ArticlePDF Available

Abstract and Figures

Language is one of the most complex of human traits. There are many hypotheses about how it originated, what factors shaped its diversity, and what ongoing processes drive how it changes. We present the Causal Hypotheses in Evolutionary Linguistics Database (CHIELD,, a tool for expressing, exploring, and evaluating hypotheses. It allows researchers to integrate multiple theories into a coherent narrative, helping to design future research. We present design goals, a formal specification, and an implementation for this database. Source code is freely available for other fields to take advantage of this tool. Some initial results are presented, including identifying conflicts in theories about gossip and ritual, comparing hypotheses relating population size and morphological complexity, and an author relation network.
Content may be subject to copyright.
CHIELD: the causal hypotheses in evolutionary
linguistics database
Sea´ n G. Roberts
, Anton Killin
, Angarika Deb
, Catherine Sheard
Simon J. Greenhill
, Kaius Sinnema¨ki
, Jose´ Segovia-Martı´n
Jonas No¨ lle
, Aleksandrs Berdicevskis
, Archie Humphreys-Balkwill
Hannah Little
, Christopher Opie
, Guillaume Jacques
Lindell Bromham
, Peeter Tinits
, Robert M. Ross
, Sean Lee
Emily Gasser
, Jasmine Calladine
, Matthew Spike
Stephen Francis Mann
, Olena Shcherbakova
, Ruth Singer
Shuya Zhang
, Antonio Benı´tez-Burraco
, Christian Kliesch
Ewan Thomas-Colquhoun
, Hedvig Skirga˚rd
, Monica Tamariz
Sam Passmore
, Thomas Pellard
and Fiona Jordan
Department of Anthropology and Archaeology, EXCD.LAB, University of Bristol, 43 Woodland Rd, Bristol,
BS8 1TH, UK,
School of Philosophy and Centre of Excellence for the Dynamics of Language, Australian
National University, Canberra, ACT, Australia,
Department of Philosophy, Mount Allison University,
Sackville, New Brunswick, E4L 1G9, Canada,
Department of Cognitive Science, Central European
University, Oktober 6 street 7, 1st floor, Budapest, 1051, Hungary,
School of Earth Sciences, University of
Bristol, Life Sciences Building, 24 Tyndall Avenue, Bristol, UK,
ARC Centre of Excellence for the
Dynamics of Language, ANU College of Asia and the Pacific, Australian National University, Canberra,
2600, Australia,
Department of Linguistic and Cultural Evolution, Max Planck Institute for the Science of
Human History, Jena, 07743, Germany,
Department of Languages, University of Helsinki, 00014, University
of Helsinki, Helsinki, Finland,
Cognitive Science and Language (CCiL), Universitat Auto`noma de
Barcelona, C/Montalegre, 6, 4
planta, despatx, Barcelona 400908001,
Centre for Language Evolution,
The University of Edinburgh, Dugald Stewart Building, 3 Charles St, Edinburgh, EH8 9AD, UK,
Swedish Language Bank, University of Gothenburg, Gothenburg, SE, 405 30, Sweden,
Communication Unit, Department of Applied Sciences, University of the West of England, Bristol, UK,
CNRS-EHESS-INALCO, Centre de recherches linguistiques sur l’Asie orientale, 105 Boulevard Raspail
75006, Paris, France,
Research School of Biology, Australian National University, 134 Linnaeus Way,
Acton ACT, 2601, Australia,
Department of Social Science, University of Tartu, Salme 1a–29, Tartu, 50103,
Department of Philosophy, Macquarie University, Level 2 North Australian Hearing Hub, NSW,
2109, Australia,
Graduate School of Asia-Pacific Studies, Waseda University, 1 Chome-21-1
Nishiwaseda, Shinjuku City, Tokyo, 169-0051, Japan,
Linguistics Department, Swarthmore College, 500
College Avenue, Swarthmore, PA, 19081, USA,
School of Philosophy and ARC Centre of Excellence for
the Dynamics of Language, Australian National University, H.C. Coombs Building, ACT, 2601, Australia,
School of Languages and Linguistics, University of Melbourne, Babel (Building 139), Parkville, 3010, VIC,
CThe Author(s) 2020. Published by Oxford University Press. All rights reserved. For permissions, please email:
Journal of Language Evolution, 2020, 1–20
doi: 10.1093/jole/lzaa001
Research article
Downloaded from by The Australian National University user on 21 April 2020
Department of Spanish, Linguistics, and Theory of Literature (Linguistics), University of
Seville. Palos de la Frontera, 41004, Seville, Spain,
Department of Psychology, Lancaster University,
Lancaster, LA1 4YF, UK,
Max Planck Institute for Human Cognitive and Brain Sciences, Stephanstraße
1a, Leipzig, 04103, Germany,
School of Culture, History and Language, Australian National University,
Canberra, ACT, Australia and
Psychology, Heriot-Watt University, Edinburgh, EH14 4AS, UK
*Corresponding author: E-mail:
Language is one of the most complex of human traits. There are many hypotheses about how it origi-
nated, what factors shaped its diversity, and what ongoing processes drive how it changes. We pre-
sent the Causal Hypotheses in Evolutionary Linguistics Database (CHIELD,, a
tool for expressing, exploring, and evaluating hypotheses. It allows researchers to integrate multiple
theories into a coherent narrative, helping to design future research. We present design goals, a for-
mal specification, and an implementation for this database. Source code is freely available for other
fields to take advantage of this tool. Some initial results are presented, including identifying conflicts
in theories about gossip and ritual, comparing hypotheses relating population size and morphological
complexity, and an author relation network.
Key words: database, causal graphs, causal inference
1. Introduction
Evolutionary linguistics is a field that uses evolutionary
principles to explain the origins of complex communica-
tion systems, as well as the similarities and differences be-
tween them (e.g., Knight, Studdert-Kennedy and Hurford
2000;Wray 2002;Botha 2003;Christiansen and Kirby
2003;Hurford, 2007;Kinsella 2009;Fitch 2010;Berwick
and Chomsky 2016;Progovac 2019). Scott-Phillips and
Kirby (2010) identified four phases in human language
evolution that are studied within this field:
Pre-adaptation—the preconditions for a language
ability, often related to genetic evolution (e.g.,
Lieberman 1984;Corballis 1999;Hurford 2003;
Slocombe and Zuberhu¨ hler 2005;Cheney and
Seyfarth 2005;Fehe´r 2017;Vernes 2017).
co-evolution—how the first human communication
systems and these pre-adaptations evolved together
(e.g., Deacon 1997;Dor, Knight and Lewis 2014;
Berwick and Chomsky 2013;Pakendorf 2014;
Vigliocco, Perniss and Vinson 2014;Power,
Finnegan and Callan 2016;Falk 2016).
cultural evolution—the initial emergence of new lin-
guistic structures (e.g., Nowak and Krakauer 1999;
Tallerman 2007;Smith and Kirby 2008;Culbertson
and Newport 2015;Progovac 2015;Kempe, Gauvrit
and Forsyth 2015;Tamariz and Kirby 2016;Goldin-
Meadow and Yang 2016;Piantadosi and Fedorenko
language change—the ongoing change in languages
(e.g., Mufwene 2001;Ritt 2004;Croft 2008;
Sampson and Trudgill 2009;Dunn et al. 2011;
Gavin et al. 2013;Majid, Jordan and Dunn, 2015;
Bowern 2015;Bybee 2015;Coelho et al. 2019).
There have been many exciting developments in recent
decades, making it perhaps possible to join them into
larger theories. However, synthesis has become difficult
as there now exists a mountain of theories and evidence,
in increasingly specialised sub-fields. Jim Hurford once
moderated a discussion between four plenary speakers
who had presented four different theories of language
evolution. His first question was ‘What do you disagree
about?’. Nobody had a reply. This showed that, al-
though the theories were internally consistent, they
weren’t connected to each other. This characterises a
problem in many fields—it is possible to have nearly as
many theories as there are researchers, and debate is
often limited to dogmatic acceptance or complete rejec-
tion of these theories, rather than trying to systematical-
ly compare, evaluate, and synthesise.
Progress in evolutionary linguistics will be made by
working towards building a chain of causal links that
join theories together. Since there are many aspects of
language evolution that cannot be tested directly, each
link should be tested with multiple methods and sources
of data—a ‘robust’ approach (Irvine, Roberts and Kirby
2013;Roberts 2018). In order to combine these different
2Journal of Language Evolution, 2020, Vol. 0, No. 0
Downloaded from by The Australian National University user on 21 April 2020
strands of evidence (experiments, models, simulations,
comparative work), researchers must coherently express
how these results relate to each other and to the real
world (Vogt and De Boer 2010). There have been previ-
ous calls for this kind of approach; for example,
Zuidema and de Boer (2010,2013) suggest that compu-
tational modelling should aim for greater ‘model paral-
lelisation’ (qualitatively comparing models against each
other) and greater ‘model sequencing’ (building chains
of models that feed into each other). However, practical
solutions are difficult to produce, due to challenges that
relate to expression, exploration, evaluation, and
Expression: Expressing complex causal hypotheses in
prose is difficult, and many hypotheses are underspeci-
fied. Without deep knowledge of a theory’s particular
sub-field, it is often difficult to identify the assumptions
and claims of a theory. In addition, different sub-fields
may use different terms to refer to very similar concepts,
or use the same term to refer to different concepts. A
classic example of this is the word ‘language’ itself,
which can be interpreted as anything relating to human
communication or only a specific syntactic ability. All
this can lead to researchers talking past each other,
along with a general lack of connection between sub-
fields. How can we better express hypotheses to avoid
these problems?
Exploration: Evolutionary linguistics now includes
many subfields within linguistics (Bergmann and Dale
2016;Berwick and Chomsky 2016), and also relates to
other larger interdisciplinary fields of research, such as
learning or cooperation (Kirby and Christiansen 2003;
Progovac 2019). The methods are diverse, ranging from
molecular genetics (e.g., Enard et al., 2002;Hitchcock,
Paracchini and Gardner 2019), to archaeology (e.g.,
Noble and Davidson 1996;Currie and Killin 2019), to
computational simulation (Steels 1997;Cangelosi and
Parisi 2012;Jon-And and Aguilar 2019). It is therefore
increasingly hard to keep up to date with all the develop-
ments in the field. How can we make theories searchable
so that researchers can access work from other fields
that relates to their own? Furthermore, how can we for-
mally relate these hypotheses to each other, in order to
find similarities and differences?
Evaluation: After relating theories to each other,
how do we then evaluate them? How do we identify the
claims that are supported by evidence, and those that re-
quire further investigation? How can we get an overview
of research conducted on a topic? Studies are getting
more complex, and there is more emphasis on large-
scale tests that can evaluate multiple competing models.
Systematically collecting these hypotheses, storing them
and converting them into statistical models is hard
(Bareinboim and Pearl 2016).
Extension: Language evolution is a field that has
inspired much debate, and even reaching a consensus on
interpretations of hypotheses is difficult. How can we
support researchers in the continuing process of refining
and extending them? How can we ensure that the tools
to do this will be useful into the future?
One possible solution is to harness the power
of causal graphs. A causal graph is a graphical tool
which breaks a complex hypothesis into individual
causal links. We present the Causal Hypotheses in
Evolutionary Linguistics Database (CHIELD, pro-
nounced like ‘shield’,, a data-
base of hypotheses expressed as causal graphs. It allows
users to apply computational search and visualisation
methods, in order to express, explore, and evaluate
hypotheses. This article describes the design and
presents three case studies to demonstrate its functional-
ity. Case Study 1 demonstrates how CHIELD can be
used to explore connections between theories. Case
Study 2 attempts to evaluate competing explanations of
the connection between population size and morpho-
logical complexity, and demonstrates some issues with
vocabulary. Of course, problems such as converging on
the same vocabulary requires more than a database to
solve, but CHIELD may at least provide a space for
spotting potential areas of disagreement. Case Study 3
demonstrates extended uses such as constructing net-
works of authors working on the same topics.
To be clear, the aim is not to build a list of theories
that have been accepted by the scientific community as
‘correct’, that are somehow more ‘prestigious’ or sup-
posedly have no counterevidence. It is, of course, very
difficult to prove a causal effect as being distinct from a
correlation. However, at the heart of any research trying
to explain a phenomenon is an idea about some kind of
causal relationship. The aim therefore is to faithfully
represent these ideas such that researchers can plan fu-
ture work. Furthermore, the aim of this database is not
to highlight one view over another, but to simply present
them on an accessible platform. Its aim is description,
not prescription. Finally, the database aims to be edit-
able and maintainable into the future. We hope that fu-
ture studies in language evolution will be enhanced by
the insights provided by causal graphs.
2. Causal inference and causal graphs
Causal inference is an approach to thinking about caus-
ality (Pearl 2000;Pearl and Mackenzie 2018;Rohrer
2018) that uses graphical tools—causal graphs—to
Journal of Language Evolution, 2020, Vol. 0, No. 0 3
Downloaded from by The Australian National University user on 21 April 2020
depict models of causal processes. Causal graphs consist
of nodes that represent measurable quantities or con-
cepts, and arrows which represent the causal influences
between these nodes. For example, consider why we
might see more shirt stains on hot days. Prof. Whippy
suggests a causal explanation: ‘High temperatures cause
more ice-cream consumption, and more ice-cream con-
sumption leads to more shirt stains’. So we can draw the
following causal graph:
Temperature !Ice cream consumption !Shirt stains
This causal graph has three measurable quantities
(nodes e.g., ‘Temperature’) and two causal links. The
links are interpreted according to an interventionist in-
terpretation of causality (see e.g., Woodward 2003,
2016). For example, the first causal link states that if
one were to ‘intervene’ by changing the temperature
sufficiently, then the amount of ice cream consumption
would also change (and the second link states that if
one were to increase the amount of ice cream consump-
tion sufficiently, then the number of shirt stains would
increase). That is, it makes a counterfactual claim
about a possible world where the measurable quantities
were different. Interventionist causality is often inter-
preted in experimental terms: if one were to experimen-
tally manipulate the temperature, then there would be
a change in ice cream consumption. Relationships
encoded by the causal arrow could be fully determinis-
tic or just probabilistic, and could be between continu-
ous, discrete, or categorical variables. This is a widely
used approach for which many helpful tools are
Note that the graph above also makes some more
claims. First, there is no link from ice cream consump-
tion to temperature, which is interpreted as there being
no causal effect of ice cream consumption on tempera-
ture. That is, if we were to intervene by forcing people
to consume ice cream, then the temperature would not
increase. Secondly, since there is no direct causal link be-
tween the first and the third node, the number of shirt
stains is causally independent of the temperature such
that temperature only influences shirt stains via ice
cream consumption. This hypothesis could be experi-
mentally tested, for example by raising the temperature
(maybe within a shopping centre) and simultaneously
banning the sale of ice cream, to see if the number of
shirt stains decreases.
A key thing to understand about causal graphs is that
they do not necessarily reflect what has been proven to
be true, but instead reflect a particular researcher’s hy-
pothesis. In other words, causal graphs represent ideas
about how the world works, they are not proof that the
world really is like that. There are many other possible
hypotheses (and causal chains) to explain the same phe-
nomenon. For example, Prof. Whippy suggested that
higher temperature causes more ice cream consumption,
which in turn causes more shirt stains. Let’s imagine
that, in another paper, Prof. Gelato attacks Prof.
Whippy’s hypothesis and suggests a different explan-
ation, namely that ice cream consumption has no effect
on shirt stains, and instead seeing a shirt stain reminds
people of ice cream and so they seek it out. In yet an-
other paper, Prof. Sorbet studies climate change and sug-
gests that refrigerated ice cream vans are contributing to
greenhouse gases, therefore affecting the temperature.
We can draw the three hypotheses in a single graph
(Fig. 1).
By representing multiple hypotheses on a single
causal graph, we can identify relationships between
them. For example, Profs Whippy and Gelato agree on
the effect of temperature on ice cream consumption.
Furthermore, Prof. Sorbet’s climate change mechanism
is not necessarily in conflict with the effect of tempera-
ture on ice cream consumption. Both could be operating
to create a feedback loop. However, Prof. Whippy and
Prof. Gelato disagree on the relationship between ice
cream consumption and shirt stains. Intuitively, Prof.
Gelato’s theory seems much less likely to be true, but the
causal graph is still a valid representation of Prof.
Gelato’s theory. Drawing out the two theories has there-
fore shown where the conflict lies, and further suggests a
future empirical test: Prof. Gelato would predict that
staining people’s shirts would cause ice cream consump-
tion to increase, while Prof. Whippy would predict that
it would make no difference.
As this example shows, expressing different hypotheses
as causal graphs allows researchers to express, explore,
evaluate, and extend the relationships between them (see
Ho¨fler et al. 2018). Constructing a causal graph is also an
Figure 1. Causal graph showing three hypotheses of the rela-
tionship between temperature, ice cream consumption, and
shirt stains. Nodes represent measurable quantities, and lines
between them represent hypothesised causal links. Lines with
arrows reflect causal effects and the line with a bar head is
interpreted as ‘no causal effect’. Lines are coloured according
to their source publication. The grey dotted rectangle high-
lights a conflict between two theories.
4Journal of Language Evolution, 2020, Vol. 0, No. 0
Downloaded from by The Australian National University user on 21 April 2020
excellent way to identify underspecified mechanisms.
Causal graphs can help readers to understand papers
(Easterday, Aleven and Scheines 2007;Cao, Sun and
Zhuge 2018;Ho¨ fler et al. 2018;althoughTubau 2008,
suggests that they may not aid reasoning in certain
domains) and facilitate debate (Easterday et al. 2009).
They also help communicate theories indirectly by provid-
ing a roadmap for writing (e.g., authors can check if they
have provided justification or evidence for each causal
link, or identify arguments that are tangential to the cen-
tral claim, see Section 4.2 below).
Expression of theories is an important part of teach-
ing, and a unified approach to expressing causal hypoth-
eses in language evolution could improve student
understanding. For example, students could point to a
particular causal link that they do not understand, and
the database could give them a quote from the paper.
Causal graphs also aid understanding in high-school stu-
dents (Hsu et al. 2015).
Of course, carrying out the experimental manipula-
tions mentioned above might be impractical. In this
case, researchers might depend on converging evidence
from multiple sources, analogies with simulations or
‘natural experiments’ (see e.g., Steels 1997;Pyers et al.
2010;De Boer and Verhoef 2012;Morgan, 2013;
Irvine, Roberts and Kirby 2013). For example, in the
case above, researchers could seek locations where the
temperature does not change, or where the temperature
varies but there is no ice-cream. However, many recent
developments in causal inference have allowed research-
ers to estimate causal effects with purely observational
data, even in cases where researchers cannot directly
measure a variable of interest (Pearl 2000;Bareinboim
and Pearl 2012).
Another advantage is in identifying conditioning sets:
which variables to control for in an analysis. Standard
methods about how to choose control variables are
often vague (Pearl and Mackenzie 2018), and many as-
sume that controlling for more variables makes the cen-
tral test more robust. However, controlling for some
variables can create spurious correlations due to col-
liders. A collider is a node on a causal path with two
causal links (arrows) pointing into it (see Supplementary
Material S5). In the example above, Prof. Gelato pre-
dicts that ‘ice cream consumption’ is a collider along the
path involving temperature and shirt stains. Gelato
would predict that temperature and shirt stains both
contribute to ice cream consumption. If this were true,
then temperature and shirt stains should be uncorre-
lated, except when controlling for ice cream consump-
tion, at which point they would become correlated (see
e.g., Elwert 2013). This property of colliders means that
spurious correlations can sometimes be introduced by
statistical controls (see Elwert and Winship 2014;Ding
and Miratrix 2015;Middleton et al. 2016;Westfall and
Yarkoni 2016;Rohrer 2018;York 2018). Drawing a
causal graph helps to identify these cases, and therefore
helps make effective decisions in study design and ana-
lysis. These benefits make causal graphs a very powerful
tool for research in the social sciences. However, causal
graphs are not substitutes for careful thought and de-
sign. They are tools for helping researchers do their job.
CHIELD aims to make these tools more accessible.
Some recent developments in causal inference theory
go further by attempting to estimate the most likely
causal graph directly from observational data on large
number of variables (Kalisch et al. 2012;Hauser and
Bu¨ hlmann 2012; see Heinze-Deml, Maathuis and
Meinshausen 2018, for applications in linguistics see
Roberts and Winters 2013;Baayen, Milin and Ramscar
2016;Blasi and Roberts 2017). Analysing the entire
causal network, rather than just explaining the variation
in a single target variable, makes it possible to study
much more complex relationships, and to form a more
comprehensive picture of complex systems such as those
found in social sciences. However, there are many pos-
sible ways of applying this method (parameters of the al-
gorithm, treatment of the data, assumptions of the
statistical tests), and these can lead to big differences in
the results and interpretation. What is needed is a way
of compiling and organising existing knowledge about
causal effects within one domain, in order to evaluate
the automatically derived causal graph. CHIELD aims
to provide this prior knowledge in order to protect
researchers from making hasty post-hoc hypotheses
from the output of automatic methods.
3. Design of the database
We reviewed various existing applications (see
Supplementary Material S1) to check whether they ad-
dress the problems discussed above. There appears to be
no single tool that allows researchers to express hypoth-
eses (visually), explore the way they interact, and evalu-
ate them. CHIELD aims to fill this gap. In this section,
we cover the design of CHIELD. The principle is to store
data in a simple spreadsheet format, with scripts that
convert them to a database that can be accessed and
edited through a customised web interface.
3.1 Expression
The aim of CHIELD is to allow researchers to express
causal graphs for a particular theory, in a simple and
Journal of Language Evolution, 2020, Vol. 0, No. 0 5
Downloaded from by The Australian National University user on 21 April 2020
intuitive format. CHIELD is based around ‘documents’
(usually a peer-reviewed, published paper), each of
which have three associated files: bibliographic informa-
tion stored as a bibtex file, a simple text file listing who
contributed the data, and a spreadsheet file which stores
information about the causal links in the document.
These files are kept together in a folder, which is named
after the unique bibtex key for the document. In order to
make finding particular files easier, we borrow the de-
sign of Glottolog: document folders are sorted into par-
ent folders for each year of publication, and then each
year is sorted into parent folders for each decade of pub-
lication (see
LD/tree/master/data/tree/documents). For example, the
files for Dunbar’s (2004) paper on gossip is stored in
‘documents/2000s/2004/dunbar2004gossip/’. This or-
ganisation is largely for the convenience of the database
developer. For most users, the online interface for cod-
ing papers will automatically create files in the correct
location, and documents can be searched using a user-
friendly interface.
Within the causal links spreadsheet, a single observa-
tion (a single row) represents a single causal link be-
tween two variables. The causal link has a number of
associated properties, shown in Table 1. The types of
causal relation are shown in Table 2. While a standard
causal graph does not encode the direction of the rela-
tionship, this information can be encoded in CHIELD
via the ‘Cor’ (correlation) field, including positive, nega-
tive and non-monotonic relationships.
The format is easiest to explain with an example
(Table 3). Lupyan and Dale (2010) suggested a link be-
tween the number of speakers a language has, and the
language’s morphological complexity. In their paper,
they hypothesise that: ‘Speakers of languages [with large
numbers of speakers] are more likely to use the lan-
guage to speak to outsiders—individuals from different
ethnic and/or linguistic backgrounds.’ This could be
coded as the first row of Table 3. They hypothesise a
causal effect (‘>’) and they support this with references
to the literature (type ¼‘review’). The main quantita-
tive results from the paper demonstrate a negative cor-
relation between population size and morphological
complexity. This can be coded separately as a ‘statistic-
al’ type of support. See Section 5.3 or the entry in
lupyan2010language) for more information. More
detailed specifications for each field are provided in the
supporting materials.
Defining hierarchies of variables is possible by using
a colon character in the variable name. For example, in
a paper on morphological complexity, Nettle (2012) dis-
tinguishes ‘paradigmatic complexity’ from ‘syntagmatic
complexity’. Ideally, these concepts should link to the
‘morphological complexity’ node of other hypotheses
while maintaining their specificity. Therefore, a coder
might use ‘morphological complexity: paradigmatic’
and ‘morphological complexity: syntagmatic’. Various
settings in the visualisation allow users to switch from
connecting nodes only if their full labels match, and con-
necting them if their higher order category matches.
It would also be possible to allow empirical measures
of causal strength to be stored for each causal link,
which are important for causal studies in fields like
medicine. However, they are not part of the core design
goals here, since many studies of evolutionary linguistics
do not estimate these kinds of measures directly, but use
experiments, models or case studies to make analogies.
Table 1. Fields in the database
Field Description Values
Var1 The source variable Any lowercase string. Hierarchies can be separated by ‘:’
Relation The type of causal relation See Table 2
Var2 The destination variable Same as Var1
Cor The direction of correlation blank, pos, neg, nm (non-monotonic)
Topic The topic of study Lowercase strings separated by ‘;’
Stage The evolutionary stage Preadaptation, co-evolution, cultural evolution, language
change (see Scott-Phillips and Kirby 2010)
Type The type of evidence used to support the link Experiment, review, model, simulation, statistical, qualita-
tive, logical, hypothesis, other
Confirmed Whether the evidence supported the link Blank, yes, no
Notes For context, e.g., a quote from the paper that supports
the observation
Any string
See the Supplementary Material S2 for the full specification.
6Journal of Language Evolution, 2020, Vol. 0, No. 0
Downloaded from by The Australian National University user on 21 April 2020
The most important design feature here is a low barrier
to adding data, and causal strength estimates may not be
very helpful for many researchers in evolutionary lin-
guistics. Having said this, it would be easy to add this
kind of information to a future version of the database.
3.1 Exploration
The web interface for CHIELD includes several ways of
searching the data. PHP scripts fetch data from the com-
piled SQLite version of the database and serve them up
into a dynamically searchable tabulated format (using
the Datatables library, For ex-
ample, users can view a list of documents, and search by
title, author or year. Each document links to a document
page, which displays the bibliographic information, the
list of causal links (with the specification information
above) and a list of other documents that have variables
in common (found using live queries of the SQLite data-
base). CHIELD also includes an interactive visualisation
of the causal graph using the vis.js javascript library
( Users can view tables which list all the
causal links or all variables. These link to similar pages
which give overviews of the variable (e.g., all causal
links involving ‘linguistic diversity’). These search fea-
tures make it easy to find documents or causal links, and
to explore links between documents.
Collaboration is an important part of language evo-
lution research (see Bergmann and Dale 2016;
Youngblood and Lahti 2018). There are now various
approaches for automatically suggesting collaborations
between authors based on network models (e.g., Lopes
et al. 2010;Xu et al. 2010;Yan and Guns 2014;Guns
and Rousseau 2014;Kong et al. 2017). By combining
the bibliographic information with the causal links data,
CHIELD may be able to find connections between
authors beyond co-authorship. Users can search a list of
authors, with links to author pages showing all docu-
ments and causal links by an author, a list of their
co-authors and a list of potential collaborators. The po-
tential collaborators are defined following a method
similar to the AXON database (see Supplementary
Materials S1): find authors whose causal graphs overlap
(have nodes in common), but who have not published
any papers together.
CHIELD includes an “Explore” mode, where users
can load multiple causal graphs into a single visualisa-
tion. This includes various tools:
An interface for combining multiple causal graphs
into a single visualisation.
Interactive manipulation, allowing adding a node
from the database, adding all nodes from a particular
document, or removing nodes or edges from the cur-
rent visualisation.
Expand links from the currently selected node (display
all causal links that connect to a particular node).
Find evidence from other documents for the currently
displayed links.
Display sub-variables under a single higher order
Find causal pathways between two given variables,
in order to discover alternative hypotheses.
Table 2. Syntax for expressing the relationship between the source and destination variable
Syntax Meaning Symbol
X>Y A change in X causes a change in Y !
X<¼>Y X and Y co-evolve $
XY X and Y are correlated ----
X id="721" />Y X does not causally influence Y j
XY X is a necessary precondition for Y !(red)
X¼ Y X is an indicator of (measured by) Y
X ^ Y X exerts an evolutionary selection pressure on Y ----"
Table 3. Example coding for two links in Lupyan and Dale (2010)
Var1 Relation Var2 Cor Topic Stage Type Confirmed Notes
Population size >Population
pos Contact Language
Review Speakers
of lang ...
Population size Morphological
neg Morphology Language
Statistical Yes Population
was ...
Journal of Language Evolution, 2020, Vol. 0, No. 0 7
Downloaded from by The Australian National University user on 21 April 2020
Exporting the causal graphs as ‘dot’ format files.
These can be used in visualisation tools, like
GraphViz (Ellson et al. 2004) or Gephi (Bastian,
Heymann and Jacomy 2009) to produce images for
use in publications (e.g. pdf, svg).
These tools can help a researcher find “upstream”
explanations that feed into their own hypothesis, and
also see how their hypothesis can generate predictions
for other “downstream” work.
CHIELD finds causal pathways using a variant of
Dijkstra’s algorithm (Dijkstra, 1959), which finds all
possible paths between two nodes. This can become
complicated if there are loops or multiple connections
between variables. However, the algorithm does not
need to find all possible paths, only the set of nodes
along all possible paths (the SQL query will find all
paths connecting these nodes later). Therefore, the algo-
rithm only follows each edge once. Another issue is
which types of causal connection should be considered
in the search. Considering only standard causal links
(‘>’) can lead to missing some connections, but includ-
ing all links can lead to very large causal networks which
are not useful. The current implementation, therefore,
only considers the following connections: ‘>’, ‘’,
<¼>’ and ‘¼’. In the future, this search process could
be customisable.
3.2 Evaluation
The explore mode allows control over the visualisation
in order to support evaluation and effective design deci-
sions for future research. Parameters of the causal graph
can be manipulated (e.g., to display hierarchical layout
or a dynamic ‘spring’ layout). For example, the colours
of edges and node positions can be manipulated to re-
flect the source publication, type of evidence, type of
causal effect, or direction of correlation (communicated
through a hideable legend). The colour schemes are
designed to be distinguishable by colourblind users.
The explore mode also includes a tool for highlight-
ing differences between hypotheses. If causal links from
more than one document are loaded, an algorithm finds
edges where the two documents disagree. At the mo-
ment, this only involves cases where one document
claims that there is a causal connection (‘>’, ‘’,
<¼>’) and the other claims that there is no causal con-
nection (‘/>’), but this could be expanded in the future.
If conflicts are detected, the relevant edges are high-
lighted and the visualisation zooms in to display them.
The final causal graph can be exported to the
DAGitty web interface (Textor, Hardt and Knu¨ ppel
2011) or as model definition code for the R package
phylopath (von Hardenberg and Gonzalez-Voyer 2013;
van der Bijl 2018), which performs phylogenetic path
analysis on multiple competing models (each document
is listed as a different model). The full database is openly
available in a range of formats, for further manipulation
by statistical software.
3.3 Extension
Constructing a comprehensive database of theories of
language evolution is not a feasible task for a single per-
son. Consequently, CHIELD has been designed to be ex-
tendable by large numbers of contributors. This makes it
important to be able to curate contributions and keep
track of changes. It is not expected that everyone will
agree on interpretations, but these issues are worth
debating, and the database should include space for dis-
cussion and revisions. If the database is to be useful in
the long-term, there are also the practical questions
about how best to store the data. Moreover, we want
the basic application to be extendable to other fields and
by other developers. Therefore, the design goals here
Integration with version control software for keeping
track of changes.
Tools for discussion and curation.
Simple file formats that can be easily edited.
Non-proprietary data formats for longevity.
Open source code for extension to other fields.
CHIELD is integrated with Git and GitHub for keep-
ing track of changes, and for managing public contribu-
tions, issues, and discussion (
CHIELDOnline/CHIELD). This also provides open ac-
cess to the source code if other fields of research want to
develop their own version. The file formats (bib, txt,
and csv) are open-source, non-proprietary, and simple
to edit. All processing scripts and external libraries are
free and open source.
Data can be added to CHIELD by coders through
the GitHub repository, but the website also includes a
simpler customised interface for adding data (using the
javascript library js-grid, This guides
coders through the process of contributing a document
to CHIELD, including how to draw the causal graph
visually or to upload a causal links template spreadsheet
from their computer (more information at https://chield. The interface sug-
gests existing variable names for the coder to re-use, in
order to maximise convergence on variable names.
Once the coder has entered their data, a script creates
the standard file formats for CHIELD and sends them to
8Journal of Language Evolution, 2020, Vol. 0, No. 0
Downloaded from by The Australian National University user on 21 April 2020
the GitHub repository (it creates a new branch, commits
the new files to it, and then makes a pull request to
merge these into the main branch). This sends an alert to
the owner of the repository who can review the contri-
bution and accept it into the database. The web manager
can now pull these changes down to the server, and run
the compiling script that creates an updated SQL data-
base and deploys it to the website. This system is per-
haps one of the more successful parts of the database
design that allows anyone to contribute without needing
to know how to use version control software. This sys-
tem also takes advantage of the existing tools of
GitHub, including user logins, bug reports, discussion
threads, and tracking changes which reduces the devel-
opment load. Finally, this system also allows rapid turn-
over: after the coder has submitted their data, the
process of adding it to the database and updating the
website takes just a few minutes.
3.3.1 Scope of data
The condition for entry into the database is that the
document proposes a hypothesis with causal claims that
relates to some part of the evolution of communication
and that it is published in a peer-reviewed publication
(e.g., journal paper, peer-reviewed conference proceed-
ings, book chapter). This could be either biological or
cultural evolution (or both) at any evolutionary stage.
As a rule of thumb, the document should relate directly
to language, or be ‘one link away’ from a relevant lan-
guage document. Entry into the database does not mean
that the hypothesis is correct or widely accepted, or even
empirically supported. The aim is not that the database
be a single theory of the evolution of communication,
but a reflection of the whole field. Potential sources for
documents include the Language Evolution and
Computation Bibliography (, includ-
ing over 2,500 papers), the Journal of Language
Evolution, the Journal of Interaction Studies,EvoLang
conferences, and so on.
3.3.2 Moderating contributions
CHIELD aims to be open and editable. Anyone can con-
tribute documents to CHIELD or edit any existing docu-
ments (as long as they have a free GitHub account). This
necessitates moderation for quality control. This is done
through GitHub, with a central administrator being able
to review each contribution or edit (as a pull request) be-
fore it is added to the database. An advantage of this sys-
tem is that it creates a record of every change to the
database. The document pages of the website include
buttons for raising issues with the coding (through a
pre-filled GitHub issues page), which may be taken up
by other users or an administrator. There are various ad-
ministrator tools for aggregate tasks, like replacing all
occurrences of a particular variable label with another.
4. Results
The CHIELD website is live (
and is updated continuously (
ELDOnline/CHIELD). The results in this article are
based on version 1.1. In general, we found coding of
causal graphs challenging but productive. We found that
a paper might take between 10 minutes and an hour to
code, depending on the complexity of the theory, the
methods used and the coder’s familiarity with the sub-
ject. Version 1.1 of CHIELD includes 400 documents
and 3,406 causal links between 1,700 variables. These
were contributed by 41 coders (see https://chield.excd.
The ratio of unique variables to links is high, suggest-
ing that many documents introduce new variables.
Ideally, if the database was approaching a ‘full picture’
of the field, the number of new variables being added
would decrease over time. Figure 2 shows the relation-
ship between the number of documents and the number
of unique variables (points are average number of
unique labels from 1,000 random orderings of docu-
ments). The curve is growing slower than strictly linear,
and if we assume a quadratic function, then we estimate
that a plateau will be reached when around 650 docu-
ments have been coded. Of course, at the moment, the
database reflects the research interests of its contributors
and might under-represent many sub-fields, so coverage
of the whole field might require many more documents.
However, another sign of convergence is that documents
are highly connected. The largest component of the net-
work includes around 75% of all documents (Fig. 3).
4.1 Case study 1: gossip, ritual, and language
The first example of CHIELD’s functionality demon-
strates how theories can be compared against each
other, and how CHIELD can be used to explore empiric-
al evidence that might help resolve debates. Figure 4
shows two theories about the coevolution of group or-
ganisation and communication in human language
emergence. The first theory, Dunbar (2004) relates
population size, brain size, and gossip: risk of predation
drives individuals into larger groups for safety, but these
groups require more time dedicated to social bonding in
order to maintain alliances. Since gossip is more efficient
at maintaining a larger number of alliances than one-on-
Journal of Language Evolution, 2020, Vol. 0, No. 0 9
Downloaded from by The Australian National University user on 21 April 2020
one physical grooming, it was selected for in humans,
leading to a coevolution between population size and
brain size (required for gossiping). However, the second
theory, Knight, Power and Watts (1995) argues that for
gossip to function as a form of social bonding, it needs
to be underwritten by another mechanism that guaran-
tees honesty. For them, the value of gossip is socially
determined through ritual. Therefore, the first symbolic
communications would have been about collaborating
in the maintenance of fictions and ritualistic acts which
enforce in-group solidarity (e.g., females banding to-
gether to conceal signals of menstruation in order to
control access to sex, in return for parental investment
from males).
Both Dunbar’s and Knight et al.’s theories are much
more complicated than these simplistic explanations.
Nonetheless, they can be suitably represented as causal
Figure 2. Basic statistics of CHIELD. The first stacked bar shows the number of links for each stage of language evolution. The se-
cond stacked bar shows the number of links for each type of evidence. The upper-right panel shows the number of documents by
decade of publication. The lower-right panel shows how the number of unique variables grows as documents are added to
CHIELD. For context, a linear grey line is shown beneath (intercept ¼0, slope ¼5).
Figure 3. The largest connected component in CHIELD (1,542
variables, 3,222 links).
10 Journal of Language Evolution, 2020, Vol. 0, No. 0
Downloaded from by The Australian National University user on 21 April 2020
graphs. Doing so leads to additional evaluative insight:
Even though the theories are seen as being totally
opposed to each other, when their causal graphs are
overlaid, there is only one place where they conflict.
CHIELD automatically identifies the critical causal link
at the heart of the disagreement: whether gossip can
maintain alliances (Dunbar argues that it does; Knight,
Power and Watts argue that it does not in the absence of
ritual bonds). Other parts of the theories are actually
mutually compatible.
One of the core strengths of CHIELD is that it can
identify studies that test critical causal links. For ex-
ample, the explore tool automatically discovered an ex-
periment by Rudnicki, Backer and Declerck (2019)
where pairs of participants played a trust game after ei-
ther gossiping for 20 minutes or interacting without gos-
siping. They found that (for prosocial people) gossiping
increased trust, in line with Dunbar’s hypothesis.
Rudnicki, Backer and Declerck do not discuss Knight,
Power and Watts’ hypothesis, but the link discovered
through CHIELD suggests that their paradigm could be
extended to compare the two theories: participants
could perform a bonding ritual together rather than gos-
siping. In this way, CHIELD can be an effective tool for
identifying critical differences between hypotheses, and
also for discovering work that might help resolve the
disagreement between them.
4.2 Case study 2: population size and
morphological complexity
The next example demonstrates the ability to evaluate
multiple different theories. Lupyan and Dale (2010),
following theories from studies of language contact,
showed that a language’s morphological complexity can
be predicted by the number of speakers who speak it.
They hypothesised that larger populations have more
adult learners and more contact with other languages.
These factors might cause a pressure for the morpho-
logical system of the language to become simpler. For
example, adults are worse at learning morphological
rules than lexical strategies. That is, languages with
large number of speakers might adapt to the adult ‘cog-
nitive niche’.
Figure 5 shows the causal graph for this hypothesis,
which highlights some key points. First, while the
hypothesised mechanism has several steps, the main
quantitative result is a correlation between population
size and morphological complexity (due to the interven-
ing variables having limited data available). The correl-
ation is consistent with the hypothesis, but alternative
data or methods could be applied to try and support
each causal link. Secondly, while most links are sup-
ported either by reviews from the literature or statistical
analyses, there is a ‘weak link’: there was no supporting
evidence for a causal effect of population size on the
proportion of adult learners. Although it makes logical
sense, ideally it should be confirmed empirically (as was
recently done in Koplenig 2019).
Lupyan and Dale’s study led to several other empiric-
al studies looking at the relationship between morpho-
logical complexity and population size, as well as the
invocation of previous studies to explain the patterns. In
Fig. 6, we show 21 of these studies represented as causal
graphs (Supplementary Material S3 include the R script
for automatically generating this figure from CHIELD).
Figure 4. Comparison of Dunbar (2004) and Knight, Power and Watts (1995), with an insert showing the conflict between them and
an additional study by Rudnicki, Backer and Declerck (2019) that was automatically discove red.
Journal of Language Evolution, 2020, Vol. 0, No. 0 11
Downloaded from by The Australian National University user on 21 April 2020
This graph includes evidence from fieldwork, cross-
cultural statistics, lab experiments and simulations.
While this visualisation may look complicated, in tan-
dem with the interactive features of the website it pro-
vides a way of systematically thinking about different
explanations. For example, Nettle (2012) and Cuskley
and Loreto (2016)’s explanation involves general proc-
esses, whereby larger populations change frequency dis-
tributions in ways that lead to simplification. In
contrast, Wray and Grace (2007) and Little (2012) sug-
gest that there are specific effects of the way adults sim-
plify their speech when talking to strangers. Ardell,
Anderson and Winter (2016) suggest that the mechan-
ism involves phonetic variation rather than morphology
directly, while Atkinson, Kirby and Smith (2015) sug-
gest that phonology is the key.
Interestingly, as the causal graph shows, many of
these theories do not necessarily conflict with each
other. In fact, there are several nodes in this network
that could be explicitly measured in a way that allowed
the explanatory power of different causal paths to be
tested against each other (e.g., in a causal path regres-
sion). Examples of conflicts include the presence or ab-
sence of a robust correlation between the proportion of
adult learners and morphological complexity (Bentz and
Winter, 2013;Koplenig 2019). Koplenig (2019) finds a
robust correlation between population size and com-
plexity in terms of entropy, but not a link to morpho-
logical complexity.
The graph also suggests areas where theory might be
extended. For example, other factors that influence mor-
phological complexity include word order (Sinnema¨ki
2010, see also Koplenig 2019) and morphological re-
dundancy (Berdicevskis and Eckhoff 2016). Perhaps
these interact with the other variables in a way that
might introduce confounds. Following Thurston (1989)
and Hymes (1971),Ross (1996) discusses an
alternative mechanism—esoterogeny—which predicts
that more contact will lead to greater morphological
complexity, due to competing groups trying to distin-
guish themselves. It is currently unclear what would be
required to test this prediction on a large scale, but some
kind of measure of between-group competition would
be needed (see Roberts 2010). These are just some of the
ways that causal graph visualisations might inspire fu-
ture work.
Another contribution to theory that can arise when
constructing these graphs, is an understanding of differ-
ences in the way terminology is used. For example, while
coding some of these studies, it became clear that ‘popu-
lation size’ was not the only way to refer to how many
speakers a language has. Table 4 shows 10 different
terms alongside their definitions. In anthropology, terms
are usually more specific in order to capture distinctions,
such as the total number of speakers of a language and
the number of people in a community (the first may be
very high while the second could be fairly low). In con-
trast, studies in modelling or demography tend to use
terms derived from biology that refer to more abstract
properties. Dissonance amongst terms is a known issue
(at least by experts), but CHIELD provides motivation
and a formal system for trying to achieve unification in
4.3 Case study 3: networks of authors
Figure 7 shows a network of connections between
authors, where each connection indicates that causal
graphs from hypotheses by the two authors have at
least one node in common. This network includes 500
of the 720 authors in the database and 6,065 connec-
tions. This kind of network is unique, since it is built
from hand-coded data about central components of
hypotheses rather than publication of co-authorship
statistics or textual analysis. Of course, these are based
on just a small sample of papers in the field, and heav-
ily biased by the research interests of the coders.
Nevertheless, we can use standard network analysis
tools like modularity to find clusters of authors. The
network splits into three main clusters which might be
characterised based on their methods: experimental
(experimental semiotics, iterated learning, computa-
tional modelling), statistical (cross-cultural statistics,
phylogenetics), and comparative (animal communica-
tion, genetics). This suggests that researchers mainly
cluster on their approaches rather than their topics,
meaning that there may be scope for more collabora-
tive work.
Network measures can be used to find ‘brokers’: indi-
viduals who provide critical bridges between clusters. For
Figure 5. Lupyan and Dale (2010)’s hypothesis, expressed as a
causal graph. Links are coloured according to the type of evi-
dence provided.
12 Journal of Language Evolution, 2020, Vol. 0, No. 0
Downloaded from by The Australian National University user on 21 April 2020
example, betweenness-centrality calculates the shortest
path between each pair of nodes, then for each node
counts the number of shortest paths that flow through it.
This estimates the number of connections that would be-
come longer or potentially break if the given node was not
there. Brokers include Kim Sterelny, Stephen Levinson,
Kenny Smith, Monica Tamariz, Simon Kirby, Claire
Bowern, Simon Greenhill, Dan Dediu, and Andrew
Whiten. These are generally researchers with interests that
span the three main clusters.
This information is used in the online interface for
CHIELD to suggest authors who might have interests in
common and might make good collaboration partners.
Authors are connected in the network if they share a
node but have not co-authored a publication (listed in
CHIELD). These suggestions are biased by the selection
Table 4. Different terms for number of individuals in a group
Field Variable Description
Sociology Group size e.g., ‘people whose members are mutually aware of each other and can potentially
interact’ (McGrath 1984); ‘the number of individuals [where we know] who
they are and how they relate to us’ (Dunbar and Dunbar 1998)
Animal behaviour Individual group size ‘the size (the number of individuals) of a group that a particular individual lives
in’ (Jovani and Mavor 2011)
Anthropology Population ‘enumerations and estimates (with dates); density (e.g., arithmetical, for arable
land); population trends; etc.’ (HRAF
ehrafe/ ocm¼161)
Anthropology Mean size of local
‘The average population of local communities, whatever the pattern of settlement,
computed from census data or other evidence.’ (D-PLACE
Anthropology Community size ‘The population size of the focal or typical community’ (Murdock and Wilson
Population size ‘Population of ethnic group as a whole’ (Ethnographic Atlas)
Effective population size ‘if the agent is connected to many others via relatively few steps (i.e. a low average
shortest path length), then its effective population size is large, and vice versa’
(Spike, 2018)
Social network size Various definitions, see e.g. Killworth, Bernard and McCarty (1984);Rogerson
Animal behaviour Broadcast network ‘the number of people with whom we can communicate directly and indirectly’
(Dunbar 2004: 103)
Linguistics Speech community Various definitions e.g. ‘All people who use a given language or dialect’. (Lyons,
1970); ‘groups that share values and attitudes about language use, varieties and
practices’ (Morgan 2014); ‘The speech community is not defined by any marked
agreement in the use of language elements, so much as by participation in a set
of shared norms’ Labov (1972: 120–1); ‘Any human aggregate characterized by
regular and frequent interaction’ (Gumperz 1968: 66)
Figure 6. Nineteen hypotheses for the mechanism that connects population size and morphological complexity.
Journal of Language Evolution, 2020, Vol. 0, No. 0 13
Downloaded from by The Australian National University user on 21 April 2020
of documents, but generally make sensible suggestions.
For example, the top suggested collaborator for Dan
Dediu, an executive editor of the Journal of Language
Evolution, is Bart de Boer, another executive editor of
the same journal.
4.4 Challenges
There were several challenges while building CHIELD,
many of which may be applicable to any attempt that
catalogues causal theories. First, we found that coding
causal graphs from publications is hard. In a few cases,
two coders coded the same paper and produced graphs
with no overlap. There was even a case where a coder
coded a paper twice, several months apart, and the
agreement was low. Such problems in coding come from
two sources: First, there is often a lack of clarity on the
author’s part (which causal graph methods might ad-
dress). Experiments have shown that researchers can
correctly identify causal structures when appropriate
information is available (Wiley and Myers 2003).
Secondly, the research priorities of the coder will influ-
ence how the graph is coded. For example, biasing the
details that they code, or how they choose to divide the
elements of the hypothesis into nodes. More specific
problems arose regarding the resolution at which to
code the paper (general processes versus specific mecha-
nisms), or how to represent structures like trade-offs or
interaction effects. While causal links into a node repre-
sent a joint conditional probability distribution that
does capture the information in an interaction effect,
there is no standard way of graphically representing
interactions in causal graphs as distinct from simple dir-
ect effects (see VanderWeele and Robins 2007: 1098–
1100). However, in many cases the issue can be resolved
by thinking about the extra steps in between (see
Supplementary Material S4). We noticed that some
coders appeared to have particular ‘styles’, such as aim-
ing to code the hypothesis as a set of discrete steps
(where different species have progressed to different
Figure 7. Author network. Connections between authors indicate that causal graphs from hypotheses by the two authors have at
least one node in common. The links are coloured by clusters discovered according to network modularity.
14 Journal of Language Evolution, 2020, Vol. 0, No. 0
Downloaded from by The Australian National University user on 21 April 2020
extents), or as a dynamic system (where different species
have different values at each node). These disagreements
lead to practical problems. For example, CHIELD
depends on agreement in variable names. Small differen-
ces in labelling lead to major changes in the network.
Very general node labels such as ‘language’ are also un-
helpful when trying to identify causal paths between the-
ories. These issues are mitigated to some extent by the
provision of term recommendations on data entry and
the ability for anyone to edit causal graphs after initial
submission. Fully unifying the vocabulary would take a
lot of editing work, but this might be worth the effort in
order to improve research communication in the field.
The final challenge was that a theory cannot be fully
understood from its graph alone. For this reason, it is
strongly encouraged to add quotes from the paper in the
‘notes’ field that include important context. However,
combined with some general background knowledge of
the field, causal graphs can provide a very helpful overview
of ongoing and existing research. Indeed, we found that
the process of coding causal graphs increased the coder’s
own clarity of the paper beyond simply reading the text.
We maintain that CHIELD is a useful tool for researchers
to organise their research, but we are aware that it would
take a huge amount of work to produce a definitive over-
view of the whole field, if this were at all possible.
The challenges above affect the scalability of the data-
base as the number of contributors and topics expand.
We are currently optimistic. Thanks to the use of git and
Github, the current system seems adequate for handling
the amount of data and edits (around 1,000 edits to
causal graphs). CHIELD serves a relatively large and di-
verse set of researchers (41 contributors from multiple
disciplines, e.g. anthropology, cognitive science, compu-
tational modelling, genetics, morphology, philosophy,
primatology, psychology, sociolinguistics, syntax, and
typology), who have been able to contribute despite
many having little technical or programming back-
grounds. Although anyone can suggest edits, there is cur-
rently only one moderator, and the project may need to
expand to one or more committees for dealing with sub-
topics or variable names. We suggest that expansion to
other fields may be best done by starting a separate data-
base using the tools and structures of CHIELD as a tem-
plate (e.g., the Hypothesis Database for Research into
the Evolution of Culture, HyDREC
winters/HyDREC), rather than trying to fit too many
fields in a single source. Refining the database may be
facilitated by dedicated sessions at workshops. One
major bottleneck is training in causal inference methods
for the coders. We hope that this can become a more cen-
tral part of research training in general in the future.
More generally, the data in CHIELD are designed to
be used with the Directed Acyclic Graph (DAG) ap-
proach to causality and the tools for dealing with them.
However, there are limitations on the kinds of causal
relations that can be represented in this way, and other
approaches are available (see e.g., Mahoney 2008;
Granger 2016,Blasi and Roberts 2017). In particular,
causal loops can be graphically represented using DAGs,
but they violate some of the assumptions that allow vari-
ous parts of the causal inference machinery to function.
Similarly, inference is complicated by non-monotonic
effects (VanderWeele and Robins 2010) or effects hap-
pening on different timescales (Aalen et al. 2016). Many
of the theories coded in CHIELD include these features,
which limits quantitative treatment and complicates
searching for ancestors and links between theories.
However, we believe that the current approach still has
qualitative value and as methods and theory develop,
CHIELD and the data therein can grow to alleviate these
We note that there are existing attempts for automat-
ic extraction of causal relations directly from publica-
tion texts (see Asghar 2016;Alshuwaier, Areshey and
Poon 2017;Mueller and Huettemann 2018;Mueller
and Abdullaev 2019;Tshitoyan et al. 2019). However,
given the problems above, we are sceptical of the effect-
iveness of this approach for evolutionary linguistics. In
any case, progress in automated coding of causal struc-
ture would require human-coded, ‘gold standard’ data,
which could be provided by CHIELD.
5. Conclusions
We presented the design and implementation of
CHIELD, a database of causal hypotheses in evolution-
ary linguistics. We demonstrated CHIELD’s uses,
including identifying critical differences between theo-
ries, discovering critical evidence, synthesising theories,
and finding collaborators. The main challenge is in the
coding of hypotheses. However, the challenge derives
mainly from the difficulty of accurately conveying causal
ideas in prose, rather than any limitation of causal infer-
ence tools. This issue would be assisted by the use of
causal graphs to express hypotheses in the first place.
One possible solution would be for journals to encour-
age that authors submit a causal graph with each publi-
cation (as a figure or as metadata). This could then be
more easily fed back into CHIELD order to strengthen
the representation of the database.
There are many ways that CHIELD could be
extended in the future. For the database itself, we plan
to add permanent links to the variables and documents,
Journal of Language Evolution, 2020, Vol. 0, No. 0 15
Downloaded from by The Australian National University user on 21 April 2020
add statistical results to edges (though there is the ques-
tion of how to allow multiple types of statistic to one
edge), and add more flexibility in searching (e.g., limit
the results to a particular stage of language evolution).
There is also the possibility of converting to database
formats that explicitly support network structures for
advanced querying (e.g., GraphML). We would like to
add support for large touchscreens and interactive de-
bate so that CHIELD could be used as a tool for live dis-
cussion between researchers. As research methods
increasingly involve collaboration with multiple
researchers, this could help people check how their
understanding of terms and concepts match. We would
also like to develop a way for CHIELD to automatically
suggest studies (e.g., for student projects). For example,
randomly choosing a causal link that currently has no
empirical evidence, detecting colliders that might require
alternative explanations, or showing a link between two
theories that might not have been noticed. It would also
be possible to link nodes to open source data, so that
CHIELD could try to discover links that currently have
no empirical support, but for which there was existing
data that could be utilised. Of course, automated proce-
dures may be susceptible to bias (e.g., publication
biases), and they can be no replacement for careful re-
search practice. But we hope CHIELD can help make re-
search more systematic. In particular, we are keen to
explore how the data in CHIELD can be used as prior
biases in more automated machine-learning processes,
such as automatic causal graph inference.
In conclusion, clearly expressing complex scientific
hypotheses is hard. There is a need for formal tools,
such as causal graphs, to help researchers communicate
their ideas. CHIELD provides easier access to these tools
and creates a space for clarifying theories. It demon-
strates that the field of evolutionary linguistics is more
connected than we thought, and there are potentially
many links between theories that are waiting to be dis-
covered. We suspect the same is true for many other
fields that would also benefit from building their own
database of causal graphs.
Supplementary data
Supplementary data is available at JOLEV online.
The following researchers contributed data to CHIELD without
contributing to this article: Lachlan Walmsley, Limor Raviv,
Andrew Buskell, Cara Evans, Michael Dunn, Robin Dunbar,
Isobel Clifford, James Winters, and Simon Kirby. Many thanks
to Robert Forkel, Tessa Alexander, Jon Hallett, and Simon
Greenhill for additional code review.
A.D., C.S., F.J., and E.T.C. were funded from the European
Research Council (ERC) under the European Union’s Horizon
2020 research and innovation programme (grant agreement no.
639291, Starting Grant VARIKIN). J.N. was supported by a
Principal’s Career Development Scholarship from the School of
Philosophy, Psychology & Language Sciences at Edinburgh.
K.S. was supported by an Academy of Finland grant (296212)
and by an ERC starting grant (805371). R.S. was supported by
the Australian Research Council (FL130100111). S.G.R. is sup-
ported by a Leverhulme early career fellowship (ECF-2016-
435). S.F.M. is supported by an Australian Government
Research Training Program (RTP) Scholarship and Australian
Research Council Laureate Fellowship Grant FL130100141.
C.S. was supported by the John Templeton Fund (40128).
Aalen, O. O. et al. (2016) ‘Can we Believe the DAGs? A
Comment on the Relationship between Causal DAGs and
Mechanisms’, Statistical Methods in Medical Research, 25/5:
Alshuwaier, F., Areshey, A., and Poon, J. (2017) ‘A
Comparative Study of the Current Technologies and
Approaches of Relation Extraction in Biomedical Literature
using Text Mining’. In 2017 4th IEEE International
Conference on Engineering Technologies and Applied
Sciences (ICETAS), 1–13.
Ardell, D., Anderson, N., and Winter, B. (2016) ‘Noise in
Phonology Affects Encoding Strategies in Morphology’, in S.
G. Roberts et al. (eds) The Evolution of Language:
Proceedings of the 11th International Conference
(EVOLANGX11). New Orleans: Evolang Scientific
Atkinson, M., Kirby, S., and Smith, K. (2015) ‘Speaker Input
Variability Does not Explain Why Larger Populations Have
Simpler Languages’, PloS One, 10/6: e0129463.
Asghar, N. (2016) ‘Automatic Extraction of Causal Relations
from Natural Language Texts: A Comprehensive Survey’, un-
defined website: </paper/Automatic-Extraction-of-Causal-
c74902d53f>accessed 15 April 2019.
Baayen, R. H., Milin, P., and Ramscar, M. (2016) ‘Frequency in
Lexical Processing’, Aphasiology, 30/11: 1174–220.
Bareinboim, E., and Pearl, J. (2012) ‘Causal Inference by
Surrogate Experiments: Z –Identifiability’ in de Freitas N. and
Murphy K. (eds) Proceedings of the Twenty-Eighth
Conference on Uncertainty in Artificial Intelligence, pp.
113–120. Corvalis, OR: AUAI Press.
, and (2016) ‘Causal Inference and the Data-Fusion
Problem’, Proceedings of the National Academy of Sciences,
113/27: 7345–52.
16 Journal of Language Evolution, 2020, Vol. 0, No. 0
Downloaded from by The Australian National University user on 21 April 2020
Bastian, M., Heymann, S., and Jacomy, M. (2009) ‘Gephi: An
Open Source Software for Exploring and Manipulating
Networks’, International AAAI Conference on Weblogs and
Social Media.
Bentz, C., and Winter, B. (2013) ‘Languages with More Second
Language Learners Tend to Lose Nominal Case’, Language
Dynamics and Change, 3/1: 1–27.
Berdicevskis, A., and Eckhoff, H. (2016) ‘Redundant Features
are Less Likely to Survive: Empirical Evidence From The
Slavic Languages’, in S. G. Roberts et al. (eds) The Evolution
of Language: Proceedings of the 11th International
Conference (EVOLANGX11). New Orleans: Evolang
Scientific Committee.
Bergmann, T., and Dale, R. (2016) ‘A Scientometric Analysis Of
Evolang: Intersections and Authorships’, in S. G. Roberts
et al. (eds) The Evolution of Language: Proceedings of the
11th International Conference (EVOLANG11). New
Orleans: Evolang Scientific Committee. Available online:
Berwick, R. C., and Chomsky, N. (2013). Birdsong, Speech, and
Language: Exploring the Evolution of Mind and Brain.
Cambridge, MA: MIT press.
, and (2016). Why Only Us: Language and
Evolution. Cambridge, MA: MIT press.
Blasi, D. E., and Roberts, S. G. (2017) ‘Beyond Binary
Dependencies in Language Structure’, in N.J. Enfield
(ed.) Dependencies in Language: On the Causal Ontology of
Linguistic Systems, Studies in Diversity Linguistics 14, pp.
117–128. Leipzig: Language Science Press.
Botha, R. P. (2003) Unravelling the Evolution of Language.
Leiden: Brill.
Bowern, C. (2015) ‘Linguistics: Evolution and Language
Change’, Current Biology, 25/1: R41–43.
Bybee, J. (2015) Language Change. Cambridge: Cambridge
University Press.
Cangelosi, A., and Parisi, D. (eds). (2012) Simulating
the Evolution of Language. London: Springer-Verlag.
Cao, M., Sun, X., and Zhuge, H. (2018) ‘The Contribution of
Cause-Effect Link to Representing the Core of Scientific
Paper—the Role of Semantic Link Network’, PloS One, 13/6:
Cheney, D. L., and Seyfarth, R. M. (2005) ‘Constraints and
Preadaptations in the Earliest Stages of Language Evolution’,
The Linguistic Review, 22/2–4: 135–59.
Christiansen, M. H., and Kirby, S. (eds). (2003). Language
Evolution. Oxford: OUP.
Coelho, M. T. P. et al. (2019) ‘Drivers of Geographical Patterns
of North American Language Diversity’, Proceedings of the
Royal Society B: Biological Sciences, 286/1899: 20190242.
Corballis, M. C. (1999) ‘The Gestural Origins of Language:
Human Language May Have Evolved from Manual Gestures,
Which Survive Today as a “Behavioral Fossil” Coupled to
Speech’, American Scientist, 87/2: 138–45.
Croft, W. (2008) ‘Evolutionary Linguistics’, Annual Review of
Anthropology, 37: 219–34.
Culbertson, J., and Newport, E. L. (2015) ‘Harmonic Biases in
Child Learners: In Support of Language Universals’,
Cognition, 139: 71–82.
Currie, A., and Killin, A. (2019) ‘From Things to Thinking:
Cognitive Archaeology’, Mind & Language, 34/2: 263–79.
Cuskley, C., and Loreto, V. (2016) ‘The Emergence of Rules and
Exceptions in a Population of Interacting Agents’, in S. G.
Roberts et al. (eds) The Evolution of Language: Proceedings
of the 11th International Conference (EVOLANGX11). New
Orleans: Evolang Scientific Committee.
Deacon, T. W. (1997) The Symbolic Species: the Co-Evolution
of Language and the Brain. New York: Norton.
De Boer, B., and Verhoef, T. (2012) ‘Language Dynamics in
Structured Form and Meaning Spaces’, Advances in Complex
Systems, 15/03n04: 1150021.
Dijkstra, E. W. (1959) ‘A Note on Two Problems in Connexion
with Graphs’, Numerische Mathematik, 1/1: 269–71.
Ding, P., and Miratrix, L. W. (2015) ‘To Adjust or Not to
Adjust? Sensitivity Analysis of M-Bias and Butterfly-Bias’,
Journal of Causal Inference, 3/1: 41–57.
Dor, D., Knight, C., and Lewis, J. (2014). The Social Origins of
Language (Vol. 19). Oxford: Oxford University Press.
Dunbar, R., and Dunbar, R. I. M. (1998). Grooming, Gossip,
and the Evolution of Language. Harvard: Harvard University
Dunbar, R. I. (2004) ‘Gossip in Evolutionary Perspective’,
Review of General Psychology, 8/2: 100–10.
Dunn, M. et al. (2011) ‘Evolved Structure of Language Shows
Lineage-Specific Trends in Word-Order Universals’, Nature,
473/7345: 79.
Easterday, M. W., Aleven, V., and Scheines, R. (2007) ‘Tis
Better to Construct than to Receive? The Effects of Diagram
Tools on Causal Reasoning’, Frontiers in Artificial
Intelligence and Applications, 158: 93.
et al. (2009) ‘Constructing Causal Diagrams to Learn
Deliberation’, International Journal of Artificial Intelligence
in Education, 19/4: 425–45.
Ellson, J. et al. (2004) Graphviz and Dynagraph—Static and
Dynamic Graph Drawing Tools. In Graph Drawing Software,
pp. 127–48. Berlin, Heidelberg: Springer.
Elwert, F. (2013) ‘Graphical Causal Models’, in Morgan S. L.
(ed.) Handbook of Causal Analysis for Social Research.
Dordrecht: Springer.
, and Winship, C. (2014) ‘Endogenous Selection Bias: The
Problem of Conditioning on a Collider Variable’, Annual
Review of Sociology, 40: 31–53.
Enard, W. et al. (2002) ‘Molecular Evolution of FOXP2, a Gene
Involved in Speech and Language’, Nature, 418/6900, 869.
Falk, D. (2016) ‘Evolution of Brain and Culture’, Journal of
Anthropological Sciences, 94: 1.
Fehe´ r, O. et al. (2017) ‘Statistical Learning in Songbirds: From
Self-Tutoring to Song Culture’, Philosophical Transactions of
the Royal Society B, 372: 1711.
Fitch, W. T. (2010). The Evolution of Language. Cambridge:
Cambridge University Press.
Journal of Language Evolution, 2020, Vol. 0, No. 0 17
Downloaded from by The Australian National University user on 21 April 2020
Gavin, M. C. et al. (2013) ‘Toward a Mechanistic
Understanding of Linguistic Diversity’, BioScience, 63/7:
Goldin-Meadow, S., and Yang, C. (2016) ‘Statistical Evidence
that a Child Can Create a Combinatorial Linguistic System
without External Linguistic Input: Implications for Language
Evolution’, Neuroscience & Biobehavioral Reviews, 81:
Granger, C. W. (2016) ‘Causal Inference’, in Palgrave
Macmillan (eds) The New Palgrave Dictionary of Economics.
London: Palgrave Macmillan.
Gumperz, J. (1968) ‘The Speech Community’ in Duranti, A.
(ed.) Linguistic Anthropology: A Reader 1: 66–73.
Guns, R., and Rousseau, R. (2014) ‘Recommending Research
Collaborations Using Link Prediction and Random Forest
Classifiers’, Scientometrics, 101/2: 1461–73.
Hitchcock, T. J., Paracchini, S., and Gardner, A. (2019)
‘Genomic Imprinting as a Window into Human Language
Evolution’, BioEssays, 41/6: 1800212.
Hauser, A., and Bu¨ hlmann, P. (2012) ‘Characterization and
Greedy Learning of Interventional Markov Equivalence
Classes of Directed Acyclic Graphs’, Journal of Machine
Learning Research, 13: 2409–64.
von Hardenberg, A., and Gonzalez-Voyer, A. (2013)
‘Disentangling Evolutionary Cause-Effect Relationships with
Phylogenetic Confirmatory Path Analysis’, Evolution, 67/2:
Heinze-Deml, C., Maathuis, M. H., and Meinshausen, N.
(2018) ‘Causal Structure Learning’, Annual Review of
Statistics and Its Application, 5: 371–91.
Ho¨ fler, M. et al. (2018) ‘Writing a Discussion Section: How to
Integrate Substantive and Statistical Expertise’, BMC Medical
Research Methodology, 18/1: 34.
Hsu, P. S. et al. (2015) ‘The Effect of a Graph-Oriented
Computer-Assisted Project-Based Learning Environment on
Argumentation Skills’, Journal of Computer Assisted
Learning, 31/1: 32–58.
Hurford, J. R. (2007) The Origins of Meaning: Language in the
Light of Evolution. Oxford: Oxford University Press.
Hurford, J. R. (2003) ‘Evolution of Language: Cognitive
Preadaptations’, in Strazny P. (ed.) Fitzroy Dearborn
Encyclopedia of Linguistics. Chicago: Fitzroy Dearborn
Hymes, D. H. (ed.). (1971) Pidginization and Creolization of
Languages. Cambridge: CUP Archive.
Irvine, L., Roberts, S., and Kirby, S. (2013) ‘A Robustness
Approach to Theory Building: A Case Study of Language
Evolution’, in Proceedings of the Annual Meeting of the
Cognitive Science Society (Vol. 35, No. 35).
Jon-And, A., and Aguilar, E. (2019) ‘A Model of
Contact-Induced Language Change: Testing the Role of
Second Language Speakers in the Evolution of Mozambican
Portuguese’, PloS One, 14/4: e0212303.
Jovani, R., and Mavor, R. (2011) ‘Group Size versus Individual
Group Size Frequency Distributions: A Nontrivial
Distinction’, Animal Behaviour, 82/5: 1027–36.
Kalisch, M. et al. (2012) ‘Causal Inference Using Graphical
Models with the R Package Pcalg’, Journal of Statistical
Software, 47/11: 1–26.
Kempe, V., Gauvrit, N., and Forsyth, D. (2015) ‘Structure
Emerges Faster during Cultural Transmission in Children than
in Adults’, Cognition, 136: 247–54.
Killworth, P. D., Bernard, H. R., and McCarty, C. (1984)
‘Measuring Patterns of Acquaintanceship’, Current
Anthropology, 25/4: 381–97.
Kinsella, A. R. (2009). Language Evolution and Syntactic
Theory. Cambridge: Cambridge University Press.
Kirby, S., and Christiansen, M. H. (2003) ‘Language Evolution:
The Hardest Problem in Science’, in S. Kirby and M.
Christiansen (eds) Language Evolution, pp. 1–15. Oxford:
Oxford University Press.
Knight, C., Power, C., and Watts, I. (1995) ‘The Human
Symbolic Revolution: A Darwinian account’, Cambridge
Archaeological Journal, 5/1: 75.
, Studdert-Kennedy, M., and Hurford, J. (eds). (2000) The
Evolutionary Emergence of Language: Social Function and the
Origins of Linguistic Form. Cambridge: Cambridge University
Kong, X. et al. (2017) ‘Exploring Dynamic Research Interest
and Academic Influence for Scientific Collaborator
Recommendation’, Scientometrics, 113/1: 369–85.
Koplenig, A. (2019) ‘Language Structure is Influenced by the
Number of Speakers but Seemingly Not by the Proportion of
Non-Native Speakers’, Royal Society Open Science, 6/2:
Labov, W. (1972). Sociolinguistic Patterns. Philadelphia:
University of Pennsylvania Press.
Lieberman, P. (1984) The Biology and Evolution of Language.
Harvard: Harvard University Press.
Little, H. (2012) ‘The Role of Foreigner-Directed Speech in
Language Evolution’, in J. Kendal, R. Kendal, J. Tehrani and
L. Boothroyd (eds) Proceedings of the European Human
Behaviour and Evolution Association Annual Meeting 2012,
pp. 7–8. Durham: EHBEA.
Lopes, G. R. et al. (2010) ‘Collaboration Recommendation on
Academic Social Networks’, in International Conference on
Conceptual Modeling, pp. 190–99. Berlin, Heidelberg:
Lupyan, G., and Dale, R. (2010) ‘Language Structure is Partly
Determined by Social Structure’, PloS One, 5/1: e8559.
Mahoney, J. (2008) ‘Toward a Unified Theory of Causality’,
Comparative Political Studies, 41/4–5: 412–36.
Majid, A., Jordan, F., and Dunn, M. (2015) ‘Semantic
Systems in Closely Related Languages’, Language Sciences,
49: 1–18.
McGrath, E. (1984) Groups: Interaction and Performance,pp.
61–62. Englewood Cliffs, NJ: Prentice-Hall.
Middleton, J. A. et al. (2016) ‘Bias Amplification and Bias
Unmasking’, Political Analysis, 24/3: 307–23.
Morgan, M. S. (2013) ‘Nature’s Experiments and Natural
Experiments in the Social Sciences’, Philosophy of the Social
Sciences, 43/3: 341–57.
18 Journal of Language Evolution, 2020, Vol. 0, No. 0
Downloaded from by The Australian National University user on 21 April 2020
Morgan, M. H. (2014). Speech Communities. Cambridge:
Cambridge University Press.
Mueller, R. M., and Huettemann, S. (2018) Extracting Causal
Claims from Information Systems Papers with Natural
Language Processing for Theory Ontology Learning. https://
Mueller, R., and Abdullaev, S. (2019) DeepCause:
Hypothesis Extraction from Information Systems Papers
with Deep Learning for Theory Ontology Learning. Retrieved
Mufwene, S. S. (2001) The Ecology of Language Evolution.
Cambridge: Cambridge University Press.
Murdock, G. P., and Wilson, S. F. (1972) ‘Settlement Patterns
and Community Organization: Cross-Cultural Codes 3’,
Ethnology, 11/3: 254–95.
Nettle, D. (2012) ‘Social Scale and Structural Complexity in
Human Languages’, Philosophical Transactions of the Royal
Society B: Biological Sciences, 367/1597: 1829–36.
Noble, W., and Davidson, I. (1996) ‘Human Evolution,
Language and Mind: A Psychological and Archaeological
Inquiry’, Cambridge: CUP Archive.
Nowak, M. A., and Krakauer, D. C. (1999) ‘The Evolution of
Language’, Proceedings of the National Academy of Sciences,
96/14: 8028–33.
Pakendorf, B. (2014) ‘Coevolution of Languages and Genes’,
Current Opinion in Genetics & Development, 29: 39–44.
Pearl, J. (2000) Causality: Models, Reasoning, and Inference
(Vol. 47). Cambridge: Cambridge University Press.
, and Mackenzie, D. (2018) The Book of Why: The New
Science of Cause and Effect. London: Allen Lane.
Piantadosi, S. T., and Fedorenko, E. (2017) ‘Infinitely
Productive Language Can Arise from Chance under
Communicative Pressure’, Journal of Language Evolution,
2/2: 141–7.
Power, C., Finnegan, M., and Callan, H. (2016) Human
Origins: Contributions from Social Anthropology. Oxford:
Progovac, L. (2015) Evolutionary Syntax. Oxford: Oxford
University Press.
(2019) A Critical Introduction to Language Evolution:
Current Controversies and Future Prospects. New York:
Springer International Publishing.
Pyers, J. E. et al. (2010) ‘Evidence from an Emerging Sign
Language Reveals that Language Supports Spatial Cognition’,
Proceedings of the National Academy of Sciences, 107/27:
Ritt, N. (2004) Selfish Sounds and Linguistic Evolution: A
Darwinian Approach to Language Change. Cambridge:
Cambridge University Press.
Roberts, G. (2010) ‘An Experimental Study of Social Selection
and Frequency of Interaction in Linguistic Diversity’,
Interaction Studies, 11/1: 138–59.
Roberts, S. G. (2018) ‘Robust, Causal, and Incremental
Approaches to Investigating Linguistic Adaptation’, Frontiers
in Psychology, 9: 166.
, and Winters, J. (2013) ‘Linguistic Diversity and Traffic
Accidents: Lessons from Statistical Studies of Cultural Traits’,
PloS One, 8/8: e70902.
Rogerson, P. A. (1997) ‘Estimating the Size of Social Networks’,
Geographical Analysis, 29/1: 50–63
Rohrer, J. M. (2018) ‘Thinking Clearly about Correlations and
Causation: Graphical Causal Models for Observational Data’,
Advances in Methods and Practices in Psychological Science,
1/1: 27–42.
Ross, M. D. (1996) ‘Contact-Induced Change and the
Comparative’, in Durie, M., and Ross, M. D. (eds) The
Comparative Method Reviewed: Regularity and Irregularity
in Language Change, p. 180. Oxford: Oxford University Press
on Demand.
Rosseel, Y. (2012) ‘Lavaan: An R Package for Structural
Equation Modeling’, Journal of Statistical Software, 48/2:
Rudnicki, K., De Backer, C. J., and Declerck, C. (2019) ‘The
Effects of Celebrity Gossip on Trust Are Moderated by
Prosociality of the Gossipers’, Personality and Individual
Differences, 143: 42.
Sampson, G., Gil, D., and Trudgill, P. (eds). (2009) Language
Complexity as an Evolving Variable (Vol. 13). Oxford:
Oxford University Press.
Scott-Phillips, T. C., and Kirby, S. (2010) ‘Language
Evolution in the Laboratory’, Trends in Cognitive Sciences,
14/9: 411–7.
Sinnema¨ ki, K. (2010) ‘Word Order in Zero-Marking
Languages’, Studies in Language, 34/4: 869–912.
Slocombe, K. E., and Zuberbu¨ hler, K. (2005) ‘Functionally
Referential Communication in a Chimpanzee’, Current
Biology, 15/19: 1779–84.
Smith, K., and Kirby, S. (2008) ‘Cultural Evolution:
Implications for Understanding the Human Language
Faculty and its Evolution’, Philosophical Transactions
of the Royal Society B: Biological Sciences, 363:
Spike, M. (2018) Language complexity as an interaction be-
tween social structure, innovation, and simplicity,
Proceedings of the 12th International Conference on the
Evolution of Language, Torun: NCU press.
Steels, L. (1997) ‘The Synthetic Modeling of Language Origins’,
Evolution of Communication, 1/1: 1–34.
Tallerman, M. (2007) ‘Did Our Ancestors Speak a Holistic
Protolanguage?’, Lingua, 117/3: 579–604.
Tamariz, M., and Kirby, S. (2016) ‘The Cultural Evolution of
Language’, Current Opinion in Psychology, 8: 37–43.
Textor, J., Hardt, J., and Knu¨ ppel, S. (2011) ‘DAGitty: A
Graphical Tool for Analyzing Causal Diagrams’,
Epidemiology, 22/5: 745.
Thurston, W. (1989) ‘How Exoteric Languages Build a Lexicon:
Esoterogeny in Western New Britain’, in Harlow, R. and
Hooper, R. (eds) VICAL 1: Oceanic Languages, Papers from
the Fifth International Conference on Austronesian
Linguistics, pp. 555–79. Auckland: Linguistic Society of New
Journal of Language Evolution, 2020, Vol. 0, No. 0 19
Downloaded from by The Australian National University user on 21 April 2020
Tshitoyan, V. et al. (2019) ‘Unsupervised Word Embeddings
Capture Latent Knowledge from Materials Science
Literature’, Nature, 571/7763: 95.
Tubau, E. (2008) ‘Enhancing Probabilistic Reasoning: The Role
of Causal Graphs, Statistical Format and Numerical Skills’,
Learning and Individual Differences, 18/2: 187–96.
VanderWeele, T. J., and Robins, J. M. (2007) ‘Directed Acyclic
Graphs, Sufficient Causes, and the Properties of Conditioning
on a Common Effect’, American Journal of Epidemiology,
166/9: 1096–104.
, and (2010) ‘Signed Directed Acyclic Graphs for
Causal Inference’, Journal of the Royal Statistical Society:
Series B (Statistical Methodology), 72/1: 111–27.
van der Bijl, W. (2018) ‘Phylopath: Easy Phylogenetic Path
Analysis in R’, PeerJ, 6: e4718.R package version 1.0.2.
Vernes, S. C. (2017) ‘What Bats Have to Say about Speech and
Language’, Psychonomic Bulletin & Review, 24/1: 111–7.
Vigliocco, G., Perniss, P., and Vinson, D. (2014) ‘Language as a
Multimodal Phenomenon: implications for Language
Learning, Processing and Evolution’, Philosophical
Transactions of the Royal Society B-Biological Sciences,
369/2014: 1–7
Vogt, P., and De Boer, B. (2010) ‘Editorial: Language Evolution:
Computer Models for Empirical Data’, Adaptive
Behavior-Animals, Animats, Software Agents, Robots,
Adaptive Systems, 18/1: 5–11.
Westfall, J., and Yarkoni, T. (2016) ‘Statistically Controlling for
Confounding Constructs is Harder than You Think’, PloS
One, 11/3: e0152719.
Wiley, J., and Myers, J. L. (2003) ‘Availability and Accessibility
of Information and Causal Inferences from Scientific Text’,
Discourse Processes, 36/2: 109–29.
Woodward, J. (2003) Making Things Happen: A Theory of
Causal Explanation. Oxford: Oxford University Press.
(2016) ‘Causation and manipulability’, in Zalta E. N.
(ed.) Stanford Encyclopedia of Philosophy (winter 2016 edi-
Wray, A. (2002) Transition to Language. Oxford: Oxford
University Press.
, and Grace, G. W. (2007) ‘The Consequences of Talking
to Strangers: Evolutionary Corollaries of Socio-Cultural
Influences on Linguistic Form’, Lingua, 117/3: 543–78.
Xu, Y. et al. (2010) ‘Network Based Approach for Discovering
Academic Researchers with Shared Interests’, in 2010
International Conference on E-Business and E-Government,
pp. 1864–7. Guangzhou: IEEE Computer.
Yan, E., and Guns, R. (2014) ‘Predicting and Recommending
Collaborations: An Author-, Institution-, and Country-Level
Analysis’, Journal of Informetrics, 8/2: 295–309.
York, R. (2018) ‘Control Variables and Causal Inference: A
Question of Balance’, International Journal of Social Research
Methodology, 21/6: 675–84.
Youngblood, M., and Lahti, D. (2018) ‘A
Bibliometric Analysis of the Interdisciplinary Field
of Cultural Evolution’, Palgrave Communications,4/1:120.
Zuidema, W., and de Boer, B. (2010) ‘Models of
Language Evolution: Does the Math Add Up?’
Models of Language Evolution Workshop
at EVOLANG 2010, April 14, 2010, Utrecht, the Netherlands.
Zuidema, W., and (2013) ‘Modeling in the Language
Sciences’, in Podesva R. J. and Sharma D. (eds) Research
Methods in Linguistics. Cambridge: Cambridge University
Press, pp. 428–445.
20 Journal of Language Evolution, 2020, Vol. 0, No. 0
Downloaded from by The Australian National University user on 21 April 2020
... Fitch 2017, Benítez-Burraco & Progovac 2020) that outline important aspects of 'language' and 'language evolution' and eschew the kind of restrictive definitions outlined in Section 3. As such, these approaches are further evidence of our view that the definitional discrepancies discussed above did not hinder progress in the field. Instead, these approaches show the importance not of top-down definitions of language, but of focusing on particular aspects of and hypotheses about language and investigating their relation to other factors relevant to language and its evolution (see also Roberts et al. 2020). ...
... An epitome of both the empirical and bottom-up approach is the Causal Hypotheses In Evolutionary Linguistics Database (CHIELD, pronounced 'shield'; Roberts et al. 2020). CHIELD contains crowd-sourced entries for over 400 publications, with over 3,400 causal links between more than 1,700 variables and aims not only at cataloguing hypotheses about language evolution but also making data on them interoperable. ...
Full-text available
Many controversies in language evolution research derive from the fact that language is itself a natural language word, which makes the underlying concept fuzzy and cumbersome, and a common perception is that progress in language evolution research is hindered because researchers do not ‘talk about the same thing’. In this article, we claim that agreement on a single, top-down definition of language is not a sine qua non for good and productive research in the field of language evolution. First, we use the example of the notion FLN (‘faculty of language in the narrow sense’) to demonstrate how the specific wording of an important top-down definition of (the faculty of) language can—surprisingly—be inconsequential to actual research practice. We then review four approaches to language evolution that we estimate to be particularly influential in the last decade. We show how their breadth precludes a single common conceptualization of language but instead leads to a family resemblance pattern, which underwrites fruitful communication between these approaches, leading to cross-fertilisation and synergies.
... For example, using quantitative methods in themselves does not automatically address the problem that multiple explanations and hypotheses might be compatible with the observed historical data (Jenset and McGillivray, 2017, p. 47). Roberts et al. (2020) present an interesting supporting tool to deal with this problem, which we find interesting and encouraging, but insufficient on its own to address this problem. Instead, we will probably need an even closer alignment of theory, hypotheses, data, and methods. ...
Full-text available
Historical linguistics is the study of language change and stability, of the history of individual languages, and of the relatedness between languages. In spite of numerous acknowledgements, the adoption of quantitative methods in historical linguistics is still far from being mainstream and it falls below the level of other branches of linguistics. This comment considers the adoption of quantitative methods in recent historical linguistics research, and compares a study on 2012 publications with a similar study conducted seven years later. This comment argues for the advantages of a wider adoption of quantitative methods among historical linguists, and considers various reasons for the relatively slow progress in this direction. It also clarifies when quantitative methods are not the preferred route.
... Since there is a large variety of criticality phenomena, we propose the examination of interlocking of causal models for different criticality phenomena as future work. One could imagine that the causal models might be stored in a connected database, thus sharing knowledge of the possibly relevant synergies, similar to the CHIELD database for causal hypotheses in evolutionary linguistics [45]. ...
Full-text available
The verification and validation of automated driving systems at SAE levels 4 and 5 is a multi-faceted challenge for which classical statistical considerations become infeasible. For this, contemporary approaches suggest a decomposition into scenario classes combined with statistical analysis thereof regarding the emergence of criticality. Unfortunately, these associational approaches may yield spurious inferences, or worse, fail to recognize the causalities leading to critical scenarios, which are, in turn, prerequisite for the development and safeguarding of automated driving systems. As to incorporate causal knowledge within these processes, this work introduces a formalization of causal queries whose answers facilitate a causal understanding of safety-relevant influencing factors for automated driving. This formalized causal knowledge can be used to specify and implement abstract safety principles that provably reduce the criticality associated with these influencing factors. Based on Judea Pearl's causal theory, we define a causal relation as a causal structure together with a context, both related to a domain ontology, where the focus lies on modeling the effect of such influencing factors on criticality as measured by a suitable metric. As to assess modeling quality, we suggest various quantities and evaluate them on a small example. As availability and quality of data are imperative for validly estimating answers to the causal queries, we also discuss requirements on real-world and synthetic data acquisition. We thereby contribute to establishing causal considerations at the heart of the safety processes that are urgently needed as to ensure the safe operation of automated driving systems.
... • Systematic comparisons using general principles (Turner & Maryanski, 2008;Cavalli-Sforza, 1997;Richerson & Boyd, 2005;Irvine et al., 2013;Dediu et al., 2013;Dor & Jablonka, 2014;Donald, 2001Donald, , 2016Donald, , 2017Roberts et al., 2020). ...
Full-text available
Studies of the origin and early evolution of language are undergoing rapid development. New directions, topics and methods of research are appearing. It becomes necessary to structure the research field, as the formation of a single space of intellectual attention contributes to a more productive communication between different groups of scientists, better mutual understanding, and strengthened arguments. The article proposes the grounds for typologizing the concepts of the origin of language and identifies three main paradigms on the basis of these typologies. The sharpest disputes are between Continualism (“human language is only quantitatively different from animal language”) and Saltationism (“human language is so fundamentally different from the communicative systems of animals, that it could only appear as a result of some amazing leap or unique mutation”). Continualists usually focus on the communicative and cognitive abilities of different species (with a clear preference for apes). Saltationists focus on the linguistic, mental and cognitive structures of humans. Both of these polar approaches are opposed by the Multistage ecosocial paradigm, which emphasizes a long, multistep process of glottogenesis with behavioral adaptations of hominid groups to the changing natural and social conditions of their existence. A list of the most plausible, theoretically substantiated propositions is given, as well as research results supported by a variety of circumstantial data. The theoretical and methodological perspectives of the multistage ecosocial paradigm related to overcoming from its drawbacks are presented. A generalizing conceptual framework corresponding to the basic principles of evolution, the laws of social interaction should be constructed. The extremely heterogeneous arsenal of methods should also be structured in a single scientific logic.
... However, we also hold that the claim that we "know essentially nothing about the evolution of our cognitive capabilities" (Lewontin 1998: 111) is just as subjective and impossible to quantify and measure. Here we find that landmark publications such as Hurford (2007Hurford ( , 2012, Fitch (2010), Tallerman & Gibson (2012, or the launch of the CHIELD database (Roberts et al. 2020) are testament to the wealth of evidence, results, and knowledge that has been accumulated in language evolution research, which can hardly be said to amount to "essentially nothing." ...
Full-text available
The question whether all languages are similarly complex is at the centre of some of the most heated debates within linguistics. These debates focus on such issues as the universality of syntactic recursion, the exceptional simplicity of creole languages, complexity trade-offs between structural levels, as well as sociolinguistic correlates of complexity profiles. Discussions concerning complexity have implications that go far beyond linguistics in the narrow sense, including e.g. the role of nature vs. nurture in human cognition and culture, or the distinction between message and noise in information theory. In consequence, debates on linguistic complexity shape our perception of human nature and variation among human populations. In this Research Topic, we investigate the motivations driving the research on linguistic complexity. Thus, Menzerath’s law about complexity trade-offs was inspired by bottom-up empirical observations. By contrast, the claim about the universality of syntactic recursion was primarily informed by theoretical considerations. Due to its normative dimension, the notion of complexity has also served as a vehicle for advancing ideological agendas, such as characterizing speakers as more or less advanced based on perceived properties of their languages. By bringing these perspectives together, we contribute to a critical assessment of how linguistic research is motivated by both epistemic and non-epistemic goals.
Striking variation exists in preferences for specific spatial linguistic strategies among different speech communities. Increasing evidence now suggests that this might not simply be a result of neutral drift, but rather a form of linguistic adaptation to the local social, cultural, or physical environment. Recent studies indicate that different factors like topography, subsistence style, and bilingualism successfully predict the choice of spatial frames of reference (FoR) on linguistic and non-linguistic tasks. However, the exact causal relationships between these variables and the cultural evolutionary mechanisms behind the selection of one FoR strategy over another are still not fully understood. In this paper, we argue that to arrive at a more mechanistic and causal understanding of the cultural evolution of spatial language, observations from descriptive fieldwork should be combined with experimental and computational methods. In the framework we present, causal relationships between linguistic and non-linguistic variables (such as FoR choice and topography) can be isolated and systematically tested in order to shed light on how sociotopographic factors motivate the variation in spatial language we observe cross-linguistically. We discuss experimental results from behavioral studies and computer simulations that illustrate how this approach can deliver empirical findings that go beyond simple correlations.
Full-text available
It is often claimed that languages with more non-native speakers tend to become morphologically simpler, presumably because non-native speakers learn the language imperfectly. A growing number of studies support this claim, but there is a dearth of experiments that evaluate it and the suggested explanatory mechanisms. We performed a large-scale experiment which directly tested whether imperfect language learning simplifies linguistic structure and whether this effect is amplified by iterated learning. Members of 45 transmission chains, each consisting of 10 one-person generations, learned artificial mini-languages and transmitted them to the next generation. Manipulating the learning time showed that when transmission chains contained generations of imperfect learners, the decrease in morphological complexity was more pronounced than when the chains did not contain imperfect learners. The decrease was partial (complexity did not get fully eliminated) and gradual (caused by the accumulation of small simplifying changes). Simplification primarily affected double agent-marking, which is more redundant, arguably more difficult to learn and less salient than other features. The results were not affected by the number of the imperfect-learner generations in the transmission chains. Thus, we provide strong experimental evidence in support of the hypothesis that iterated imperfect learning leads to language simplification.
Diachronic studies have played an increasingly important role in recent Cognitive Linguistics. This introductory paper provides an overview of some major lines of research in this field, starting with the inherently panchronic approach that characterizes most flavors of usage-based theory from Cognitive Grammar to recent complex adaptive systems approaches. In particular, the “constructionist turn” and the “quantitative turn” in Diachronic Cognitive Linguistics are discussed in detail. Diachronic Cognitive Linguistics is introduced as a multi-faceted, dynamic framework that aims at providing a holistic and nuanced picture of the complex interplay between language, cognition, and cultural evolution. In addition, this paper introduces the contributions to the present volume in some detail and discusses their relation to current research trends and paradigms within the broader framework of Diachronic Cognitive Linguistics.
Full-text available
The overwhelming majority of scientific knowledge is published as text, which is difficult to analyse by either traditional statistical analysis or modern machine learning methods. By contrast, the main source of machine-interpretable data for the materials research community has come from structured property databases1,2, which encompass only a small fraction of the knowledge present in the research literature. Beyond property values, publications contain valuable knowledge regarding the connections and relationships between data items as interpreted by the authors. To improve the identification and use of this knowledge, several studies have focused on the retrieval of information from scientific literature using supervised natural language processing3–10, which requires large hand-labelled datasets for training. Here we show that materials science knowledge present in the published literature can be efficiently encoded as information-dense word embeddings11–13 (vector representations of words) without human labelling or supervision. Without any explicit insertion of chemical knowledge, these embeddings capture complex materials science concepts such as the underlying structure of the periodic table and structure–property relationships in materials. Furthermore, we demonstrate that an unsupervised method can recommend materials for functional applications several years before their discovery. This suggests that latent knowledge regarding future discoveries is to a large extent embedded in past publications. Our findings highlight the possibility of extracting knowledge and relationships from the massive body of scientific literature in a collective manner, and point towards a generalized approach to the mining of scientific literature.
Full-text available
Humans spend large portions of their time and energy talking to one another, yet it remains unclear whether this activity is primarily selfish or altruistic. Here, it is shown how parent‐of‐origin specific gene expression—or “genomic imprinting”—may provide an answer to this question. First, it is shown why, regarding language, only altruistic or selfish scenarios are expected. Second, it is pointed out that an individual's maternal‐origin and paternal‐origin genes may have different evolutionary interests regarding investment into language, and that this intragenomic conflict may drive genomic imprinting which—as the direction of imprint depends upon whether investment into language is relatively selfish or altruistic—may be used to discriminate between these two possibilities. Third, predictions concerning the impact of various mutations and epimutations at imprinted loci on language pathologies are derived. In doing so, a framework is developed that highlights avenues for using intragenomic conflicts to investigate the evolutionary drivers of language. Patterns of genomic imprinting, and the intragenomic conflicts that drive them, can provide new, underexplored, empirical tests of the social evolutionary drivers of human language.
Full-text available
Language change is accelerated by language contact, especially by contact that occurs when a group of speakers shifts from one language to another. This has commonly been explained by linguistic innovation occurring during second language acquisition. This hypothesis is based on historical reconstructions of instances of contact and has not been formally tested on empirical data. In this paper, we construct an agent-based model to formalize the hypothesis that second language speakers are responsible for accelerated language change during language shift. We compare model predictions to a unique combination of diachronic linguistic and demographic data from Maputu, Mozambique. The model correctly predicts an increased proportional use of the novel linguistic variants during the period we study. We find that a modified version of the model is a better fit to one of our two datasets and discuss plausible reasons for this. As a general conclusion concerning typological differences between contact-induced and non-contact-induced language change, we suggest that multiple introductions of a new linguistic variant by different individuals may be the mechanism by which language contact accelerates language change. © 2019 Jon-And, Aguilar. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Full-text available
Although many hypotheses have been proposed to explain why humans speak so many languages and why languages are unevenly distributed across the globe, the factors that shape geographical patterns of cultural and linguistic diversity remain poorly understood. Prior research has tended to focus on identifying universal predictors of language diversity, without accounting for how local factors and multiple predictors interact. Here, we use a unique combination of path analysis, mechanistic simulation modelling, and geographically weighted regression to investigate the broadly described, but poorly understood, spatial pattern of language diversity in North America. We show that the ecological drivers of language diversity are not universal or entirely direct. The strongest associations imply a role for previously developed hypothesized drivers such as population density, resource diversity, and carrying capacity with group size limits. The predictive power of this web of factors varies over space from regions where our model predicts approximately 86% of the variation in diversity, to areas where less than 40% is explained.
Full-text available
Large-scale empirical evidence indicates a fascinating statistical relationship between the estimated number of language users and its linguistic and statistical structure. In this context, the linguistic niche hypothesis argues that this relationship reflects a negative selection against morphological paradigms that are hard to learn for adults, because languages with a large number of speakers are assumed to be typically spoken and learned by greater proportions of adults. In this paper, this conjecture is tested empirically for more than 2000 languages. The results question the idea of the impact of non-native speakers on the grammatical and statistical structure of languages, as it is demonstrated that the relative proportion of non-native speakers does not significantly correlate with either morphological or information-theoretic complexity. While it thus seems that large numbers of adult learners/speakers do not affect the (grammatical or statistical) structure of a language, the results suggest that there is indeed a relationship between the number of speakers and (especially) information-theoretic complexity, i.e. entropy rates. A potential explanation for the observed relationship is discussed.
This major 2001 work explores the development of creoles and other new languages, focusing on the conceptual and methodological issues they raise for genetic linguistics. Written by an internationally renowned linguist, the book discusses the nature and significance of internal and external factors or 'ecologies' that bear on the evolution of a language. The book surveys a wide range of examples of changes in the structure, function and vitality of languages, and suggests that similar ecologies have played the same kinds of roles in all cases of language evolution. Drawing on major theories of language formation, macroecology and population genetics, Mufwene proposes a common approach to the development of creoles and other new languages. The Ecology of Language Evolution will be welcomed by students and researchers in sociolinguistics, creolistics, theoretical linguistics and theories of evolution.