Babies scream for attention—and to get what they want. But, in some cases, their vocalizations may point to medical problems.

Is your baby hungry, sleepy or in pain? A mobile phone app claims to know the answer. The program can tell users why a child younger than six months is crying, according to its developers from the Yun-Lin branch of the National Taiwan University Hospital. To do so, the “Baby Cries Translator” analyzes the frequencies of the baby’s wails, looking for small acoustic fluctuations. It then compares the recorded pattern with a database and determines the likely reason for the outburst. The program asks the parents for feedback. It thereby learns to better guess what the baby wants and gauges how well it is doing: the app claims to correctly pinpoint why a newborn cries 92 percent of the time—a high success rate that reportedly drops as the child grows older. A Spanish company offers a comparable product: its “Cry Translator” runs on smartphones (there is also a baby monitor) and takes only a few seconds to suggest what might be bugging the kid. Simultaneously, it advises its users on how to soothe their little one.

Of course, no algorithm will be able to substitute for good parental instincts. But cry analysis could support the child’s caregivers—or possibly their doctors: in past decades researchers have found that infant squeals contain a treasure trove of information. Instead of focusing on the wants of the baby, as the app developers do, scientists have been trying to tease out information about potential health issues from baby cries. In infants’ vocalizations, they have searched for signs of neurological damage and genetic defects. This diagnostic approach has an obvious advantage: the toddler might be spared more uncomfortable or even dangerous examinations.

French pediatrician Jérôme Lejeune pioneered this research in the 1960s. He discovered that some babies’ high-pitched screams—almost catlike in sound—signal that these children are suffering from a genetic defect similar to Down syndrome. Lejeune aptly named the disease cri du chat, which translates as “cat cry.” The eponymous shrill squeals are caused by a malformation of the infant’s larynx. Affected children show various other symptoms, including growth defects, muscular dystrophy and a small, elongated head with a round face. To diagnose the disease, doctors always confirm their suspicions with a genetic test. Still, the distinct cries are a first, clear indication of the condition.

The larynx has a central role in almost all human sound production. Part of the breathing apparatus, it separates the throat from the windpipe. Together with the so-called vocal folds (also called vocal cords), it creates vocalization and speech. Any utterance starts here, with a tightening of the muscles around the vocal cords. When air is expelled from the lungs, the taut vocal cords begin to vibrate, which lets off a sound. Depending on the tautness of the vocal cords, the pitch rises—a cry of a healthy newborn produces 250 to 450 oscillations a second.

A child’s scream is unique in many ways. Its basic pitch is determined by the interaction of the vocal cords with the larynx. Together they create a “dominant frequency,” which forms the base of an individual’s vocalizations. Yet a voice is not static—it can be modulated to some degree. Properties such as volume, rhythm and overlaid tones produce variation within the vocal spectrum. These features are created predominantly by areas below the larynx, including the diaphragm, lungs and chest. The upper vocal tract mostly takes on the fine-tuning: it amplifies some frequencies but leaves others unaltered or suppresses them. The complex interplay creates the entire spectrum of human vocalization.

The impulse to cry originates in the brain in the limbic system and hypothalamus. From here neuronal signals spread to other brain regions, such as the brain stem and the cerebellum, which coordinate the formation of sounds. Signals are then sent to the muscles of the vocal cord, larynx, chest and stomach via a highway of nerves that runs through the spinal cord. The different components work like an orchestra: all parts contribute to the final makeup of the vocalization. If one or several of the contributors botch their part, the tune is off. Certain kinds of brain damage interfere with this complex interplay, and in this way, they may alter a baby’s cry.

Determining which parts are out of tune often needs more than an attentive listener. Scientists use technical aids that break a sound into its components and pick up on even the smallest abnormalities (irregularities). In 2013 doctors and engineers at Brown University announced they had developed a frequency analyzer that could screen an infant’s voice recordings for 80 different acoustic properties. According to the researchers, each of them may hint at potential health problems.

Telling Patterns: Spectrograms (visual representations of frequency spectrums) sometimes reveal health problems in an infant at first sight. The top pattern was recorded from a healthy newborn; the middle and the lower spectrograms were from a baby with microcephaly and one with oxygen deprivation, respectively. Source: From “Acoustic Measures of the Cry Characteristics of Healthy Newborns and Newborns with Pathologies,” by Yasmina Kheddache and Chakib Tadj, in Journal of Biomedical Science and Engineering, Vol. 6; August 2013

Their analysis is a two-step process: Initially, the software cuts the recorded cry into 12.5-millisecond snippets and scans them for sound frequency, volume and voicing (indicating the degree of involvement of the vocal cords). In the second step, the researchers use the gathered insights to categorize longer sections of the recording into “continued vocalization,” “silence” and “single scream.” Finally, the software analyzes different characteristics such as pauses in-between the cries, average pitch and change in tonality over time. Stephen Sheinkopf, a pediatrician at the Women and Infants Hospital of Rhode Island and one of the scientists involved in the development of the tool, envisions that this analysis could help diagnose autism at an early age. “It has long been known that [autistics] produce unusual sounds,” he explains. Moreover, the spectrum of disorders that could manifest in babies’ cries may be large. Trauma and brain damage, for example, are rare birth complications that are tricky to diagnose. “Cry analysis may enable doctors to identify children suffering from these conditions earlier,” Sheinkopf believes. They could then monitor the infant carefully and respond quickly should sudden problems arise.

Infants’ cries are like “a window to their brain,” says Barry Lester, a psychiatrist at Brown and primary investigator of the acoustic analyzer study. He had started to investigate the hidden messages in babies’ squeals in the 1970s. At that time instruments were much less sophisticated than the tool he helped to develop in 2013, Lester asserts. Researchers still had to work with simple spectrograms—graphic representations of the spectrum of frequencies of the sounds. Technicians analyzed the graphs, a task that was mostly done by hand. Yet even with these limited means, several important discoveries were made: in the 1960s Vincent R. Fisichelli and Samuel Karelitz of the Long Island Jewish Medical Center deduced that specific irregularities in the frequency analysis hinted at brain damage in newborns. The cries of these infants were too high-pitched, too short and featured double tones. Moreover, the babies showed a delayed response after being exposed to pain triggers.

Years later a team led by child care expert Katarina Michelsson of the University of Helsinki found another such link in infants: unusually shrill sounds coupled with an irregular basal frequency point toward a higher risk of death by suffocation. She went on to uncover several other syndromes tied to changes in the screams. Among them were encephalitis, hydrocephalus and Krabbe’s disease—a genetic defect that leads to progressive nerve damage and sometimes triggers bouts of screaming in affected children.

Sudden infant death syndrome (SIDS) has no known cause and displays no early symptoms. A group of researchers led by Lester and Michael J. Corwin of Boston University decided to screen babies’ cries for warning signs. In a large study in 1995, the team recorded cries of 20,000 healthy newborns and analyzed them for abnormalities using a computer-based method. In the course of the investigation, 12 infants died from SIDS. In their cries the researchers discovered certain traits that point toward a constriction of the upper vocal tract and a perturbation of the neuronal control of this area. Many children who did not die from SIDS, however, showed the same characteristic pattern. Thus, although the method identified a risk group, it was not suitable for routine medical screening, because there were too many false positives.

In the 1970s researchers became interested in whether drug consumption during pregnancy could affect the offspring’s screams. A team led by George Blinick, at the time at the Mount Sinai School of Medicine, had discovered that children from opioid-addicted mothers wailed at higher pitches. Lester became interested in that observation two decades later and set out to investigate this relation in greater detail. The babies he and his team studied in the following years had been exposed to drugs such as marijuana, alcohol, opiates and cocaine in their mother’s womb. The scientists observed various unnatural scream patterns, such as extremely high-pitched cries and an excessive amount of short interruptions. That, in combination with other irregularities, clearly points to problems in the control of breathing and the vocal tract in these infants—changes that are probably caused by neuronal damage and developmental defects of their nervous system.

Teasing out auditory effects of maternal drug consumption during pregnancy is still a hot topic in the cry analysis field. The technical means to study these phenomena have improved greatly since the 1970s. In 2014, using modern voice recording and analyzing equipment, a team led by Philip S. Zeskind of the University of South Carolina found evidence for differential effects in infants whose mothers had consumed cocaine while pregnant. The main determining factor was gender: affected boys screamed at a higher pitch and their vocalizations sounded unnaturally coarse. Girls, on the other hand, cried at a lower volume with fewer repeats and longer pauses.

Stressful stimuli do not affect males and females in the same way—that much was known previously. Scientists had also already suspected that contact with cocaine in the womb could have sex-specific effects on infants. They had found that girls displayed reduced responsiveness to their environment, whereas boys appeared chronically overstimulated. Only recently have Sheinkopf and Lester uncovered a factor that may help explain a differential response: in an article in June 2016 the researchers reported that the screaming tone depends on the expression of a gene that contributes to shaping the body’s stress response. They thereby observed a link between the sound of the wails and the effects of the drug withdrawal on the unborn child.

Over the years a long list of scream traits has been analyzed. Aside from obvious characteristics such as duration, volume and interruptions, scientists have looked at tonic keynote, superimposed tones, variability and dysphonia (which includes hoarseness and roughness of the voice, for example). For every feature, they investigated how it can deviate from the norm. A single disruption—for example, a fault in the regulation of the breathing apparatus or of the vocal cords—often causes an unusual pattern to emerge. These observations do not yet suffice to diagnose disease reliably enough. In many cases, a variety of health problems can affect the screams in a similar way. Conversely, different irregularities in the cries are sometimes caused by the same dysfunction. Doctors can detect traits that are out of the norm, but further examinations are necessary to make conclusive diagnoses.

Clinics are currently not using the method because of these limitations. Sheinkopf believes that more validation studies could persuade them to embrace the technique. “Cry analysis might become an independent diagnostic tool or part of a package that aims to estimate risk for diseases such as autism,” he hopes. Until then, it may be best to trust in parental instincts and listen for hidden messages in your infant’s cries. If your inherent translator fails, there is always trial and error: feed, cuddle, sing. An app would not recommend anything more sophisticated either.


Mother Tongue Affects the Sound of Cries

Babies do not only cry for attention—the wails also help them to learn how to talk. While they are screaming, they practice melodies that will later help them speak. A team led by biologist Kathleen Wermke of the University of Würzburg discovered that French newborns wailed differently than their German counterparts. While the French babies often produced ascending sound sequences, the Germans did the opposite.

The scientists believe this is because of the differing patterns of emphasis in the two languages. The fetus already notices the speech melody in the womb and practices to reproduce it once born. The older the babies, the more they learn to vary and combine the building blocks that make up the language. At some point, this allows them to speak their first word and then sentence—and ultimately to become fluent in the language.