演講 II：Phillip Endicott
27 MARCH 23
NCKU-EFEO Talk 專題演講
主講人：
Dr. Phillip Endicott（愛沙尼亞塔爾圖大學研究員︑巴黎國立自然歷史博物館）
講題：
Joint quantitative analysis of genes and language using Contact Graphs applied to Papuan and Austronesian speaking communities of East Timor
日期：
二〇二三年四月十二日（三）下午二時半
地點：
國立成功大學力行校區考古學研究所R201教室
摘要：
Accounts of human prehistory face two related challenges: ensuring scientific falsifiability and integrating different aspects of the past into a single history. Of the possible lines of evidence, language and genes offer the best prospect of synthesis because of their shared historical processes, which include vertical inheritance and horizontal transfer (that is, admixture in genetics, and borrowing in linguistics). Combining two different data sets, however, increases super exponentially the number of possible trees and many of the statistical models capable of testing hypotheses for the structure of a data set turn out to have intractable likelihood functions. Consequently, studies combining genetic and linguistic data are typically restricted to qualitative comparisons for independence between distance-based summary matrices.We propose a framework where languages and genes of the same populations are analysed jointly, constrained to share a history, or `contact’, graph describing the interactions between their ‘carrier’ societies within a shared scaffolding. In contrast to phylogenetic trees, contact graphs provide a topology that can include mixtures between different language families, and multiple genetic populations, to model detailed, local histories. While our focus is on current genetic and linguistic diversity, the approach can facilitate the inclusion of human ancient DNA and archaeological data. This provides a first step towards a statistical framework towards joint analysis, in order to advance beyond a narrative account of the past and have statistically testable and, therefore, falsifiable hypotheses.Our study region is east of the Wallace line in Island Southeast Asia, from the Lesser Sundas to West Papua. In this pilot study, we look at the zone of ongoing contact between patterns of structurally dissimilar genetic and linguistic ancestries in East Timor, where widespread language shift/replacement has occurred towards both Papuan as well as Austronesian. Our Contact Graph for East Timor, learned using a prototype of our proposed method, is based on all data (Genetics: SNPs and Haplotypes; Linguistics: Lexical, and Phonetic) for five Austronesian (Kemak, Mambae, Tetun, Tokodede, Waima’a) and three Papuan (Bunak, Fataluku, Makasae) speaking communities. We describe proportions of transfer inferred for the minimum number of contact events required to explain the difference between the predicted structure, given the graph, and that of the actual data.
此演講由國立成功大學考古學研究所所長趙金勇教授、法國遠東學院台北中心主任梅豪方教授生共同主持。
此為英文演講。自由入座，無須報名。
演講
主講人：
Dr. Phillip Endicott（愛沙尼亞塔爾圖大學研究員︑巴黎國立自然歷史博物館）
講題：
Joint quantitative analysis of genes and language using Contact Graphs applied to Papuan and Austronesian speaking communities of East Timor
日期：
二〇二三年四月十二日（三）下午二時半
地點：
國立成功大學力行校區考古學研究所R201教室
摘要：
Accounts of human prehistory face two related challenges: ensuring scientific falsifiability and integrating different aspects of the past into a single history. Of the possible lines of evidence, language and genes offer the best prospect of synthesis because of their shared historical processes, which include vertical inheritance and horizontal transfer (that is, admixture in genetics, and borrowing in linguistics). Combining two different data sets, however, increases super exponentially the number of possible trees and many of the statistical models capable of testing hypotheses for the structure of a data set turn out to have intractable likelihood functions. Consequently, studies combining genetic and linguistic data are typically restricted to qualitative comparisons for independence between distance-based summary matrices.We propose a framework where languages and genes of the same populations are analysed jointly, constrained to share a history, or `contact’, graph describing the interactions between their ‘carrier’ societies within a shared scaffolding. In contrast to phylogenetic trees, contact graphs provide a topology that can include mixtures between different language families, and multiple genetic populations, to model detailed, local histories. While our focus is on current genetic and linguistic diversity, the approach can facilitate the inclusion of human ancient DNA and archaeological data. This provides a first step towards a statistical framework towards joint analysis, in order to advance beyond a narrative account of the past and have statistically testable and, therefore, falsifiable hypotheses.Our study region is east of the Wallace line in Island Southeast Asia, from the Lesser Sundas to West Papua. In this pilot study, we look at the zone of ongoing contact between patterns of structurally dissimilar genetic and linguistic ancestries in East Timor, where widespread language shift/replacement has occurred towards both Papuan as well as Austronesian. Our Contact Graph for East Timor, learned using a prototype of our proposed method, is based on all data (Genetics: SNPs and Haplotypes; Linguistics: Lexical, and Phonetic) for five Austronesian (Kemak, Mambae, Tetun, Tokodede, Waima’a) and three Papuan (Bunak, Fataluku, Makasae) speaking communities. We describe proportions of transfer inferred for the minimum number of contact events required to explain the difference between the predicted structure, given the graph, and that of the actual data.
此演講由國立成功大學考古學研究所所長趙金勇教授、法國遠東學院台北中心主任梅豪方教授生共同主持。
此為英文演講。自由入座，無須報名。
演講