those birds copied significantly less of the tutor song than controls, while birds given AP5 on non-tutoring days, or into the cerebellum, learned normally. Treated birds could still tell zebra finch song from canary song, so they were not simply deaf to it. The result places a requirement for NMDA receptors in that region in the step of memorizing. Hearing a tutor also stabilizes dendritic spines on HVC neurons in young zebra finches, a structural trace of the memory forming.

Design problem

You want to know whether a zebra finch has a sensitive period for learning its tutor’s song, and whether experience itself helps to close that window. Design the experiment, and say what result would tell you that hearing a tutor closes the window rather than age alone.

Language: what the age effects do and do not show

In 1967 Eric Lenneberg proposed a critical period hypothesis for language: a first language can be acquired only between early childhood and puberty. It is hard to test, because nearly every child hears language from birth. The case of a girl known as Genie, found at age 13 after years of severe isolation and abuse, is often cited, but a child deprived so completely was deprived of much more than language, and such cases cannot isolate a language window.

Second languages offer a test with better controls. In 1989 Jacqueline Johnson and Elissa Newport studied 46 native speakers of Korean or Chinese who had moved to the United States between the ages of 3 and 39. Each judged whether English sentences were grammatical. People who arrived at ages 3 to 7 scored like native speakers. Scores fell steadily with age of arrival up to about puberty. Among later arrivals, scores were lower and varied widely, and no longer tracked age of arrival.

Read the result carefully. It shows a strong age effect on grammar. It does not show a door shutting at puberty: the decline began in childhood and was graded. A study of about 670,000 online participants by Joshua Hartshorne, Joshua Tenenbaum and Steven Pinker in 2018 estimated that grammar-learning ability stays high until about age 17 and then declines. Whether such declines reflect a change in the brain’s plasticity, competition from an entrenched first language, or differences in the kind and amount of input is still debated. The evidence fits a graded age effect better than Lenneberg’s boundary at puberty, although Hartshorne and colleagues themselves read their data as a sharply defined critical period with a late offset.

Sight: the world of the newborns is blurry

A newborn sees, but coarsely. Acuity and contrast sensitivity, the ability to detect faint differences in brightness, both improve over the first years of life, steeply in the first year. How to ask babies questions? You use where they look. Infants prefer to look at a pattern rather than a plain field. In forced-choice preferential looking, a striped grating appears on one side and a gray field of equal brightness on the other; an observer who cannot see which side holds the stripes judges from the infant’s gaze. The finest grating the infant reliably looks at gives its acuity in cycles per degree, one cycle being one dark and one light stripe.

Snellen denominator ≈ 20 × (30 ÷ finest grating in cycles per degree)

By this method a newborn’s acuity is often near 1 to 2 cycles per degree, roughly 20/600 to 20/300. Numbers depend on the method: recordings of visual evoked potentials from the scalp give better values at the same age than looking tests do.

Children born with dense cataracts in one or both eyes show what visual input does. In 1999 Daphne Maurer, Terri Lewis, Henry Brent and Alex Levin reported 28 infants treated between 1 week and 9 months of age. Right after treatment their acuity was no better than a newborn’s, whatever their age, and some improvement appeared after as little as an hour of visual input. Improvement needs patterned input, not age alone. Yet the same group found lasting deficits in some functions after early deprivation, among them fine acuity and processing of the spacing between facial features, even when the cataracts were treated within the first months. Different visual abilities have different sensitive periods, and some are affected by deprivation before the ability itself has appeared. The cortical mechanisms behind such windows are taken up in ocular dominance and critical period control.