Phylogenetic signal in the lexicon: Are parental terms influenced by baby talk?

Words for one’s parents have recurring (possibly non-random) features across the world. Jakobson (1963), Murdock (1959), and numerous works since have noted the probable grounding of parental terms in child language speech. Parental terms, unlike some other kinship terms, are universally lexicalized and acquired early by children across the world. They disproportionately tend to contain highly stable segments (labial and apical stops); all this suggests that parental terms may, despite their universality and semantic stability, not be phylogenetically stable if measured by lexicon.

In this work, I present a phylogenetic approach to these questions by studying the lexical evolution of words for ‘mother’ and ‘father’ across three large language phylogenies: Pama-Nyungan (Australia; Bouckaert, Bowern, and Atkinson 2018), Austronesian (Greenhill et al. 2009), and Indo-European (Grey and Atkinson 2003). All families exhibit multiple lexical cognates for parental terms, and as Murdock (1959) noted, there is a strong skew to the phonemes found in such words. This paper presents a number of results regarding the phylogenetic signal found in this domain.

Transcript

1. Phylogenetic signal in the lexicon: Are parental terms influenced by baby talk? Claire Bowern • ICHL, Canberra, 2019 Work in progress
3. “Mama”/“Papa”: Stable Words or Baby Talk? • Case study of phylogenetic signal and evolution of terms for parents (i.e. ‘mother’ and ‘father’) • Quantitative evaluation of claims • “mama/papa/dada” and their effect on reference terms ◦ cf. Jakobson 1962, Murdock 1959
4. Parental Terms: • universally lexicalized • acquired early • tend to contain phonologically stable segments (e.g. fewer affricates, ejectives, etc) • influenced by baby talk phonology • replaced by baby talk forms • address vs reference forms • register variation < more phyl’ic signal more random variation>
5. Today: Aims • Illustration of quantitative approach to old question in historical linguistics • Contribution to understanding evolution of lexicon [small slice!] • Investigating how arbitrary non-arbitrary can be • Better understanding relationships between constraints on lexical transmission and phylogenetic signal • Case study of ‘process’ in micro and its outcomes in macro-change
6. Data and Methods
7. 3 phylogenies • Pama-Nyungan [243 langs] • Austronesian [369 langs] • Indo-European [100 langs] • Translation equivalents for ‘mother’ and ‘father’ • c 700 languages in sample • families represent c. 30% of the world’s languages
8. Coding • 2 meaning classes (‘mother’ and ‘father’ • Each cognate coded as ‘state’ • States converted to binary characters (presence/absence of form X in meaning Y) ‘mother’ *ŋama *ŋapaŋ *ŋaŋka Djinang ama 1 0 0 Kungardutyi ŋama 1 0 0 Batyala 0 ŋavaŋ 1 0 Waka-Kabi 0 ŋapaŋ 1 0 Panyjima 0 0 ŋaŋka 1 Kurrama 0 0 ŋaŋka 1 Martuthunira 0 0 ŋaŋka 1
9. Methods: suite of investigation 1. PCA (Principle components analysis) – mother/father in the context of basic vocabulary: which meaning classes behave like each other in the data? Are mother/father like each other? Like ‘high cognacy’ or ‘low cognacy’ meanings? [Pama-Nyungan only] 2. Phylogenetic signal: D-statistics and measuring conservatism/phylogenetic signal [All families] 3. Phonology: Apicals and Labials: do words with ‘baby talk’ features (e.g. reduplication) behave differently from words with other consonants? Are words with baby talk consonants disproportionately frequent in parental terms? Do such words show different characteristics? [All families]
10. D statistic • sister languages & sister nodes within the tree • observed distinctness from one another • vs expected distinctness for evolution under Brownian motion • compared against two null hypotheses: ◦ Brownian evolution – random evolution down tree ◦ un-tree-like randomness – random distribution along tips
11. D statistic Random distribution at tips Brownian evolution
12. D Statistic Fritz & Purvis (2010)
13. Results
14. PCA
15. PCA • Max.class.size: number of langs in largest cog set [measure of retention] • Number of singletons/number of cog classes per meaning: [measure of innovation] • Missing: [utility for phylogenetic comparison] • Mother/father are similar • Mother/father are conservative, but not the most conservative • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 1dl.excl 1dl.incl 1pl.excl 1pl.incl 1sg 2pl 2sg 3dl 3pl 3sg above again all alone and angry ankle ant armpit ashamed ashes ask at baby back backbone bad bag bandicoot bat be.alive beard belly below big bird bite black black.cockatoo black.duck blind blood Bloodwood blow blowfly Blue−Tongue.Lizard blunt body bone boomerang boy brain break breast breathe brother.in.law burn butterfly buttocks calf camp canoe Carpet.snake carry cave Centipede charcoal cheek chest chew child chin climb cloud cold come cook corroboree cough count creek crooked Crow cry curlew cut dark daughter day Death.adder die dig digging.stick dingo dirty dog dream drink dry Duck dust eaglehawk ear earth east eat Echidna egg elbow Emu enter eye eyebrow face faeces fall far fat father fathers.father fathers.mother fathers.sister fear feather fighting.stick find fingernail fire fish flow flower fly Flying.Fox fog foot forehead friend fruit ghost girl give goanna good grass grasshopper greedy grow hair hand hard head hear heart heat heavy heel hide hill hip hit hit.w.hand hold honey house how hungry husband hut if inside intestines itch jaw jump kangaroo kidney kill kiss knee know lake language laugh leaf left leg lie.down lightning lip liver long long.ago louse lung maggot Magpie man many marrow meat milk moon morning mosquito mother mothers.brother mothers.father mothers.mother mouth mud name nape native.cat navel near neck Nest new night no north nose old old.man old.woman older.brother older.sister one open other painful.sore penis person pimple pregnant pull quiet rain rainbow raw red rib right river road roof root rope rotten run salt.water sand sandhill say Scorpion scratch search see shade sharp shin shoot short shoulder sick sit skin sky sleep slow small smell smoke snake soft soon south speak spear spearthrower spider spirit spit split squeeze stab stand star steal stick stone stone.axe string strong suck sun swallow swan sweat sweet swell swim tail take termite testicles that there thick thigh thin think this three throat throw thumb thunder tie.up tired tobacco tongue tooth tree tree.bark true turn turtle two urinate urine vegetable.food vomit walk wallaby water west wet what when where whistle white who wife Willy.Wagtail wind wing winnow Witchetty.Grub woman wood work Yam yawn yellow yes younger.brother younger.sister singletons max.class.size number.cognates missing −6 −4 −2 0 2 −2 0 2 4 Dim1 (62.4%) Dim2 (30.4%) 0.25 0.50 0.75 cos2 PCA − Biplot
16. Phylogenetic Signal Proliferation of Terms
17. Number of forms • IE and Aus’n have terms that dominate; PNy doesn’t • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • 0.0 0.2 0.4 0.6 Aus IE Pny Family Percentage of Languages Parent • • Father Mother Cognate class size as percentage of languages in family *tama *ʔina *peh2 -ter *meh2 -ter *ŋama(tyi) *mama
18. Mother • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • papa tyapu ngapiRi piya(ng) pi:pa ngatyi mama • • • • • • • • • • • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • ngama(tyi) yanga(na nga:ndi ka:tya papan yaku Father
19. Phylogenetic Signal (D statistics)
20. Overall Results • D Statistics are mostly negative [ie strong Phylogenetic Signal] • Averages: • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • −4 0 4 Father Mother Parent Dstat Family • • • Aus IE Pny Father Mother Aus -0.909 -0.439 IE -1.634 -1.108 Pny -0.972 -1.157 Overall -1.041 -0.785
21. Any significant differences between families? • No. • (linear regression with family, term, and their interaction as variables) • [Only significant difference is between terms] • • • • • • • −4 0 4 Aus IE Pny Family Dstat Parent Father Mother
22. Compared to other vocabulary? [PNy only] • All parts of speech show variation (borrowing, inheritance, etc) • Means all similar • Mother/Father DStats show same type of range. • • • • • • • • −2 0 2 Adj Adj/Verb Adpos Adv Grammar Noun Verb PoS Dstat MF Other Parent
23. Phonology
24. Are D statistics higher for baby-talk forms? • • • • • −4 0 4 Father Mother Parent Dstat Apical no yes • • • • • • • • −4 0 4 Father Mother Parent Dstat Labial no yes Labials Apicals
25. Reduplicated Terms • Higher average D statistic • But not a significant difference • • • • • • • • • −4 0 4 Father Mother Parent Dstat Redup no yes
26. Are baby-talk consonants disproportionately frequent in parental terms? • Chi-Square test [test of fit between observed and expected distributions of segments to test for interactions between variables] • Tested for Pama-Nyungan only, against other kin terms [MM, MF, B, Z, BW, MB, FZ, ZH] Reduplicated Not.Reduplicated Parental.Term a b Not.Parental c d
27. Disproportionate baby-talk forms? Redup Labial Apical Palatal Parents ✕ ✕ ✓ ✕ G’parents ✓ ✕ ✕ ✕ Siblings ✕ ✕ ✕ ✕
28. Conclusions
29. Conclusions • Findings: ◦ Terms show phylogenetic signal. ◦ The “baby talk” terms show less. ◦ But not significantly so. • The results are consistent across language families, though the terms are less conservative in Pama-Nyungan. • There is probably an effect of child acquisition, but not enough to swamp regular transmission.
30. Thank you! • Traditional owners of Ngunnawal Country • Eliza Scruton (initial data coding) • Simon Greenhill and ABVD • IELex • Pama-Nyungan Lab @ Yale