Search

A foundational goal of linguistics has been to understand why languages look the way they do. A range of possible explanations exist – from domain-specific representations to cognition-external factors like history and grammaticalization – and all of these undoubtedly play some role. But determining exactly how these link with specific features of language remains challenging, and the role of domain-specific mechanisms has been particularly contentious. In this paper, I highlight a growing new approach, which uses artificial language experiments to link individual-level biases to cross-linguistic trends in language structure. Using word and morpheme order as case studies, I will show how a range of different paradigms and learner populations allow us to make progress on this crucial issue in linguistics. I will focus on typological trends in word and morpheme order. For some ordering trends, experimental evidence points to variation across populations, suggesting that the best explanation for these patterns likely has its root in language history and grammaticalization. In other cases, the evidence points to the role of universal but domain-general cognitive biases, like transparency and simplicity. These domain-general biases interact with linguistic representations in important ways. Taken together these studies help adjudicate between alternative explanations for a number of specific ordering patterns and suggest a new sense of domain-specificity in the evolution of language.

Collocations, semi-productive lexical combinations with onefigurative and one literal word, are said to be a “pain in theneck” for researchers and L2 learners. The present study aims:(i) to conceptually replicate the processing costs incurred byL1 speakers when processing collocations using a larger andmore diverse set of items, (ii) to use literalness judgementsto test whether L1 speakers are aware of the semi-transparentmeaning of a collocation, and (iii) to test whether the presence of processing costs associated with collocations can be predicted from literalness judgements. If so, we propose that literalness judgements could be used as a diagnostic for reli-ably identifying collocations. We replicate the L1 processingcosts with a larger stimulus set and demonstrate that speakersare aware of the semi-transparent meaning of the collocation.We further show that L1 speaker judgements about the literal-ness of a word combination can be used to predict its status as a collocation.

Animacy is prevalent in as a semantic basis for noun classification systems (i.e., grammatical gender, noun classes and classifiers), but colour is completely absent, despite its visual salience. The absence of colour in such systems is sometimes argued to suggest domain-specific constraints on what is grammatically encodable. Here, we investigate whether this tendency could instead be explained by the superior predictive power of animacy (i.e., the degree to which it predicts other features) compared to colour. In a series of experiments, we find that animacy-based noun classes are learned better than colour-based ones. However, when participants are encouraged, by manipulating predictive power, to sort images based on colour, they are worse at learning animacy-based noun classes. The results suggest the animacy bias in grammar may have its roots in domain-general categorisation principles. They further serve as evidence for the role of cognitive biases in constraining cross-linguistic variation.

Agreement, a systematic formal mapping between linguistic elements, adds redundant complexity to languages (e.g., in ‘she writes' the -s adds no information), and yet is crosslinguistically prevalent. A prominent hypothesis argues that the ubiquity of agreement may be due to a functional advantage it confers for child learners. Here, we test this using an artificial language learning experiment with 56 English-speaking children (mean age 5;11). We investigate whether agreement can facilitate learning of noun classes (e.g., ‘masculine'/'feminine'). In one condition, agreement appeared as a redundant cue to noun classes, whereas in the other condition there was no agreement. Following exposure, we tested children on noun classification for both nouns they were trained on and novel nouns. Results reveal that children classified nouns equally well in both conditions. However, novel nouns were classified better in the agreement condition compared to the no-agreement condition, suggesting agreement can facilitate generalization for child learners.

Cross-linguistically, lexicons tend to be more phonetically clustered than required by their phonotactics; that is, words are less distinct than they could be. We use an agent-based exemplar model to investigate how this property arises over generations of language transmission under different functional pressures from learning and communication. We find that, in isolation, learnability pressures rapidly give rise to maximally clustered lexicons. When communicative pressures are also at play, clustering increases in line with a producer-side pressure to maximise similarity between words, but the rate of change is modulated by a listener-side preference for dispersion of word forms: a speaker who is trying to be understood considers what the listener is likely to understand before choosing a word to send. Overall, this work sheds light on how organisational properties of the lexicon may arise as a result of an ongoing trade-off between pressures from language learning, production and comprehension.

Generalisation of a linguistic rule can be facilitated by certain distributional characteristics. Previous work has shown that a rule is better generalised if it applies to items that (i) follow a skewed frequency distribution, or (ii) follow a uniform frequency distribution over many distinct item types. These two observations cannot be unified under explanations of rule generalisation that are based on entropy of the frequency distributions (since skewed distributions have low entropy, while a greater type count increases the entropy), nor explanations that focus on one highly-frequent type providing a basis for analogical extension (since all types in uniform distributions are equally frequent). Using an artificial language learning experiment and an agent-based model, we show that participants' generalisation behaviour is best matched by a model encoding preferential generalisation of rules containing long-tailed distributions—that is, containing a greater number of low-frequency types.

Languages seem to be designed for efficient communication. For example, shorter forms are used for more predictable meanings, a tendency argued to stem from speakers’ efficient language use. However, no study to date has systematically tested whether communicative efficiency shapes children’s language use. Investigating whether such a pressure is already present in children will shed light on the development of children’s’ communicative behaviour and the respective roles of adults and children in shaping language structure. Here, we investigate the development of communicative efficiency using a novel experimental paradigm with children ages 4-10. Results show that communicative efficiency is attested already in young children and becomes more adult-like with age: as children grow, they are more likely to shorten messages (minimize effort) when a short message is sufficient for accurate communication. We discuss the implications of our results for cognitive development and for theories of language evolution and change.

Languages exhibit a tremendous amount of variation in how they organise and order morphemes within words; however, regularities are also found. For example, gender and number inflectional morphology tend to appear together within a single affix, and when they appear in two separate affixes, gender marking tends to be placed closer to the stem than number. Formal theories of gender and number have been designed (in part) to explain these tendencies. However, determining whether the abstract representations hypothesised by these theories indeed drive the patterns we find cross-linguistically is difficult, if not impossible, based on the natural language data alone. In this study we use an artificial language learning paradigm to test whether the inferences learners make about the order of gender and number affixes—in the absence of any explicit information in the input—accord with formal theories of how they are represented. We test two different populations, English and Italian speakers, with substantially differ- ent gender systems in their first language. Our results suggest a clear preference for placing gender closest to the noun across these populations, across different types of gender systems, and across prefixing and suffixing morphology. These results expand the range of behavioural evidence for the role of cognitive representations in determining morpheme order.

Across many semantic domains, cross-linguistic regularities in categorization systems (e.g., color or kinship terms) have been taken to reflect constraints on how humans perceive and conceptualize the world. Such conceptual representations are often assumed to be universal, and independent of an individual's experience with a particular language. However, in most cases, representational constraints have not been observed empirically on language-independent grounds. This study comes to fill in this gap. We use a card sorting task to provide the first empirical evidence for a common, language-independent representation of pronominal referents, shared by speakers of different languages.

Numeral classifiers are a noun categorization device found in many East and Southeast Asian languages. Their syntactic status is debated: some theories treat them as forming a constituent with the noun, others with the numeral. Typological data on how classifiers, numerals, and nouns are ordered largely support the classifier-with-numeral hypothesis: orders in which the noun intervenes between the numeral and the classifier are unattested. However, typological evidence is problematic in this case, because the distribution of classifiers is highly geographically constrained, and ordering tendencies may reflect language contact. In this study, we look for evidence of the syntactic status of classifiers using artificial language learning experiments. We test whether learners’ preference among patterns are predicted by one of these two accounts. Results suggest that, if anything, learners in fact prefer orders that group classifiers with the numeral.

Active production helps L1 and L2 learners to notice linguistic generalisations better than passive learning does. Here, we explore whether a production task encourages learners to identify less obvious generalisations when more than one analysis is possible. We taught participants an artificial language compatible with two grammatical analyses: one that marks semantic roles using word order, the other using case suffixes. After training, half of the participants did a comprehension task, and the other half did a production task. We predicted that the comprehension group would adopt the English-like word order analysis, while the production group would be more likely to learn the recurring case suffixes and thus adopt the alternative analysis. Although this prediction was not supported, exploratory analysis suggests that production did increase participants' certainty about their chosen analysis. We suspect that participants arrived at the familiar word order analysis first and then did not consider alternative hypotheses.

In natural languages and in experimental studies of artificial language learning, case marking of grammatical arguments is more likely to be used in languages with flexible word order due to an efficiency trade-off between production effort and communicative accuracy. However, experimental evidence suggests that language learners are less efficient when there is a social bias in favour of a group whose productions are inefficient. Here, we examine the impact of autistic traits on efficient communication. We find that autistic people's use of case in the absence of a social bias is comparative to their neurotypical peers. However, we also find evidence that autistic people adhere more to social biases; they increase production effort in order to behave more like the group they are biased towards. We argue that some autistic people may be more likely to adhere to a social bias as a result of learnt social behaviours. More generally, these results underscore the importance of studying more diverse populations in language evolution research.

Agreement, a systematic formal mapping between linguistic elements, adds redundancy to languages (e.g., in ‘she writes’ the -s adds no information), and yet is crosslinguistically prevalent. Here, we suggest that agreement may be functionally advantageous by providing additional cues for language learning. We conducted an artificial language learning experiment to test whether agreement, and especially, alliterative agreement – where the agreement is expressed by repetition of the same form, can facilitate learning of noun classes (e.g., ‘masculine’/’feminine’). To this end, we compared the learnability of noun class systems in three input conditions: no agreement, alliterative agreement, and non-alliterative agreement. We found that participants who learned the non-alliterative agreement were better at generalizing the noun class system to novel nouns with the relevant semantic features. There was no difference between the alliterative and the no-agreement conditions, suggesting that the possible learnability advantage of agreement marking lies in having distinct forms as cues.

Humans use both linear and hierarchical representations in language processing, and the exact role of each has been debated. One domain where hierarchical processing is important is noun phrases. English noun phrases have a fixed order of prenominal modifiers: demonstratives - numerals - adjectives (these two green vases). However, when English speakers learn an artificial language with postnominal modifiers, instead of reproducing this linear order they preserve the distance between each modifier and the noun (vases green two these). This has been explained by a hierarchical homomorphism bias. Here, we investigate whether RNNs exhibit this bias. We pre-train one linear and two hierarchical models on English and expose them to a small artificial language. We then test them on noun phrases from a study with humans and find that only the hierarchical models can exhibit the bias, supporting the idea that homomorphic word order preferences arise from hierarchical, and not linear relations.

Syntactic harmony occurs when heads and dependents align within and across different types of phrases in a language. Harmony is a well-known (statistical) typological universal: in most languages, many if not all heads and dependents are consistently ordered (i.e., either head-dependent, or dependent-head). Despite decades of work, from every conceivable theoretical perspective, the origins of syntactic harmony remain opaque. However, recent work using artificial language learning has suggested that harmonic patterns are easier to learn than their non-harmonic counter-parts. Thus at least part of the explanation for this tendency may be linked to learning. Here, we explore whether the mechanism behind the learning bias for syntactic harmony is fundamentally domain-general by instantiating harmony in non-linguistic stimuli. Our findings support the claim that the origins of syntactic harmony lie in a domain-general bias for simplicity acting on linearized, language-specific categories.

Responsive predicates are clause-embedding predicates like English 'know' and 'guess' that can take both declarative and interrogative clausal complements. The meanings of responsive predicates when they take a declarative complement and when they take an interrogative complement are hypothesized to be constrained in systematic ways across languages, suggesting that these constraints represent semantic universals. We report an artificial language learning experiment showing that one of these proposed constraints is indeed reflected in the inferences participants make while learning a novel responsive predicate. Our results add support to a growing body of evidence linking semantic universals to learning.

Typological data shows a tendency for languages to exhibit harmonic (i.e. consistent) ordering between heads and dependents. However, some categories seem to contradict this tendency. Here we investigate one such case, the order of the noun with respect to two dependents—adjectives, which tend to follow the noun and genitives which precede. We report two silent gesture experiments examining (i) whether there are cognitive biases favouring postnominal adjective and prenominal genitive order in a single trial judgement task, and (ii) if those preferences continue to influence order when participants learn a complete word order system. Our results shed light on how biases for individual categories of elements interact with biases that affect the wider linguistic system. While participants strongly prefer postnominal adjectives and prenominal genitives when these are judged in isolation, when they learn a system of ordering, these biases are obscured and (at least in some cases) harmony emerges.

There is ample evidence that individual‐level cognitive mechanisms active during language learning and use can contribute to the evolution of language. For example, experimental work suggests that learners will reduce case marking in a language where grammatical roles are reliably indicated by fixed word order, a correlation found robustly in the languages of the world. However, such research often assumes homogeneity among language learners and users, or at least does not dig into individual differences in behavior. Yet, it is increasingly clear that language users vary in a large number of ways: in culture, in demographics, and—critically for present purposes—in terms of cognitive diversity. Here, we explore how neurodiversity impacts behavior in an experimental task similar to the one summarized above, and how this behavior interacts with social pressures. We find both similarities and differences between autistic and nonautistic English‐speaking individuals, suggesting that neurodiversity can impact language change in the lab. This, in turn, highlights the potential for future research on the role of neurodivergent populations in language evolution more generally. [ABSTRACT FROM AUTHOR]