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.

Grammatical redundancy is a widespread feature across languages. Although redundant cues can be seen to increase the complexity and processing burden of structures, it has been suggested that they can assist language acquisition. Here, we explored if this learning benefit can be observed from the very initial stages of second language (L2) acquisition and whether the effect of redundancy is modulated by the perceptual salience of the redundant linguistic cues. Across two experiments, three groups of adult native speakers of English were incidentally exposed to three different artificial languages; one that had a fixed word order, Verb-Object-Subject, and two in which thematic role assignment was additionally determined by a low-salient (Experiment 1) or a high-salient (Experiment 2) redundant case marker. While all groups managed to learn the novel language, our results pointed towards a hindering role of redundancy, with participants in the non-redundant condition achieving greater learning outcomes compared to those in both redundant conditions. Results also revealed that this impeding effect of redundancy on L2 learners can be attenuated by the salience of the redundant cue (Experiment 2). In conjunction with earlier findings, the present results suggest that the effect of redundancy on L2 acquisition can be differentially manifested depending on the stage of L2 development, learners’ first language biases and age.

This study investigates substantive bias, a phenomenon wherein learners exhibit a preference for phonetically motivated patterns in language acquisition. The study presents evidence that variable input, rather than categorical input, can activate substantive bias. Native speakers of Hong Kong Cantonese were randomly assigned to categorical or variable training conditions for vowel backness harmony or disharmony, or to a no-training control condition. Results indicate that participants in the categorical and control conditions did not exhibit a bias towards either pattern. However, participants in the variable conditions displayed a bias towards vowel harmony, suggesting that input variability can strengthen the effect of substantive bias. These findings contribute to our comprehension of the role of input variability in phonological learning and the mechanisms involved in the acquisition of phonetically motivated and unmotivated phonological patterns.

Agreement markers that refer to the same feature or argument tend to be found in the same position (e.g., all subject agreement markers as suffixes, all object agreement markers as prefixes). However, little is known about the exceptions to this trend: cases where different values of the same feature are marked in different positions in the word (i.e., positional splits). In this study, we explore the positional properties of subject and object person-number agreement markers in a phylogenetically diverse sample of 227 languages. We find that the recurrence of a positional split is proportional to its degree of naturalness, that is, to the amount of shared feature values amongst the forms with the same positional arrangement. Natural patterns (e.g., where prefixal forms all share SG and suffixal forms all share PL) are over-represented in natural languages compared to a random baseline. The most unnatural patterns are underrepresented, and splits with an intermediate level of unnaturalness occur at around chance levels. We hypothesise that this graded bias for naturalness is grounded in a preference for morphological similarity amongst semantically similar forms during language learning. To test this hypothesis we conducted two online artificial language learning experiments where we trained and tested participants on person-number verbal agreement paradigms of different sizes with positional splits of different degrees of naturalness. We found that their relative learnability is also gradient, again proportional to the amount of feature value overlap, thus matching the observed cross-linguistic tendencies. Our findings support the notion that semantic similarity shapes the evolution of morphological structure in person-number verbal agreement systems and that it does so in a gradient way.

Artificial language learning research has become a popular tool to investigate universal mechanisms in language learning. However, often it is unclear whether the found effects are due to learning, or due to artefacts of the native language or the artificial language, and whether findings in only one language will generalise to speakers of other languages. The present study offers a new approach to model the influence of both the L1 and the target artificial language on language learning. The idea is to control for linguistic factors of the artificial and the native language by incorporating measures of wordlikeness into the statistical analysis as covariates. To demonstrate the approach, we extend Linzen and Gallagher (2017)’s study on consonant identity pattern to evaluate whether speakers of German and Mandarin rapidly learn the pattern when influences of L1 and the artificial language are accounted for by incorporating measures assessed by analogical and discriminative learning models over the L1 and artificial lexicon. Results show that nonwords are more likely to be accepted as grammatical if they are more similar to the trained artificial lexicon and more different from the L1 and, crucially, the identity effect is still present. The proposed approach is helpful for designing cross-linguistic studies.

Response particles, like English ‘yes’ and ‘no’, are used to respond to polar questions or assertions and are found in all languages. However, the number of particles and the specific meanings they convey vary across languages. For example, in some languages particles mainly convey whether the response itself is positive or negative, while in others they convey whether the response is agreeing or disagreeing with previous discourse. Further, some languages have two response particles, while others have three, or even four. Recent work suggests that how meanings tend to be mapped to forms cross-linguistically might nevertheless be constrained. Roelofsen & Farkas (2015) suggest that indicating disagreement with a negative question or assertion (e.g., A: ‘Ally doesn’t eat meat.’ B: ‘Yes, he does.’) is more marked than indicating agreement with a positive assertion (e.g., A: ‘Ally eats meat.’ B: ‘Yes, he does’.). This difference in semantic markedness is argued to lead to a difference in form: more marked meanings are mapped to more specialized forms. Here we investigate this hypothesis in a series of behavioral experiments. Across our experiments, we find that participants are indeed sensitive to the differences in meaning that particles can convey. However, not all of the differences implicated by the hierarchy hypothesized in Roelofsen & Farkas (2015) are supported by our results, and we find evidence highlighting an unexpected special role for Positive Agreement—the least marked meaning.

Prediction error is known to enhance priming effects for familiar syntactic structures; it also strengthens the formation of new declarative memories. Here, we investigate whether violating expectations may aid the acquisition of new abstract syntactic structures, too, by enhancing memory for individual instances which can then form the basis for abstraction. In a cross-situational artificial language learning paradigm, participants were exposed to novel syntactic structures in ways that either violated their expectations (Surprisal group) or that conformed to them (Control group). In a delayed post-test, participants were tested on their structural knowledge both by means of structure test trials (cross-situational learning trials focusing on the active / passive distinction, with both familiar and novel verbs), and by a grammaticality judgment task. Participants in the Surprisal group were significantly more accurate than Control on the structure test trials using novel verbs and in the grammaticality judgment task, suggesting they had developed stronger abstract structural knowledge and were better at generalising it to novel instances, even though they were not significantly more likely to become aware of the functional distinction between the two structures.

In order to become proficient native speakers, children have to learn the morpho-syntactic relations between distant elements in a sentence, so-called non-adjacent dependencies (NADs). Previous research suggests that NAD learning in children comprises different developmental stages, where until 2 years of age children are able to learn NADs associatively under passive listening conditions, while starting around the age of 3–4 years children fail to learn NADs during passive listening. To test whether the transition between these developmental stages occurs gradually, we tested children’s NAD learning in a foreign language using event-related potentials (ERPs). We found ERP evidence of NAD learning across the ages of 1, 2 and 3 years. The amplitude of the ERP effect indexing NAD learning, however, decreased with age. These findings might indicate a gradual transition in children’s ability to learn NADs associatively. Cognitively, this transition might be driven by children’s increasing knowledge of their native language, hindering NAD learning in novel contexts. Neuroanatomically, maturation of the prefrontal cortex might play a crucial role, promoting top-down learning, affecting bottom-up, associative learning. In sum, our study suggests that NAD learning under passive listening conditions undergoes a gradual transition between different developmental stages during early childhood.

This study investigated the effect of speaking style on speech segmentation by statistical learning under optimal and adverse listening conditions. Similar to the intelligibility and memory benefits found in previous studies, enhanced acoustic-phonetic cues of the listener-oriented clear speech could improve speech segmentation by statistical learning compared to conversational speech. Yet, it could not be precluded that hyper-articulated clear speech, reported to have less pervasive coarticulation, would result in worse segmentation than conversational speech. We tested these predictions using an artificial language learning paradigm. Listeners who acquired English before age six were played continuous repetitions of the ‘words’ of an artificial language, spoken either clearly or conversationally and presented either in quiet or in noise at a signal-to-noise ratio of +3 or 0 dB SPL. Next, they recognized the artificial words in a two-alternative forced-choice test. Results supported the prediction that clear speech facilitates segmentation by statistical learning more than conversational speech but only in the quiet listening condition. This suggests that listeners can use clear speech acoustic-phonetic enhancements to guide speech processing dependent on domain-general, signal-independent statistical computations. However, there was no clear speech benefit in noise at either signal-to-noise ratio. We discuss possible mechanisms that could explain these results.

Recent work has used artificial language experiments to argue that hierarchical representations drive learners’ expectations about word order in complex noun phrases like these two green cars (Culbertson & Adger 2014; Martin, Ratitamkul, et al. 2019). When trained on a novel language in which individual modifiers come after the Noun, English speakers overwhelmingly assume that multiple nominal modifiers should be ordered such that Adjectives come closest to the Noun, then Numerals, then Demonstratives (i.e., N-Adj-Num-Dem or some subset thereof). This order transparently reflects a constituent structure in which Adjectives combine with Nouns to the exclusion of Numerals and Demonstratives, and Numerals combine with Noun+Adjective units to the exclusion of Demonstratives. This structure has also been claimed to derive frequency asymmetries in complex noun phrase order across languages (e.g., Cinque 2005). However, we show that features of the methodology used in these experiments potentially encourage participants to use a particular metalinguistic strategy that could yield this outcome without implicating constituency structure. Here, we use a more naturalistic artificial language learning task to investigate whether the preference for hierarchy-respecting orders is still found when participants do not use this strategy. We find that the preference still holds, and, moreover, as Culbertson & Adger (2014) speculate, that its strength reflects structural distance between modifiers. It is strongest when ordering Adjectives relative to Demonstratives, and weaker when ordering Numerals relative to Adjectives or Demonstratives relative to Numerals. Our results provide the strongest evidence yet for the psychological influence of hierarchical structure on word order preferences during learning.

Linguistic animacy reflects a particular construal of biological distinctions encountered in the world, passed through cultural and cognitive filters. This study explores the process by which our construal of animacy becomes encoded in the grammars of human languages. We ran an iterated learning experiment investigating the effect of animacy on language transmission. Participants engaged in a simple artificial language learning task in which they were asked to learn which affix was assigned to each noun in the language. Though initially random, the language each participant produced at test became the language that the subsequent participant in a chain was trained on. Results of the experiment were analysed in terms of learnability, measured through the accuracy of responses, and structure, using an entropy measure. We found that the learnability of languages increased over generations, as expected, but entropy did not decrease. Languages did not become formally simpler over time. Instead, structure emerged through a reorganisation of noun classes around animacy-based categories. The use of semantic animacy distinctions allowed languages to retain morphological complexity while becoming more learnable. Our study shows that grammatical reflexes of animacy distinctions can arise out of learning alone, and that structuring grammar based on animacy can make languages more learnable.

The tendency for languages to use harmonic word order patterns—orders that place heads in a consistent position with respect to modifiers or other dependents—has been noted since the 1960s. As with many other statistical typological tendencies, there has been debate regarding whether harmony reflects properties of human cognition or forces external to it. Recent research using laboratory language learning has shown that children and adults find harmonic patterns easier to learn than nonharmonic patterns (Culbertson & Newport, 2015; Culbertson, Smolensky, & Legendre, 2012). This supports a link between learning and typological frequency: if harmonic patterns are easier to learn, while nonharmonic patterns are more likely to be targets of change, then, all things equal, harmonic patterns will be more frequent in the world’s languages. However, these previous studies relied on variation in the input as a mechanism for change in the lab; learners were exposed to variable word order, allowing them to shift the frequencies of different orders so that harmonic patterns became more frequent. Here we teach adult and child learners languages that are consistently nonharmonic, with no variation. While adults perfectly maintain these consistently nonharmonic patterns, young child learners innovate novel orders, changing nonharmonic patterns into harmonic ones.

Statistical learning (SL) involving sensitivity to distributional regularities in the environment has been suggested to be an important factor in many aspects of cognition, including language. However, the degree to which statistically-learned information is retained over time is not well understood. To establish whether or not learners are able to preserve such regularities over time, we examined performance on an artificial second language learning task both immediately after training and also at a follow-up session 2 weeks later. Participants were exposed to an artificial language (Brocanto2), half of them receiving simplified training items in which only 20% of sequences contained complex structures, whereas the other half were exposed to a training set in which 80% of the items were composed of complex sequences. Overall, participants showed signs of learning at the first session and retention at the second, but the degree of learning was affected by the nature of the training they received. Participants exposed to the simplified input outperformed those in the more complex training condition. A GLMM was used to model the relationship between stimulus properties and participants’ endorsement strategies across both sessions. The results indicate that participants in the complex training condition relied more on an item’s chunk strength than those in the simple training condition. Taken together, this set of findings shows that statistically learned regularities are retained over the course of 2 weeks. The results also demonstrate that training on input featuring simple items leads to improved learning and retention of grammatical regularities.

Plural morphology exhibits differing interpretations across languages. For example, in downward entailing contexts in English, the plural receives a one or more (or inclusive) interpretation, whereas in Korean-like languages the plural always receives a more than one (or exclusive) interpretation, regardless of context. Previous experimental work using an artificial language suggests that such differences may follow from structural properties of these languages (Liter, Heffner & Schmitt 2017), namely lack of grammaticalization of the plural/singular distinction. In this paper we adopt Sauerland, Anderssen & Yatsushiro’s (2005) implicated presupposition analysis of the plural (the English plural is semantically unmarked, whereas the Korean plural is semantically marked, carrying a presupposition that the cardinality of its referent is greater than one) in order to test two hypotheses about the interpretation of the plural. Using an artificial language learning paradigm identical to that in Liter, Heffner & Schmitt (2017) with non-grammaticalized number but with a much greater frequency of singular/plural NPs in the input, we test (i) whether semantic markedness of the plural should be linked to the non-grammaticalization of the number paradigm; or (ii) whether semantic markedness follows from insufficient statistical evidence for simplifying the lexical entry for the plural. Our results show that participants continue to assign an exclusive interpretation to plural morphology under the scope of negation, which is compatible with the hypothesis that non-grammaticalized number entails semantic markedness.

Several theoretical proposals for the evolution of language have sparked a renewed search for comparative data on human and non-human animal computational capacities. However, conceptual confusions still hinder the field, leading to experimental evidence that fails to test for comparable human competences. Here we focus on two conceptual and methodological challenges that affect the field generally: 1) properly characterizing the computational features of the faculty of language in the narrow sense; 2) defining and probing for human language-like computations via artificial language learning experiments in non-human animals. Our intent is to be critical in the service of clarity, in what we agree is an important approach to understanding how language evolved.

In an artificial language-learning task, two groups of English and French participants learned one of two language rules: 1) stress the first heavy (CVC) syllable, else the first syllable, or, 2) stress the first light (CV) syllable, else the first syllable. French and English participants were chosen to compare learning outcomes by speakers of different native stress systems, fixed and variable. Participants were trained on the target language by listening to a set of nonsense familiarization words exemplifying the stress rule. This was followed by a forced-choice task to choose the correct version of the words they had just learned. Following the training procedure, participants were tested on novel words with the same stress pattern to which they were familiarized. The result of the novel word testing was that the natural rule with stress on heavy syllables was learned significantly better than the unnatural, stress light syllables, rule. To account for the learnability of both the natural and the unnatural rules, I argue for the interaction of a general cognitive mechanism that facilitates learning in general and a domain-specific language mechanism that can access universal phonological principles to aid in learning a natural language rule.

Language acquisition in both natural and artificial language learning settings crucially depends on extracting information from ordered sequences. A shared sequence learning mechanism is thus assumed to underlie both natural and artificial language learning. A growing body of empirical evidence is consistent with this hypothesis. By means of artificial language learning experiments, we may therefore gain more insight in this shared mechanism. In this paper, we review empirical evidence from artificial language learning and computational modeling studies, as well as natural language data, and suggest that there are two key factors that help deter-mine processing complexity in sequence learning, and thus in natural language processing. We propose that the specific ordering of non-adjacent dependencies (i.e. nested or crossed), as well as the number of non-adjacent dependencies to be resolved simultaneously (i.e. two or three) are important factors in gaining more insight into the boundaries of human sequence learning; and thus, also in natural language processing. The implications for theories of linguistic competence are discussed.

Being able to track dependencies between syntactic elements separated by other constituents is crucial for language acquisition and processing (e.g., in subject-noun/verb agreement). Although long assumed to require language-specific machinery, research on statistical learning has suggested that domain-general mechanisms may support the acquisition of nonadjacent dependencies. In this study, we investigated whether individuals with specific language impairment (SLI)—who have problems with long-distance dependencies in language—also have problems with statistical learning of nonadjacent relations. The results confirmed this hypothesis, indicating that statistical learning may subserve the acquisition and processing of long-distance dependencies in natural language.