Also known as likelihood-free methods, approximate Bayesian computational (ABC) methods have appeared in the past ten years as the most satisfactory approach to intractable likelihood problems, first in genetics then in a broader spectrum of applications. However, these methods suffer to some degree from calibration difficulties that make them rather volatile in their implementation and thus render them suspicious to the users of more traditional Monte Carlo methods. In this survey, we study the various improvements and extensions brought on the original ABC algorithm in recent years.

The Sino-Tibetan language family is one of the world’s largest and most prominent families, spoken by nearly 1.4 billion people. Despite the importance of the Sino-Tibetan languages, their prehistory remains controversial, with ongoing debate about when and where they originated. To shed light on this debate we develop a database of comparative linguistic data, and apply the linguistic comparative method to identify sound correspondences and establish cognates. We then use phylogenetic methods to infer the relationships among these languages and estimate the age of their origin and homeland. Our findings point to Sino-Tibetan originating with north Chinese millet farmers around 7200 B.P. and suggest a link to the late Cishan and the early Yangshao cultures.

International audienceWe develop a formal semantic analysis of the alarm calls used by Campbell’s monkeys in the Tai forest (Ivory Coast) and on Tiwai island (Sierra Leone)—two sites that differ in the main predators that the monkeys are exposed to (eagles on Tiwai vs. eagles and leopards in Tai). Building on data discussed in Ouattara et al. (PLoS ONE 4(11):e7808, 2009a; PNAS 106(51): 22026–22031, 2009b and Arnold et al. (Population differences in wild Campbell’s monkeys alarm call use, 2013), we argue that on both sites alarm calls include the roots krak and hok, which can optionally be affixed with -oo, a kind of attenuating suffix; in addition, sentences can start with boom boom, which indicates that the context is not one of predation. In line with Arnold et al., we show that the meaning of the roots is not quite the same in Tai and on Tiwai: krak often functions as a leopard alarm call in Tai, but as a general alarm call on Tiwai. We develop models based on a compositional semantics in which concatenation is interpreted as conjunction, roots have lexical meanings, -oo is an attenuating suffix, and an all-purpose alarm parameter is raised with each individual call. The first model accounts for the difference between Tai and Tiwai by way of different lexical entries for krak. The second model gives the same underspecified entry to krak in both locations (= general alarm call), but it makes use of a competition mechanism akin to scalar implicatures. In Tai, strengthening yields a meaning equivalent to non-aerial dangerous predator and turns out to single out leopards. On Tiwai, strengthening yields a nearly contradictory meaning due to the absence of ground predators, and only the unstrengthened meaning is used

We argue that rich data gathered in experimental primatology in the last 40 years can benefit from analytical methods used in contemporary linguistics. Focusing on the syntactic and especially semantic side, we suggest that these (iii) how are the meanings of individual calls combined? (iv) how do calls or call sequences compete with each other when several are appropriate in a given situation? (v) how did the form and meaning of calls evolve? We address these questions in five case studies pertaining to cercopithecines (Putty-nosed monkeys, Blue monkeys, and Campbell's monkeys), colobinae (Guereza monkeys and King Colobus monkeys), and New World monkeys (Titi monkeys). The morphology mostly involves simple calls, but in at least one case (Campbell's -oo) we find a root-suffix structure, possibly with a compositional semantics. The syntax is in all clear cases simple and finite-state. With respect to meaning, nearly all cases of call concatenation can be analyzed as conjunction. But a key question concerns the division of labor between semantics, pragmatics and the environmental context ('world' knowledge and context change). An apparent case of dialectal variation in the semantics (Campbell's krak) can arguably be analyzed away if one posits sufficiently powerful mechanisms of competition among calls, akin to scalar implicatures. An apparent case of non-compositionality (Putty-nosed pyow-hack sequences) can be analyzed away if one further posits a pragmatic principle of 'urgency', whereby threat-related calls must come early in sequences (another potential case of non-compositionality -Colobus snort-roar sequences -might justify assigning non-compositional meanings to complex calls, but results are tentative). Finally, rich Titi sequences in which two calls are re-arranged in complex ways so as to reflect information about both predator identity and location are argued not to involve a complex syntax/semantics interface, but rather a finegrained interaction between simple call meanings and the environmental context. With respect to call evolution, we suggest that the remarkable preservation of call form and function over millions of years should make it possible to lay the groundwork for an evolutionary monkey linguistics, which we illustrate with cercopithecine booms, and with a comparative analysis of Blue monkey and Putty-nosed monkey repertoires. Throughout, we aim to compare possible theories rather than to fully adjudicate between them, and our claims are correspondingly modest. But we hope that our methods could lay the groundwork for a formal monkey linguistics combining data from primatology with formal techniques from linguistics (from which it does not follow that the calls under study share non-trivial properties, let alone an evolutionary history, with human language).

We explain why general techniques from formal linguistics can and should be applied to the analysis of monkey communication -in the areas of syntax and especially semantics. An informed look at our recent proposals shows that such techniques needn't rely excessively on categories of human Theoretical Linguistics 2016; 42(1-2): 173-201 Brought to you by | Utrecht University Library Authenticated Download Date | 11/24/16 3:04 PM language: syntax and semantics provide versatile formal tools that go beyond the specificities of human linguistics. We argue that "formal monkey linguistics" can yield new insights into monkey morphology, syntax, and semantics, as well as raise provocative new questions about the existence of a pragmatic, competition-based component in these communication systems. Finally, we argue that evolutionary questions, which are highly speculative in human language, can be addressed in an empirically satisfying fashion in primate linguistics, and we lay out problems that should be addressed at the interface between evolutionary primate linguistics and formal analyses of language evolution.

Abstract. Cäsar et al. 2013 show that the structure of Titi monkey call sequences can, with just two call types (A and B), reflect information about predator type and predator location. Using the general methods of Schlenker et al. 2014, to appear, a, b, we ask what these observations show about the 'linguistic' structure of Titi calls. We first demonstrate that the simplest behavioral assumptions make it challenging to provide lexical specifications for A-and B-calls: B-calls rather clearly have the distribution of highly underspecified calls; but A-calls are also found in highly heterogeneous contexts (e.g. they are triggered by 'cat in the canopy' and 'raptor on the ground' situations). We discuss two possible solutions to the problem. One posits that entire sequences are endowed with meanings that are not compositionally derived from their individual parts (a related idea was proposed by Arnold and Zuberbühler to analyze pyow-hack sequences in Putty-nosed monkeys). The second solution, which we consider to be superior, takes sequences to have no structure besides concatenation: the Bcall is a general call, the A-call is used for serious non-ground threats, and each call reflects information about the environment at the time at which it is uttered. The composition of Cäsar et al.'s sequences is seen to follow from the interaction between call meaning, rules of competition among calls, and more sophisticated assumptions about the environmental context. In the end, a detailed analysis of the division of labor between semantics, pragmatics and the environmental context yields a simple and explanatory analysis of sequences that initially seemed to display a complex mapping between syntax and semantics. * We are very grateful to three anonymous referees and to Editor Ad Neeleman for very helpful comments and Grant acknowledgments: Cäsar: The research leading to these results received funding from the CAPES-Brazil, FAPEMIG-Brazil, S.B. LEAKEY TRUST and the University of St Andrews.

22 pagesInternational audienceNicholls and Gray have described a phylogenetic model for trait data. They used their model to estimate branching times on Indo-European language trees from lexical data. Alekseyenko and co-workers extended the model and gave applications in genetics. We extend the inference to handle data missing at random. When trait data are gathered, traits are thinned in a way that depends on both the trait and the missing data content. Nicholls and Gray treated missing records as absent traits. Hittite has 12% missing trait records. Its age is poorly predicted in their cross-validation. Our prediction is consistent with the historical record. Nicholls and Gray dropped seven languages with too much missing data. We fit all 24 languages in the lexical data of Ringe and co-workers. To model spatiotemporal rate heterogeneity we add a catastrophe process to the model. When a language passes through a catastrophe, many traits change at the same time. We fit the full model in a Bayesian setting, via Markov chain Monte Carlo sampling. We validate our fit by using Bayes factors to test known age constraints. We reject three of 30 historically attested constraints. Our main result is a unimodal posterior distribution for the age of Proto-Indo-European centred at 8400 years before Present with 95% highest posterior density interval equal to 7100-9800 years before Present

Approximate Bayesian computation methods are useful for generative models with intractable likelihoods. These methods are however sensitive to the dimension of the parameter space, requiring exponentially increasing resources as this dimension grows. To tackle this difficulty, we explore a Gibbs version of the Approximate Bayesian computation approach that runs component-wise approximate Bayesian computation steps aimed at the corresponding conditional posterior distributions, and based on summary statistics of reduced dimensions. While lacking the standard justifications for the Gibbs sampler, the resulting Markov chain is shown to converge in distribution under some partial independence conditions. The associated stationary distribution can further be shown to be close to the true posterior distribution and some hierarchical versions of the proposed mechanism enjoy a closed form limiting distribution. Experiments also demonstrate the gain in efficiency brought by the Gibbs version over the standard solution.