When the Bough Breaks: A Contribution to Falk’s Hypothesis

Falling, not waving: the sigh startle as a vestige of failure to cling

Falk has proposed that a crucial event in the emergence of language arose when hominin infants could no longer cling to their mothers’ underbellies. As offspring came to be placed on the ground temporarily while their mothers foraged, the adaptive pressure favored the development of a protolanguage or motherese (Falk 2004, 2016). Falk’s hypothesis is important for many reasons, not least of which is its focus on the mother–child dyad, yet it omits what I believe to be a significant piece of the puzzle, which when put in place both strengthens her argument and opens new lines of inquiry.

The missing piece is the gradual nature of adaptation. Between the closest quadripedal ancestor of humans and the first fully upright hominin lies a period of some millions of years. If hominin infants ceased to be able to cling to their mothers’ underbellies as a result of bipedalism, the change is not likely to have occurred from 1 day to the next. Even if there were a relatively abrupt decision by mothers to put their offspring down, it would likely have been preceded by a phylogenetic phase during which infants held onto their mothers more tenuously. This would have been true regardless of the driving force behind bipedalism, due simply to the gradual nature of the progression from clinging to no longer being able to cling. For hominin infants, in the glacial process of evolution, there must have been a protracted period marked by the anticipation of falling from their mothers’ underbellies, strategies for avoiding those falls and failures to avoid them, with corresponding reflexive responses adapted to these conditions.

My hypothesis is that a vestige of this transitional phase survives in the sleep of today’s infants in the form of the sigh startle. Granting in advance the dangers of extrapolating evolutionary conclusions from contemporary conditions, the sigh startle contains clues that, I propose, allow us to go further and reconstruct specific aspects of this transitional phase. No new experiments are offered. Rather, the following is presented in the spirit of the amateur astronomer who notices something novel in the night sky and endeavors to report it as cogently as possible to others with greater expertise.

As identified by Thach and associates (Thach and Lijowska 1996; Lijowska et al. 1997; Wulbrand et al. 1998, 2008; McNamara et al. 2002; Franco et al. 2010), the sigh startle follows a distinct sequential pattern. With an increase in CO₂ and a decrease in O₂, the sleeping infant responds first with a sigh, or augmented breath, coupled with a startle response consistent with the Moro reflex (extension of the neck, arms thrown wide with fingers completely extended, then arms adducted), followed by thrashing of the limbs and, finally, arousal (eyes open, awake) and crying. This stereotyped sequence is not always completed, but retains the order of responses in such cases and is proposed to be a chained reflex, in which one response elicits the next in cascade fashion. By the end of the first year, the sigh startle generally disappears or is greatly diminished. Prior to its disappearance, however, statistics suggest that the phenomenon is common: Gerard et al. (2002) observed a group of 22 infants between the ages of 24 and 180 days and recorded a mean frequency of more than 40 sighs per hour and more than 20 startles per hour in unswaddled infants during non-REM sleep, with higher frequencies in the same cohort during REM sleep.

The sigh startle has been extensively studied for its possible role in sudden infant death syndrome, or SIDS, which poses the greatest risk from the second to fourth month (Kinney and Thach 2009), but the connection remains unclear. Lijowska, for example, notes that the thrashing aspect of the sequence actually does not greatly reduce the risk of asphyxiation. Of a group of infants placed face down, fully 40% did not reduce the levels of CO₂ by thrashing (Lijowska et al. 1997; Franco et al. 2010). Moreover, in response to spontaneous airway closure during sleep, no more than 18% of sleep obstructions terminate with EEG arousals (Katz et al. 2012). There are also indications that thrashing actually increases the risk of death for a prone infant sleeping on an uneven surface, due to the possibility of the face becoming wedged in irregularities in the sleeping surface and the airway more severely obstructed (Thach and Lijowska 1996).

The leading explanation accounts for these findings on the basis of cost–benefit: because infants benefit from uninterrupted sleep, the sigh-startle sequence is an adaptation selected for lower-risk respiratory challenges for which cortical involvement would be counterproductive (ibid). By this reasoning, however, a set of reflexes that yields longer periods of uninterrupted sleep would also increase the opportunities for those reflexes to fail, which leaves the adaptive pressure for its genesis in doubt.

It has also been suggested that the separate elements of the sequence persist due to known benefits specific to each of them: the sigh for maintaining lung compliance, the startle for early limb development, and thrashing for even development of the skull (through periodic change of position in the sleeping infant) (ibid). These benefits, however, do not explain why the separate reflexes should be linked, as they so clearly are. It is proposed here that the sigh startle can be more adequately explained by considering it within the context of gradual evolutionary changes that introduced the threat of imminent falls.

The presence of the Moro response offers perhaps the most easily accepted connection to falling. This reflex can be elicited by holding a newborn face up and allowing its head to drop backward into one hand. While it can be elicited by other means, the head drop stimulus indicates vestibular involvement, which has led some to theorize that it owes its origins to the attempt by primate infants to cling to their mothers’ fur. Indeed, the Moro reflex is regularly observed in newborn apes and monkeys in cases where elicitation of the palmar grasp reflex is inhibited (Futagi et al. 2012).

Thrashing is not part of the classic Moro response in humans, however, nor does it accompany the clinging response in other non-human primates. Drawing from their field observations, Savage-Rumbaugh and Fields note a clear distinction in the rotational hand waving and leg kicking movements of human infants when compared with the clinging behavior of their primate counterparts (Savage-Rumbaugh and Fields 2011). Other mammals have been observed to sigh and startle during sleep, but not to thrash as humans do; in a study of sleeping kittens, for example, somatic activity was limited to the neck and eyes (McGinty et al. 1979). How then might thrashing have come to be part of the sigh startle sequence?

Wulbrand et al. (2008) provides data that may be relevant to the question. Monitoring infants from 34 to 134 days, this study found that the intensity of the startle steadily increased even as the latency of the overall response grew shorter. The authors were unable to account for these results and proposed that they might indicate an increase in the infant’s metabolic rate, or a lowering of the chemoreceptor thresholds for arousal. An alternative explanation, and not necessarily a mutually exclusive one, is that the increased intensity and greater speed of the response are due to an overlapping of the Moro reflex and—for lack of a better term—the ‘adult’ startle reflex.

Although the Moro reflex and the adult startle reflex are often conflated in non-scientific literature, they are considered to be distinct behaviors (Hunt and Landis 1938). The former is a primitive reflex said to disappear in humans at around 6 months, while the latter, a protective response to danger, actually appears almost at birth and persists throughout life. The Moro is also slow (450 ms to the onset of abduction, 998 ms to arm embracing arc) (Roennqvist 1995; Bijesh et al. 2013) relative to the adult startle (head: 60–120 ms, neck: 75–121 ms, shoulder: 100–121 ms, arms 125–195 ms, legs: 145–395 ms) (Davis 1984).

The thrashing movements in the sigh startle are broadly indicative of a superimposition of these two reflex responses. Unlike the Moro reflex, the adult startle reflex does not include any significant abduction, or extension, of the arms, but is characterized by their pronounced adduction. During any overlap, then, the infant would sometimes experience near simultaneous but contradictory reflexes—one to extend, the other to flex—and so would be likely to generate the uncoordinated upper limb movements, that is, thrashing. In other words, thrashing is not necessarily random but may be evidence of the competing responses of “cling” or “break your fall,” with the second startle indicators being masked by those of the first.

The increasing startle magnitudes and decreasing latencies reported in the Wulbrand study also make sense in an evolutionary context. Hominin infants, like infants today, would have grown heavier as they aged, becoming more likely to fall than to succeed in clinging. As greater anxiety levels have been correlated with higher adult startle magnitudes (Poli and Angrilli 2015), the result would presumably have been magnitudes of the adult startle that were proportionally higher with the increasing threat of injury. By the same reasoning, the adult startle would also tend to occur sooner in the sequence for an older infant than it would for a younger one (i.e., the loss of purchase on the underbelly would occur sooner at a greater weight than it would at a lesser weight). An adult startle occurring earlier and more intensely in a given sigh startle episode would then exert an earlier and more definitive adductive response, pulling the arms in earlier, thereby reducing the overall latency of the superimposed reflexes within the individual infant as it developed.

The sigh may also gain new significance when considered in relation to the danger of falling. Although obviously not unique to humans, sighs have been shown to decrease physiological tension (Vlemincx et al. 2016), and so could be expected to loosen the grip of the clinging hominin infant. The timing of sigh startle responses in Lijowska et al. (1997) is consistent with this sequence of events. In an infant whose face is covered with a cloth, a near simultaneous sigh and reflex response initiate the Moro abduction, which reaches its fullest extension after about 1 s and ends after about 2.5 s. Within this time frame, the sigh would cause the limb muscles to relax, triggering the association of an imminent fall and an accompanying adult startle response. As described above, the thrashing is then proposed to consist of the Moro abduction–adduction superimposed on the adult startle adduction.

Hitting the ground running: the possible presence of the stepping response in the sigh startle

As mentioned, thrashing is not a signature trait of the Moro reflex. It should be noted, however, that the leg movements of the sigh startle do not follow the same kinematic pattern as do the arm movements. Where the arm movements of the sigh startle are asymmetrical, the leg movements are generally symmetrical (synchronous) and involve repeated flexion and extension (ibid), making them characteristic of kicking. This merits closer attention because infant kicking is closely associated with the infant stepping response.

Thelen has shown that the infant stepping response, while present at birth, becomes unsustainable in upright infants by the age of about 2 months due to the increasing weight of the legs. Stepping does not disappear, however, but continues to be present in air stepping (stepping while supported above a surface) and the kicking of supine infants, during which the tibialis anterior is active and gravity substitutes for the action of the gastrocnemius soleus (Thelen and Fisher 1982). Absent special intercessions (i.e., stepping under assisted conditions), the ‘practice arena’ for the suppressed stepping response is then largely found in the action of kicking (Thelen 1985).

A comparison of contemporary data to the fossil record supports the early involvement of kicking in the transition away from clinging. Ontogenetically, our legs grow faster than all other body parts except the head during childhood. Phylogenetically, human legs are longer than those of other primates (Bogin and Varela-Silva 2010)—an anatomical change in the hominin line that became pronounced around the time of Homo ergaster/early Homo erectus (Anton 2012). It is also generally agreed that H. ergaster/early H. erectus no longer had prehensile feet (Harcourt-Smith and Aiello 2004), with even the partial foot grasp of the Dikika child (DeSilva et al. 2018) a thing of the distant past. As a result, infants at this stage of evolution would have been likely to lose their grip in the lower limbs before the upper limbs, because of their relatively greater weight and more tenuous foot grip, and then, abandoning hope of holding on with the legs, to begin kicking and air stepping in advance of meeting the ground.

Moreover, infants today step in a manner that accords well with the hypothesis of infants falling from their mothers’ underbellies. The transitional phase of evolution is most logically envisioned as taking place when hominins were finally abandoning arboreal life completely in favor of committed terrestrial bipedalism—again, around the time of H. ergaster—simply because of the high mortality rate that would have ensued had infants fallen to the ground from any great height. (Indeed, if infants began to fall while still arboreal, it would have exerted a strong evolutionary pressure for hominins to descend from the trees, even after accounting for the opposite pressure of avoiding predators.) In this respect, the difficulty in clinging would have occurred primarily when the mother was walking on the ground. The logical consequence, then, is that the clinging hominin infant would have fallen facing its mother’s ventrum, 180 degrees away from her direction of motion.

This supposition is in line with findings related to kicking and air stepping in human infants exposed to an optic flow. Recent studies by Barbu-Roth and associates have shown that infants held above a visual treadmill will air step and kick with greater frequency when the optic flow below translates as a motion away from them (Barbu-Roth et al. 2014). As with the increasing startle magnitudes in the Wulbrand study, the researchers were unable to account for these results. The data become more explicable, however, if it formerly was the case that hominin infants fell “feet first, facing backward” in relation to their mothers’ direction of travel, as the optic flow below would then have been moving away from them.

Of course, kicking while clinging or being supported above a surface is not necessarily the same phenomenon as kicking during a sigh startle. A stronger correlation between kicking/air stepping, and the leg movements of the sigh startle awaits a close inspection of experimental data, both existing and yet to be generated. The accumulated evidence is strong enough now, however, to hazard a plausible sequence of responses for our ancestors at some indeterminate midpoint between their clinging and post-clinging lives.

Prior to the transitional phase, hominin infants clutched onto their mothers’ underbellies, as non-human primates do today. Gradually, as they began to have more difficulty clinging, a cascade of responses emerged: (a) a more pronounced Moro response (i.e., hugging in a wider, tighter arc), (b) an obstructed airway due to tighter grip, (c) a sigh to increase airflow, (d) a decrease in muscle tension from the sigh and a subsequent loss of grip in the lower limbs, leading to, (e) a stepping/kicking response, and (f) an adult startle, which when combined with the Moro response induced thrashing. In short, the transitional sequence between clinging and being put down is proposed to have resembled a backward ‘hang glider landing’, in which the infant engaged in some combination of air stepping, kicking, and landing stepping backward, or simply thrashed and fell, or both.

It is also likely that mothers attempted to rescue their infants from falling—with varying degrees of success—at any of these stages. Eventually, however, both the increasing phylogenetic upright stature of mothers and the increasing ontogenetic weight of infants, especially in the legs, would have made clinging untenable.

While this paper takes its inspiration from Falk, its premise actually entails no commitment to specific maternal adaptations once the transitional phase was over. Mothers who placed their children down momentarily, carried them in slings, left them with groups of caretakers, or devised other solutions yet to be imagined, could have done so just easily in response to falling or near-falling as to any other transition away from clinging. In this respect, the immediate concern is not maternal adjustments to the inability to cling but rather the destiny of the sigh startle once it was putatively in place.

Waving, not falling: the sigh startle as a possible exaptation for speech

The preceding analysis gives a new explanation as to how the sigh startle came to exist as a series of chained reflexes, but it does not tell us why it should have persisted even after the danger of imminent falls had been resolved. Nor does it explain why it should have persisted when it presented new dangers of its own (the risk of asphyxiation). Falk’s hypothesis holds that the transition from away from clinging created the conditions for a protolanguage between mother and child. Savage-Rumbaugh and Fields have elaborated further on these conditions through a close investigation of the physical separation of mother and child and the affordance of that separation for new semantic expressions (the distinction between ‘I’ and ‘me’) (Savage-Rumbaugh and Fields 2011). Clearly, the emergence of language would count as an adaptive advantage for hominins who no longer clung to their mothers. The question accordingly arises as to whether the sigh startle is implicated in some way not just in strategies for breaking a fall but in the motor production of speech as well.

Infants do not step between the ages of 2 and 6–8 months, but do exhibit both sigh startle and babbling behavior during this period, which theoretically allows for an exaptation of sigh-startle stepping behavior during the non-stepping interim. More pointedly, one notes that the sigh startle is analogous to the production of babble. Iverson has shown that stereotyped limb movements—kicking, banging, and arm waving—typically precede attempts at canonical babble, suggesting the emergence of multimodal communication (Iverson 2010). The sigh startle follows a similar pattern—kicking and thrashing (competing arm movements) followed by a cry. Nominally, the sigh startle is distinguished from the stereotyped movements of babble by pronounced respiratory challenges, especially the challenge of hypoxia. We might wonder, then, what role hypoxic conditions play, if any, in early speech production.

That hypoxia leads to negative outcomes is almost too obvious to mention. At sufficient levels and durations, it can result in brain injury, stroke, cancer (Gilany and Vafakhah 2010), and, of course, at the far end of the spectrum, death. The case against hypoxia is not so simple, however, because intermittent hypoxia, as distinct from severe hypoxia, has been shown at times to be detrimental and at others to be markedly beneficial—not only rehabilitating walking in patients with spinal cord injuries (Ting et al. 2015) but, as a general matter, enhancing neural plasticity (Dale-Nagle et al. 2010). Any adaptation that may have preserved or built upon the sigh startle response while capitalizing on the benefits of hypoxia thus warrants further investigation.

Phylogeny at least does not rule out the possibility of such adaptations. Drawing on the existing literature on speech production, Davidson compared the skulls of modern humans, H. erectus and chimpanzees, and concluded that bipedalism and klinorynchy (the migration of the jaw to a position under the skull) forced the upper and lower portions of the supralaryngeal vocal tract into a right angle bend, thus bequeathing H. sapiens with both obstructive sleep apnea and the capacity for vowels—and, more broadly, for buccal speech (Davidson 2003). By this account, the evolutionary conditions for both the sigh startle and full-blown speech coincide. Whether hypoxia may have played a causal role in speech production was not a concern of the analysis. There is, however, at least one plausible ontogenetic scenario in which it does. This scenario essentially extends the chained reflex of the sigh startle to include the effect of thrashing on hypoxic levels and, as a direct consequence, to a pharyngeal response that can be interpreted as a precursor to canonical babble.

Limb movements reduce intermittent hypoxia

Kesavan et al. (2016) have shown that limb activity in preterm infants reduces incidents of intermittent hypoxia compared with baseline in the same subject. Using a simple, vibrating device (0.3 gm/128 Hz) that excited the proprioceptors of the hands and feet, the authors were able to reduce episodes of oxygen desaturation from levels below 90% by as much as 28%. If this study is any guide for achievable results in term infants, it suggests that the sigh startle may serve a self-limiting function, insofar as hypoxic conditions elicit limb movements, and limb movements reduce hypoxic levels. In other words, thrashing may contribute to milder hypoxic levels, which in turn reduce the intensity of thrashing, creating a negative feedback loop that continues until the chain of responses subsides independent of arousal. In keeping with the prevailing belief that thrashing contributes to arousal (Thach and Lijowska 1996; Katz et al. 2012), it may also be that the negative feedback loop fails in extreme hypoxic events and that thrashing then contributes to arousal by virtue of its unabated intensity.

So much remains within the etiology of the sigh startle as a protective response. The relevance of this feedback loop to speech can be seen in parallel findings in respiratory and phonological research.

Intermittent hypoxia decreases glottal constriction

One of the negative effects of hypoxia is that it decreases upper airway resistance (Rowley et al. 2007). Specific to the present discussion, arterial oxygen tension (a measurement for which oxygen saturation acts as a proxy) (Collins et al. 2015) and the threshold for the glottal closure reflex are inversely proportional. As the amount of oxygen in the blood drops, so is the glottis less able to close (Sasaki 2006). Intermittent hypoxia has also been demonstrated to cause a decrease in glottal constriction post-inspiration in rats (Moraes and Machado 2015). From a respiratory perspective, these results are undesirable, because they destabilize the upper airway (ibid). However, this same destabilization may yield a phonological byproduct.

Decreased glottal constriction is implicated in early speech-like behavior

Although research on speech production tends to emphasize the mechanisms of the oral cavity (tongue, lips, teeth, hard palate), Esling locates the pharynx (i.e., the upper airway minus the oral cavity) as the site of the earliest vocalizations upon which articulated speech is based. In particular he identifies the first ‘speech sound’ from a stricture point of view as the epiglottal stop, which is a function of the airway protection reflex upon inspiration of liquids or solids, followed at about 2–3 months of age by glottal stops, which involve careful openings of the glottis and slightly more laryngeal openness (Esling et al. 2015). Moreover, Esling has elucidated the larynx and the pharynx in which it is housed (as opposed to the oral cavity) as the site from which infants begin their explorations of other speech sounds that can be made, in a process he has called ‘pharyngeal priming’ (ibid).

From a purely morphological standpoint, the use of the glottis for articulation (incremental opening for phonation) can be compared with the glottal response to oxygen desaturation (an incremental inability to close), and at 2–3 months falls well within the developmental stage in which the sigh startle is observed. The prediction then is that, in curtailing hypoxic challenges, thrashing yields a respiratory benefit often enough that the sigh startle persists as a chained reflex despite the increased risk of asphyxiation, with phonological novelty appearing as a spandrel.

There seem to be two ways in which this result could come about. In one, adaptation operates by a kind of moral hazard, in which severe hypoxia is survived more often than it would be otherwise because thrashing can attenuate it; in this case, the pharynx undergoes plastic changes during sleep. In the other, thrashing is unable to overcome the severity of a hypoxic event and so gives way to arousal and to crying, with the glottis unable to close entirely post-inspiration until normoxia returns, thereby providing a window for the long-term facilitation of novel phonological formations in the pharynx.

If we hold fast to H. ergaster as the proximal heir of the original sigh startle, the foregoing begins to provide an interesting evolutionary picture. Being relatively hairless and long-legged, H. ergaster is often described as the first hominin to clearly exhibit many of the traits of modern humans, with one exception being in its spine, which was too thin to accommodate the number of nerves necessary for respiratory control—a crucial requirement for speech (MacLarnon and Hewitt 1999). But what if this lack of control is precisely what was needed for speech to arise? In order to avoid the circularity that tends to occur when the mother is assumed to be the first hominin speaker in the mother–child dyad, it may make sense to understand the earliest origin of speech behavior in terms of neonatal conditions that forced novel, uncontrolled articulations equivalent to present-day marginal babble, which only later came to be harnessed and rendered as syllables. For these articulations to have been exchanged between mother and child would then seem to depend minimally on the persistence of hypoxic-induced pharyngeal plasticity to reproductive age, a matter that lies beyond the scope of this paper.

Conclusion

An incremental model of Falk’s hypothesis reveals a possible inner structure to the transition between clinging and speaking, wherein falling contributes to the chained reflex of the sigh startle, which in turn contributes to the production of marginal babble. To this end, two explanations have been offered here for otherwise unaccountable results—the increasing magnitude of startles during the sequence (due to the appearance of the adult startle as provoked by the increasing threat of falling) and a heightened kicking response to an optic flow moving-away (due to a hereditary tendency to fall facing backward in relation to the direction of the mother’s travel). In addition, a negative motor-respiratory feedback loop that facilitates phonological exploration has been proposed.

As a speculative excursion, the model naturally leaves a great many questions unanswered. This was to be expected. The exercise was not to prove Falk’s hypothesis but rather to explore what would be entailed if an additional premise were true. In a sense, the result is several corollaries, which stand or fall somewhat independently of one another. It may be, for example, that genesis of the sigh startle is incorrect while its implication in marginal babble obtains, or vice versa. The decrease in sigh startle latencies, for its part, may be evidence of Hebbian learning rather than the result of weight increases; backward stepping may be a vestige of falling, yet have no bearing on speech production. And so on.

One cannot predict when the next fossil will be found or what it will reveal. In this respect, perhaps the most direct line of attack would be to investigate the presence of glottal activity during the sigh startle itself and, in parallel, to scrutinize any significant instances or patterns of hypoxia, induced or voluntary, during the early stages of babble production, especially in the context of the multimodal developments described by Iverson. Studies of this kind might show more clearly how novel infant pharyngeal activity, hypoxia, and the sigh startle interact with each other, and in turn serve to verify or refute whether these interactions can be linked to waking vocalizations in human infants.

Conflict of interest statement. None declared.

References