Viki’s First Words: A Comparative Phonetics Case Study

It remains an open question whether nonhuman great apes are capable of vocal learning—modification of uttered signals from experience with other individuals. Evidence of such capacities would provide insight into when human imitative abilities—crucial for spoken language—first emerged. However, an important case study with bearing on this topic has been neglected in the literature—that of Viki the chimpanzee (Pan troglodytes). Only a few days old, Viki was adopted by Keith and Catherine Hayes of the Yerkes Laboratory of Primate Biology and raised in a human home. Whilst in the Hayeses’ care, Viki became an early case in a series of attempts by scientists to teach human language to nonhuman great apes. Ultimately, Viki learned to consistently produce four words: “mama,” “papa,” “cup,” and “up” (Hayes & Hayes, 1951) and was claimed to babble like a human child. This would suggest that chimpanzees are capable of vocal learning. It is surprising, then, that no phonetic analysis of recordings of Viki’s speechlike behavior has ever been presented. Providing such an analysis is the purpose of this brief report and case study. For the study, I curated audio from a segment of the Television movie The Alphabet Conspiracy (Warner Bros., 1959). To my knowledge, this film contains the only widely available recording of Viki’s speechlike utterances. Across the recording, I identified three pulses of supposed “babble,” three utterances of “papa,” and a single utterance of “cup,” which were all analyzed by way of spectrograms. I was unable to identify the “up” and “mama” utterances purported by the Hayeses (Hayes & Hayes, 1951). Because of genetic relatedness between chimpanzees and modern humans, I assumed that human approximation of speech-like sounds uttered by a chimpanzee should recruit much of the same articulatory organs and mechanisms. Thus, to compare Viki’s speechlike utterances with human phonemes, I, a male speaker aged 26, made recordings imitating Viki’s speechlike production. This was done to elucidate potential

limitations on Viki's articulation. During recordings, I listened to each of Viki's utterances and vocally approximated the sound heard. Each of Viki's utterances was imitated three times and the approximation closest to that of Viki (to the ear) was rendered and analysed. All sounds were rendered as high-pass filter spectrograms (300 Hz) in the Sopran software (Svante Granqvist, Karolinska Institute) (tolvan.com) ( Table I).

Babble
Human infants typically begin exhibiting babbling behavior around the age of 6 months. In developmental sciences, babble is defined as syllabic prespeech behavior and is widely believed to lay the groundwork for speech proper. If indeed chimpanzees were capable of vocal learning in the wild (Crockford et al., 2004), it is conceivable that infant chimpanzees should occasionally babble also (but see Kojima, 2001). The sequence of supposed babble by Viki consists of three pulses in quick succession, which I realized as glottal stops [ʔ] -formed via airstream release following complete occlusion by the glottis ("uhoh"). However, Viki's so-called babbling does not exhibit the telltale acoustic features of either syllabic human speech (MacNeilage, 1998), nor does it resemble her other speechlike utterances. Given its singular occurrence, and the controversial nature of the claim, Viki's supposed babbling is more parsimoniously interpreted as playful grunting.

Papa
Viki's /papa/ lacks any apparent vowel-like features, which are readily identifiable by their arrangement as broad bands of concentrated energy. Instead, utterances are arranged as sequences of two clicks or bursts (consistent with plosive consonants). However, the acoustic profiles of segments were less consistent with bilabial plosive [p] ("pot"), and more with tenuis (voiceless, unaspirated, unglottalized obstruents) bilabial clicks [ʘ] (rare in natural languages, occurring only in the Tuu and Kx'a language families of southern Africa).
Crucially, across all iterations of /papa/, the spectrographic profile of the utterance is well maintained, indeed suggesting a learned motor sequence (i.e., a word). Importantly, consistent execution of the corresponding speech gestural score indicates that Viki had indeed acquired a novel voiceless call. Because vocal learning typically refers to voiced sounds, such learned gesture may be tentatively described as nonphonatory vocal learning, reflecting the lack of laryngeal involvement.

Cup
As with, /papa/, there is no apparent vowel-like articulation following /c/ in Viki's /cup/. Furthermore, the spectrographic profile of the /c/ in cup was not consistent with voiceless velar plosive [k] but more so with my production of voiceless velar fricative [x] (German: "Bach"). In humans, velar consonants are executed via the dorsal tongue against the velum-but [x] is executed posteriorly in comparison. Viki's potential realization of [x] (a fricative) as opposed to [k] (a plosive)

3
Viki's First Words: A Comparative Phonetics Case Study

Viki
Human may be suggestive of limitations of the chimpanzee tongue, which is longer and flat in comparison with humans, with fewer articulatory degrees of freedom inside the oral cavity accordingly (Takemoto, 2008). However, the interpretation of Viki's /c/ as [x] should be considered tentative, as intra-oral articulatory gestures actively employed by chimpanzees in call production remain poorly understood. The /p/ in "cup" appears similarly to the /p/ in Viki's "papa"-but with an evidently weaker release (diminished in intensity). This is consistent with the interpretation of the /p/ as [p]. I also realized it as such.
The . This indicates that (1) articulatory capacities of chimpanzees may ultimately be defined by species-typical tongue dimensions and degrees of freedom (Takemoto, 2008), and (2) at least in rare occurrences, chimpanzees may be capable of nonphonatory or voiceless vocal learning. While Viki was seemingly incapable of simultaneous recruitment of consonantal "frames" and vowel-like "content" (MacNeilage, 1998), she successfully produced a small sample of humanlike consonantal speech sounds, including labial and, possibly, velar articulations.