Sean G. Roberts

Universal Principles in the Repair of Communication Problems.

There would be little adaptive value in a complex communication system like human language if there were no ways to detect and correct problems. A systematic comparison of conversation in a broad sample of the world’s languages reveals a universal system for the real-time resolution of frequent breakdowns in communication. In a sample of 12 languages of 8 language families of varied typological profiles we find a system of ‘other-initiated repair’, where the recipient of an unclear message can signal trouble and the sender can repair the original message. We find that this system is frequently used (on average about once per 1.4 minutes in any language), and that it has detailed common properties, contrary to assumptions of radical cultural variation. Unrelated languages share the same three functionally distinct types of repair initiator for signalling problems and use them in the same kinds of contexts. People prefer to choose the type that is the most specific possible, a principle that minimizes cost both for the sender being asked to fix the problem and for the dyad as a social unit. Disruption to the conversation is kept to a minimum, with the two-utterance repair sequence being on average no longer that the single utterance which is being fixed. The findings, controlled for historical relationships, situation types and other dependencies, reveal the fundamentally cooperative nature of human communication and offer support for the pragmatic universals hypothesis: while languages may vary in the organization of grammar and meaning, key systems of language use may be largely similar across cultural groups. They also provide a fresh perspective on controversies about the core properties of language, by revealing a common infrastructure for social interaction which may be the universal bedrock upon which linguistic diversity rests.

Linguistic Diversity and Traffic Accidents: Lessons from Statistical Studies of Cultural Traits

The recent proliferation of digital databases of cultural and linguistic data, together with new statistical techniques becoming available has lead to a rise in so-called nomothetic studies [1]–[8]. These seek relationships between demographic variables and cultural traits from large, cross-cultural datasets. The insights from these studies are important for understanding how cultural traits evolve. While these studies are fascinating and are good at generating testable hypotheses, they may underestimate the probability of finding spurious correlations between cultural traits. Here we show that this kind of approach can find links between such unlikely cultural traits as traffic accidents, levels of extra-martial sex, political collectivism and linguistic diversity. This suggests that spurious correlations, due to historical descent, geographic diffusion or increased noise-to-signal ratios in large datasets, are much more likely than some studies admit. We suggest some criteria for the evaluation of nomothetic studies and some practical solutions to the problems. Since some of these studies are receiving media attention without a widespread understanding of the complexities of the issue, there is a risk that poorly controlled studies could affect policy. We hope to contribute towards a general skepticism for correlational studies by demonstrating the ease of finding apparently rigorous correlations between cultural traits. Despite this, we see well-controlled nomothetic studies as useful tools for the development of theories.

Climate, vocal folds, and tonal languages: Connecting the physiological and geographic dots

Significance The sound systems of human languages are not generally thought to be ecologically adaptive. We offer the most extensive evidence to date that such systems are in fact adaptive and can be influenced, at least in some respects, by climatic factors. Based on a survey of laryngology data demonstrating the deleterious effects of aridity on vocal cord movement, we predict that complex tone patterns should be relatively unlikely to evolve in arid climates. This prediction is supported by careful statistical sampling of climatic and phonological data pertaining to over half of the world’s languages. We conclude that human sound systems, like those of some other species, are influenced by environmental variables. We summarize a number of findings in laryngology demonstrating that perturbations of phonation, including increased jitter and shimmer, are associated with desiccated ambient air. We predict that, given the relative imprecision of vocal fold vibration in desiccated versus humid contexts, arid and cold ecologies should be less amenable, when contrasted to warm and humid ecologies, to the development of languages with phonemic tone, especially complex tone. This prediction is supported by data from two large independently coded databases representing 3,700+ languages. Languages with complex tonality have generally not developed in very cold or otherwise desiccated climates, in accordance with the physiologically based predictions. The predicted global geographic–linguistic association is shown to operate within continents, within major language families, and across language isolates. Our results offer evidence that human sound systems are influenced by environmental factors.

Commentary: Large-scale psychological differences within China explained by rice vs. wheat agriculture

Talhelm et al. (2014) test the hypothesis that activities which require more intensive collaboration foster more collectivist cultures. They demonstrate that a measure of collectivism correlates with the proportion of cultivated land devoted to rice paddies, which require more work to grow and maintain than other grains. The data come from individual measures of provinces in China. While the data is analyzed carefully, one aspect that is not directly controlled for is the historical relations between these provinces. Spurious correlations can occur between cultural traits that are inherited from ancestor cultures or borrowed through contact, what is commonly known as Galtons problem (Roberts and Winters, 2013). Effectively, Talhelm et al. treat the measures of each province as independent samples, while in reality both farming practices (e.g., Renfrew, 1997; Diamond and Bellwood, 2003; Lee and Hasegawa, 2011) and cultural values (e.g., Currie et al., 2010; Bulbulia et al., 2013) can be inherited or borrowed. This means that the data may be composed of non-independent points, inflating the apparent correlation between rice growing and collectivism. The correlation between farming practices and collectivism may be robust, but this cannot be known without an empirical control for the relatedness of the samples. Talhelm et al. do discuss this problem in the supplementary materials of their paper. They acknowledge that a phylogenetic analysis could be used to control for relatedness, but that at the time of publication there were no genetic or linguistic trees of descent which are detailed enough to suit this purpose. In this commentary I would like to make two points. First, since the original publication, researchers have created new linguistic trees that can provide the needed resolution. For example, the Glottolog phylogeny (Hammarstrom et al., 2015) has at least three levels of classification for the relevant varieties, though this does not have branch lengths (see also “reference” trees produced in List et al., 2014). Another recently published phylogeny uses lexical data to construct a phylogenetic tree for many language varieties within China (List et al., 2014). In this commentary I use these lexical data to estimate cultural contact between different provinces, and test whether these measures explain variation in rice farming pracices. However, the second point is that Talhelm et al. focus on descent (vertical transmission), while it may be relevant to control for both descent and borrowing (horizontal transmission). In this case, all that is needed is some measure of cultural contact between groups, not necessarily a unified tree of descent. I use a second source of linguistic data to calculate simple distances between languages based directly on the lexicon. These distances reflect borrowing as well as descent. The historical relatedness between cultures can be measured through language. Similarities and differences between languages and varieties reflect the histories of the human populations that speak them (e.g., Gray et al., 2009). Indeed, several studies have investigated the co-diffusion of language and rice growing in Austroasiatic languages (e.g., Sidwell and Blench, 2011; van Driem, 2011). For example, languages that descend from common ancestors have vocabularies with common etymologies, although the words may have changed over time to become different. Thus, languages that have close historical links tend to have more similar vocabularies and languages that have been isolated from each other for longer tend to have more dissimilar vocabularies. This measure may be a proxy for the transmission of cultural practices or knowledge. Similarly, speakers borrow words from other languages and varieties meaning that varieties that are geographically close tend to be more similar. This may also be a proxy for the adoption of other cultural traits, such as farming practices (or, indeed, collectivism, though Minkov, 2012 argues that certain cultural values cannot be borrowed directly). This data could be used in two ways. First, the linguistic similarities between cultures could be used as a statistical control for non-independence in the synchronic data. The relationship between rice growing and collectivism would be robust if it persists when removing the variance explained by shared history. Secondly, one could use the linguistic data to look at diachronic changes. It is possible to reconstruct the historical contingencies between cultural groups using language data and phylogenetic techniques (e.g., Dunn et al., 2011). This involves reconstructing past changes as cultural groups come into contact and divide. Support for Talhelms theory would come from demonstrating that when the prevalence of rice growing increased, so did the estimated collectivism measure. The estimated historical relations could be integrated with what is known about the actual historical spread of farming practices (e.g., Kovach et al., 2007; Fuller, 2011; Dodson et al., 2013). This kind of quantitative study is beyond the scope of this commentary, but it is at least possible to test whether rice-growing exhibits historical contingencies. I use a comparison of the vocabularies of the languages spoken in each province as a proxy for historical relations. A list of vocabulary and cognate coding for basic concepts for (Han) varieties within China were taken from List et al. (2014) (the cognate coding was derived automatically, see List et al. for details). The average distance between varieties was computed as the number of shared cognates (words which are related by descent) between the two varieties (calculation done using LingPy, List and Moran, 2013). Varieties were associated with the provinces in which their data was collected. The distance between two provinces was calculated as the average of the distance between each pair of varieties within each province. In this way, a matrix of distances was produced which represented how dissimilar the vocabularies of each pair of provinces was. A second distance matrix was produced which represented the difference in the proportion of rice growing between each pair of provinces (taken from Talhelm et al.). Complete data was available for 40 language varieties from 18 provinces. The rice and linguistic distances were compared using a Mantel test (which compares the correlation between two matrices with the distribution of the correlation when one of the matrices is permuted). The difference in rice growing was significantly correlated with the linguistic distance measure (r = 0.27, p = 0.02, 10,000 permutations). That is, provinces which are more similar in the proportion of rice growing are also more similar in their vocabularies. The same test was done using linguistic data from the ASJP database (Wichmann et al., 2013). Languages with origins within China were identified from the Ethnologue catalog (Lewis et al., 2015) along with the provinces where each was mainly spoken (these include languages associated with many ethnic groups, including Han). The distances between vocabularies were computed as the normalized edit distances (LDND) used in Wichmann et al. (2011). Given two words from different varieties for the same concept, the edit distance is the number of changes it takes to convert one word to the other, normalized by maximum number of possible edits. Similar words have a lower distance. The distance between varieties was taken as the average distance between each word in the vocabulary. The LDND measure further normalizes the distance between varieties by eliminating meaning-specific variation (see Wichmann et al.). Distances between provinces were calculated as above. Data were available for 139 languages from 18 provinces. Again, the differences in proportion of rice growing correlated significantly with the difference in vocabulary (r = 0.38, p = 0.0002, 10,000 permutations). A tree of historical relations can be estimated from the historical distances using hierarchical clustering. Figure ​Figure11 shows this tree (from the first dataset) projected onto a map of China. The “rice–wheat” border from Talhelm et al. is highlighted. This separates the high rice production areas in the south from low rice production areas in the north. Its clear that the historical relations align with this border. Indeed, the root of the tree splits provinces in the north from those in the south and sub-branches of the tree spread east-to-west. Figure 1 A tree of historical relationships between language varieties constructed from hierarchical clustering of similarities in vocabulary. Provinces on the “rice–wheat” border are highlighted in gray (provinces south of the border produce ... These results suggest that the prevalence of rice growing is related to cultural contact. This means that a more careful consideration of historical relationships between provinces is warranted before the link between rice production and collectivism can be confirmed. One part of Talhelm et al.s study which may be more robust to the findings presented here are the results at the county-level for neighboring provinces. This analysis uses more fine-grained groups than the varieties used in this paper. However, cultural contact can also be assessed at the level of accent and dialect (e.g., Spruit et al., 2009). Further surveys similar to Talhelm et al.s may consider eliciting linguistic data from participants, as well as psychological or sociological data, with the aim of using them as controls for relatedness. In general, researchers should be wary of correlations that do not control for shared cultural history. On a more positive note, as an increasing amount of data becomes available, cultural and linguistic data can be used together with other sources to investigate human history.

The Interactive Origin of Iconicity

We investigate the emergence of iconicity, specifically a bouba-kiki effect in miniature artificial languages under different functional constraints: when the languages are reproduced and when they are used communicatively. We ran transmission chains of (a) participant dyads who played an interactive communicative game and (b) individual participants who played a matched learning game. An analysis of the languages over six generations in an iterated learning experiment revealed that in the Communication condition, but not in the Reproduction condition, words for spiky shapes tend to be rated by naive judges as more spiky than the words for round shapes. This suggests that iconicity may not only be the outcome of innovations introduced by individuals, but, crucially, the result of interlocutor negotiation of new communicative conventions. We interpret our results as an illustration of cultural evolution by random mutation and selection (as opposed to by guided variation).

The Effects of Generation Turnover and Interlocutor Negotiation on Linguistic Structure

Cultural transmission has a key impact on the evolution of linguistic structure (Kirby, Cornish & Smith, 2008). During transmission, new individuals learn the conventions of the language through observation and participation in communicative interactions in context. We present a series of experiments which disentangle the role of generation turnover, or new minds coming into the linguistic community (e.g. Kirby et al. 2008) from that of negotiation of shared form-meaning conventions during interlocutor interaction (studied in graphical, but not linguistic, systems by e.g. Fay et al. 2007, Galantucci 2005). In the experimental task, two interlocutors played a communicative cooperative game using an artificial miniature language. Players were trained on a random language, which they then used to ask each other for specific objects; they scored a point for each successful interaction. The language produced by one (randomly selected) participant was then used as training data for the following training-and-usage round. We manipulated generation turnover: in dyads, two participants played six training-and-usage rounds; in chains, six different participant pairs played one round each. We also manipulated negotiation. In the negotiation condition there were two human players; in the no-negotiation condition, one of the players was replaced with a simulated computer agent who had perfect memory (i.e. it always used the training language without errors, and never offered or adopted suggestions during usage). The training languages in this condition came from the human player. We examined the effects of generation turnover and negotiation on the systematic structure of the resulting languages (measured using the technique described in Kirby et al., 2008) and found a significant effect of both factors. Structure increased over the six rounds in chains, but not in dyads. Looking at chains only, the level of structure at the final round was markedly higher when there was negotiation between two interlocutors than when a single participant

Innovation, selection and the emergence of transparent signals in interaction

We review recent work in interactive experimental semiotics to discuss how biases in innovation and selection during interaction lead to the cultural evolution of transparent signals. Recent studies suggest that individuals are not good at innovating transparent signals. For example, Sulik and Lupyan (2016) show that there are large individual differences in perspective taking abilities, with most participants in communication games being poor at choosing a signal that will be easy for their partners to interpret (though there are ‘rare geniuses’). Verhoef, Roberts, and Dingemanse (2015) found that iconic signals could take generations to emerge, even with only 4 meanings and where the stimuli were designed to have obvious iconic mappings. Tamariz et al. (2017) found that innovations are equally likely to increase or decrease iconicity. While studies such as Tamariz et al. (2014) find that transparent signals are innovated in early generations, there are large individual differences in the ability to do this (though see Ortega, Schiefner, & Ozyurek, 2017; Schouwstra & de Swart, 2014). This would make innovation random (unbiased). This is supported to some extent by cross-linguistic studies arguing that iconicity in the lexicon both increases and decreases over cultural evolutionary time (Blasi et al., 2016). Even classic examples of individual innovation of transparent signals such as ‘universal’ sentence structures of creoles have recently been called into question (Blasi et al., 2017). How, then, do transparent linguistic conventions emerge? One answer is that interaction provides the key mechanisms. Interaction can be thought of as an independent level of cultural evolution. In the broad model of genetic evolution (Dawkins, 1982; Hull, 1980), the gene is a replicator and an organism is a vehicle that interfaces with the environment to allow the replicator to replicate. According to Croft (2000) this model also applies to language: the word or phrase is a replicator and the individual speaker is the vehicle. However, this misses out a level between the individual’s brain and the spoken phrases: turns at talk (sequences) in conversational interaction. This is highlighted in Buyn et al. (2016) which studies signers converging on a shared lexicon. They find that 419

Interactive sequences modulate the selection of expressive forms in cross-signing

This study looks at how linguistic conventions arise in the context of face-toface, real-time interaction. This topic is difficult to study because initial contact events for most languages happened long ago. Two approaches try to tackle this. The first is experimental semiotics: putting people in a situation where they need to improvise a new communication system. Studies have considered the relative role of different biases in this process (Tamariz, et al., 2014) such as frequency bias (where the more frequent signs prevail), content bias (where the more iconically motivated, and/or easily articulated form is selected), and coordination bias (where participants attempt to match each other’s usage). However, these experiments typically involve an artificial language or a restricted or unfamiliar communication medium which are used by participants who already share a common language and culture. This limits the ecological validity of these experiments, especially relating to face-to-face interaction. Another approach has been to study the emergence of new signed languages which emerge spontaneously from scratch, allowing the study of the formation processes of modern human languages in real life. This process has been welldocumented in the case of Nicaraguan Sign Language, which emerged over the course of several decades in a deaf school (Senghas & Coppola, 2001), as well as in multiple ‘deaf villages’ where a local sign language has emerged from the interaction of deaf and hearing community members (Meir et al. 2010). Lesser-known instances of de novo signed communication arises between deaf and hearing nonsigners (trans-languaging), and among deaf signers who do not know a common written or signed language (cross-signing, Kusters et al. 2017; Buyn et al., 2017). Cross-signing is of particular interest as it creates a real-time pressure to establish a shared communicative repertoire. However, these studies are often not designed to be experimentally controlled, and rarely capture the very first period of the emergence of a signed language. In this study we combine the control of experimental semiotics with the eco67

CHIELD: Causal hypotheses in evolutionary linguistics database

Evolutionary linguistics is now a well established field with several conferences and its own journal. The ultimate goal of the field is to explain how complex communication systems emerge and change. A coherent, comprehensive explanation would involve a long chain of causal claims, stretching from genetics to cognition and from prehistorical adaptations to modern language change, supported by a range of methods from experiments to computer simulations. Because of the range of disciplines feeding into language evolution theories, producing such an explanation is a daunting task. In order to help this process, this paper presents a schema and implementation for a database of causal hypotheses about language evolution. Researchers can edit and contribute through a custom web application or through a GitHub repository.

Robust, Causal, and Incremental Approaches to Investigating Linguistic Adaptation

This paper discusses the maximum robustness approach for studying cases of adaptation in language. We live in an age where we have more data on more languages than ever before, and more data to link it with from other domains. This should make it easier to test hypotheses involving adaptation, and also to spot new patterns that might be explained by adaptation. However, there is not much discussion of the overall approach to research in this area. There are outstanding questions about how to formalize theories, what the criteria are for directing research and how to integrate results from different methods into a clear assessment of a hypothesis. This paper addresses some of those issues by suggesting an approach which is causal, incremental and robust. It illustrates the approach with reference to a recent claim that dry environments select against the use of precise contrasts in pitch. Study 1 replicates a previous analysis of the link between humidity and lexical tone with an alternative dataset and finds that it is not robust. Study 2 performs an analysis with a continuous measure of tone and finds no significant correlation. Study 3 addresses a more recent analysis of the link between humidity and vowel use and finds that it is robust, though the effect size is small and the robustness of the measurement of vowel use is low. Methodological robustness of the general theory is addressed by suggesting additional approaches including iterated learning, a historical case study, corpus studies, and studying individual speech.