CHAPTER 2

Building a Communication System in Infancy

ATHENA VOULOUMANOS AND AMY YAMASHIRO

DEPARTMENT OF PSYCHOLOGY, NEW YORK UNIVERSITY, 6 WASHINGTON PLACE, NEW YORK, 10003

FOUNDATIONS: EARLY SPEECH PREFERENCES AND THEIR ORIGINS

Many animals differentiate between different kinds of sounds and preferentially respond to the sounds that are relevant for their survival. One of the most important sounds in a fledgling animal’s environment is the vocalizations of conspecifics, or members of the same species. For example, swamp sparrows will preferentially perch on a branch that elicits the playing of their own species’ song relative to another branch playing the song of its close relative, the song sparrow, and when given equal exposure to both species’ songs during their song-learning phase will preferentially learn the song of their own species (Marler, 1990; Marler & Peters, 1977). Baby crocodiles that are near hatching listen for the calls of other baby crocodiles to start pecking to crack their shell (Vergne & Mathevon, 2008). Hearing white noise bursts at the same time during their development does not elicit pecking behavior. And baby ducklings, who preferentially approach mature members of their own species, will preferentially approach an audio speaker that is playing their conspecific calls over another audio speaker that is playing the calls of a different species – unless the other species’ calls are very similar in frequency and timing characteristics to the ducklings’ own species’ calls (Gottlieb, 1997). This suggests that, by birth, many animals have already developed broad templates for recognizing the vocalizations of their own species. Animals don’t necessarily recognize the functional benefit of responding to their own species’ calls, but their perceptual system has evolved to prefer and respond to these sounds.

Like other animals, human infants show early-emerging perceptual biases that allow them to recognize and orient to the vocalizations of their conspecifics. From birth, humans differentiate between different kinds of sounds and show systematic biases for responding preferentially to some sounds over others. These early listening preferences may emerge for different reasons: based on auditory experiences in utero or more abstract experience-independent factors. Infants may show these increased responses in their brain responses, their physiological reactions, and how they control their own behavior to elicit the presentation of different sounds.

Some newborn listening preferences are clearly rooted in prenatal experience. For example, newborns prefer their mother’s voice, their native language, and soap opera theme songs, based on hearing these sounds before birth (Byers-Heinlein, Burns, & Werker, 2010; DeCasper & Fifer, 1980; Hepper, 1988; Mehler et al., 1988). Some newborn listening preferences appear rooted in the structure of the auditory system. For example, newborns’ muscular activity increased the most to acoustic information clustered around 125–250 Hz, including sine wave tones, square wave tones, and naturally occurring human speech (Hutt, Hutt, Lenard, van Bernuth, & Muntjewerff, 1968).

Other listening preferences may be more abstract, and not based on infants’ specific prenatal experiences. Human infants prefer listening to speech relative to a wide range of sounds, including warbled tones (Samples & Franklin, 1978), white noise (Butterfield & Siperstein,1970; Colombo & Bundy, 1981), filtered speech (Spence & DeCasper, 1987), and melodic non-speech analogues of speech (Vouloumanos & Werker, 2004, 2007a). This last comparison is particularly noteworthy as speech was compared to a melodic non-speech counterpart consisting of sine wave analogues closely matched to the fundamental frequency and first three formants of natural speech (Vouloumanos & Werker, 2007a). Newborns modulated their high amplitude sucking responses to elicit more presentations of speech compared with melodic non-speech counterparts. This experimental paradigm allows infants to control the presentation of the sounds they hear by their sucking strength on a non-nutritive pacifier – making it analogous to infant-controlled procedures used with older infants and making it a strong comparative test. Infants may exercise deliberate control over their high-amplitude sucking, or infants may simply be aroused by the sounds they hear and suck harder because of this arousal. Whether cognitive or arousal based, within 48 hours of birth, newborns change their sucking behavior to elicit the presentation of speech more than non-speech. This listening preference may even be independent of prenatal exposure because when speech and non-speech sounds are filtered to preserve only the lower frequencies that are audible in the womb prenatally (Griffiths, Brown, Gerhardt, Abrams, & Morris, 1994), newborns suck equally for filtered speech and filtered non-speech (Vouloumanos & Werker, 2007b). This finding suggests that the relevant information for differentiating speech and non-speech analogues – and thus eliciting preference – in infants after birth is not present in the low frequencies that permeate the womb. Thus, part of the basis for this speech bias might be independent of specific prenatal experience, and may be triggered by a brief amount of experience after birth.

DEVELOPMENT: SHIFTING SPEECH PREFERENCES

Although newborns prefer to listen to speech over other sounds, like white noise or melodic non-speech analogues, they don’t prefer speech compared to all sounds. For example, newborns listen equally to speech and calls of rhesus monkeys (Vouloumanos, Hauser, Werker, & Martin, 2010). By 3 months, infants prefer speech compared to rhesus monkey calls (Shultz & Vouloumanos, 2010). Newborns may listen to speech and rhesus calls equally because of the acoustic properties they share from being biological or primate-produced vocalizations. Younger infants might find the harmonic structure that is available in both sounds to be equally interesting. The human auditory system may have evolved, similarly to that of ducklings or songbirds, with a general template for frequency and timing characteristics that is shared by both human and rhesus vocalizations.

Some listening preferences for speech that newborns show during the first year of life remain stable, for example newborns’ preference for speech over melodic non-speech analogues of speech appears stable in the first year of life: it is observed at birth, 2 months, 4 months, 6 months, and 12 months (Vouloumanos & Curtin, 2014; Vouloumanos & Werker, 2004, 2007a). Such stable preferences might reflect evolutionarily derived general tuning of the auditory system to particular acoustic information in environment-specific frequency bands or timing intervals.

Other preferences for speech change over the first year of life. For example, typically developing infants show no preference for speech over rhesus monkey calls at birth, but prefer speech at 3 months, and prefer rhesus monkey calls at 9 months (Sorcinelli, Ference, Curtin, & Vouloumanos, 2019). This changing pattern might tap into infants’ ability to assign value to different acoustic signals. At birth, newborns orient to the shared acoustic characteristics of both species’ vocalizations. At 3 months, infants who are figuring out their native language sounds and sound structures may listen to speech more to better analyze it, and less to monkey calls because monkey calls are novel and have no social context. At 9 months, single canonical native-like words might no longer require in-depth analysis, while the novelty of a rhesus monkey vocalization induces infants to listen more. The amount of information that infants may glean from different sounds may motivate these shifting listening preferences (Hunter & Ames, 1988; Kidd, Piantadosi, & Aslin, 2014).

FOUNDATIONS: BRAIN RESPONSES TO SPEECH

Systematic biases for speech might also be found in infant brains, specifically in brain areas responding preferentially to speech compared with non-speech sounds, similar to adults (e.g., Hickok & Poeppel, 2015; Vouloumanos, Kiehl, Werker, & Liddle, 2001). A small number of studies of infants’ neural responses support some early functional specialization for speech sounds and for specific acoustic properties of speech. Neuroimaging studies with newborns have used functional magnetic resonance imaging (fMRI) and near-infrared spectroscopy (NIRS), which attempt to localize active brain areas by measuring changes in local oxygenation of blood, and electroencephalography (EEG) and magnetoencephalography, which attempt to examine how and when the newborn brain processes speech sounds.

Neuroimaging studies with newborns thus far have found mixed and sometimes contradictory results for a neural specialization for speech. When speech was compared to music, there was weak evidence for a left hemisphere advantage for speech in newborns with a significant group-level difference for speech but a smaller difference to music (Kotilahti et al., 2010). When sounds with more comparable acoustic properties were used, some studies showed that native language speech activated the left hemisphere more than backward speech or silence using fMRI (Peña et al., 2003), and backward speech and non-native (foreign) speech using functional NIRS (Sato et al., 2011), while another showed no specialized response to forward speech compared with backward speech (both native and non-native) using low-pass-filtered speech (May, Byers-Heinlein, Gervain, & Werker, 2011). A more recent NIRS study compared...