Search

Grammatical regularities may correlate with semantics; e.g. grammatical gender is often partially predictable from the noun's semantics. We explore whether learners generalise over semantic cues, and whether the extent of exposure (1 versus 4 sessions) and number of exemplars for each semantic class (type-frequency) affect this. Six-year-olds and adults were exposed to semi-artificial languages where nouns co-occurred with novel particles, with particle usage fully or partially determined by the semantics of the nouns. Both adults and children generalised to novel nouns when semantic cues were fully consistent. Adults (but not children) also generalised when cues were partially consistent. Generalisation increased with exposure, however there was no evidence that increasing type-frequency (i.e. more nouns per semantic class) increased generalisation. Post-experiment interviews also suggested that successful generalisation depended on explicit awareness. These results suggest that semantic cues are particularly difficult for children to exploit during the early stages of language acquisition.

This is the preregistration for Experiment2 of Harmony Block. The predictions are essentially the same, because the changes in the design hopefully will avoid some of the issues in learnability for the patterns

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.

We combine two established research methods (trust games and artificial miniature languages) to investigate whether listeners use linguistic cues to infer ingroup membership, and whether they treat speakers who speak their own language as more trustworthy. Additionally, we investigate how linguistic cues compare in this respect to other cultural and biological markers of group membership.

Here we extend the experiment pre-registered in https://osf.io/z872c/?view_only=3ca6c44c6e4e48389585fa51f5184d8c to a population of Tamil rather than English native speakers. In that experiment, we used an extrapolation paradigm to test whether English-speaking learners distinguish between different tripartitions of the first person space. Participants were first taught three pronominal forms, corresponding to the exclusive minimal (1), inclusive minimal (12), and exclusive plural (1+) categories. They were then asked to generalize (or extrapolate) the forms they learned to cases where they have to refer to the speaker, the addressee and others (inclusive augmented category, 12+). English-speaking participants in this experiment did not have a consistent extrapolation pattern: some of them assimilated meanings on the basis of a person natural class (12=12+), and some of them assimilated meanings on the basis of a number natural class (1+=12+). Unlike English, Tamil is a language that makes both a number (singular vs. plural) and a clusivity (exclusive vs. inclusive) distinction in its first person pronominal system. That is, Tamil has three different pronominal forms: one for the speaker alone (1= ‘nan’), one for the speaker plus others (1+= ‘naanga’/’nam’) and one for the speaker plus the addressee plus potential others (12=12+=‘namba’). In the present experiment, we test whether Tamil speaking participants, when learning a new pronominal system, have a preference for making the same distinctions present in their native language.

This is a follow-up from the two previous experiments pre-registred in this project, which investigates the Jespersen-Zeijlstra generalization. This generalization makes a connection between the interpretation of sentences involving multiple negative elements and formal properties of the negative marker. We test whether learners are sensitive to the correlation between type of negative marker and being NC or DN. Are English-speakers more likely to treat a language as negative concord if the negative marker is an affix than if it’s an adverb? What about treating it as double negation? Previous experiments in this project have failed to show that the Marker has an influence on the type of interpretation. Instead, these experiments have only suggested a overall bias against double-negation languages. While in previous experiments we addressed these questions by using an ease of learning paradigm, in this experiment we use an extrapolation design. English-speaking participants are first taught a miniature language, including verbs, nouns, a negative marker, and negative indefinites. Depending on the condition, learners will be taught a negative marker that is either an affix or an adverb. Crucially, during training, participants are not exposed to sentences involving both the negative marker *and* a negative indefinite. In the testing phase, participants are asked to assign a meaning to these critical sentences by performing a sentence-picture matching task. We will test whether the interpretation that participants assign to sentences involving two negative elements is determined by the type of negative marker. In addition, we will investigate whether participants continue to show a dispreference for double-negation interpretations in the extrapolation paradigm.

The first person space can be thought of as covering all possible groups of individuals that include the speaker of the conversational context. Research on the typology of person systems suggests that languages carve-up this space by making use of features, which are smaller than the groupings or categories themselves. For example, a language that makes use of a ±addressee person feature will distinguish pronouns as a function of whether their reference includes both speaker and addressee, or the speaker but not the addressee. Similarly, a language that makes use of only a number distinction (such as ±atomic) will just differentiate between a pronoun that refers to the speaker alone and a pronoun that refers to the speaker plus any others (e.g., English). Person and number features can also be combined. Languages like Tamil make an ±atomic contrast on top of the clusivity distinction, and have three first person pronominal forms. Previous work (Maldonado & Culbertson 2019, Maldonado & Culbertson, to appear) investigated whether English-speaking learners differentiate between alternative bipartitions of the first person space as defined by homophony (i.e., two-way person systems). Besides having a strong bias for first person paradigms that resemble their native language, English learners were shown to prefer systems that feature person homophony over those that feature number homophony. We interpreted this as a consequence of the fact that their native-language makes no clusivity distinction, so participants show a bias against systems that only make such distinction. In this experiment, we aim at testing whether English-speaking learners distinguish different tripartitions of the first person space, where three pronominal forms are used to cover the first person space. We will use the same artificial language learning methodology used in Maldonado & Culbertson (2019) to test whether English speakers continue to prefer person homophony over other patterns when they are learning a first person tripartition.

This is a follow-up from the previous experiments pre-registred in this project. From a logical point of view, one would expect every negative element to contribute a negation in the semantics. This is the case for languages like Dutch: the combination of two negative elements give rise to a double negation interpretation, resulting in a positive meaning. (1) Niemand rent niet. N-body run NEG. “Nobody doesn’t run” → “Everybody runs” However, this is not the case in all languages. For example, despite involving two negative elements, a Serbian sentence like (2) only contains one semantic negation. In other words, the two negative elements in (2) do not each contribute a semantic negation, but rather convey a negative meaning together, yielding what is known as a negative concord interpretation. 2) Niko ne trci N-word NEG run "Nobody runs" The contrast between (1) and (2) illustrates the existence of cross-linguistic variation wrt the interpretation of negative elements. A natural question to ask is what determines the distribution and behaviour of double negation (DN) and negative concord (NC) languages. That is, where does the difference between languages like Dutch and languages like Serbian come from. Accounts of the difference between DN and NC languages have often relied on the existence of other linguistic properties that correlate with the kinds of interpretations that sentences like (1)/(2) can get. One observation that has played an important role in current theories of negative concord was made by Jespersen in 1917. Jespersen noted that whether a language is double negation or negative concord correlates with the phonological and syntactic nature of the negative marker. That is, whether the negative marker is an adverb or a particle/affix (see Zanuttini 1997 for tests to tease these apart). More specifically, Jespersen argued that languages which have only a negative adverb always exhibit double negation, while languages which have only a negative particle/affix always exhibit negative concord. While this generalization, as it is, is too strong (there are negative concord languages that only have negative adverbs; e.g., Quebecois, Deprez 1997), it can still be accurately reformulated in an unidirectional way: a language may lack negative concord if and only if it has a negative adverb (and not a negative particle or affix) (Zeijlstra 2004). The Jespersen-Zeijlstra generalization is important for current theories of negation because it suggests that the contrast between DN and NC languages is partially due to a difference in the syntactic status of adverbs and particles, which would in turn have consequences on how they compose with negative indefinites. For instance, according to Z., only negative particles/affixes can license n-words, so only languages that have those should be able to have negative concord interpretations. The exact strength of this generalization, however, remains unclear: Jespersen’s original generalization had to be weakened as more typological data became available, leaving open the possibility that even a uni-directional interpretation may be too strong. In this project, we investigate the Jespersen-Zeijlstra generalization by testing whether learners are sensitive to the correlation between type of negative marker and being NC or DN. Are English-speakers more likely to treat a language as negative concord if the negative marker is an affix than if it’s an adverb? What about treating it as double negation? Here, we will use an artificial language learning experiment to test whether English-speaking participants find it easier to learn a DN language if the marker is an adverb than if it is a an affix. Participants will be taught a miniature language, including verbs, nouns, negative markers, and negative indefinites. Depending on the condition, learners will be taught a negative marker that is either an affix or an adverb. Finally, participants will be tested on their that comprehension and production of sentences two negative elements (i.e., a negative indefinite and a negative marker).

We combine two established research paradigms (trust games and artificial miniature languages) to investigate whether listeners use linguistic cues to infer ingroup membership, and whether they trust speakers who speak their own language more than speakers who speak a different language. More specifically, in our experiment, a participant travels to a fictional Alien Planet, on which red* and yellow* aliens live. The participant answers three personal questions and is told that they are themselves transformed into an alien based on their replies to these questions. (In fact, the allocation is pre-determined based on the treatment condition). Thus, the participant becomes a member of one of the alien groups. Each of the two alien groups speaks a distinct made-up language*, which the participant is familiarised with. (*Colours and languages are randomized). Subsequently, the participant meets two aliens, who they believe are played by other real participants but are actually bots. The aliens' colours are masked in darkness but the participant can briefly hear them speak. Then, the participant plays a trust game with both aliens simultaneously, in which the participant is the sender and the two other aliens are the receivers. All three get an initial endowment of 10 Alien-Dollars. The participant can divide their 10 Alien-Dollars between the two other aliens. The participant knows that whatever amounts they pass on will be tripled, and that the other aliens can either keep those tripled amounts or return a share. Thus, the amounts the participant is willing to share with each receiver-alien reflect the extent of their trust in them.

From a logical point of view, one would expect every negative element to contribute a negation in the semantics. This is the case for languages like Dutch: the combination of two negative elements (the negative marker “niet” and the negative indefinite “niemand”) in both (1-a) and (1-b) give rise to a double negation interpretation, resulting in a positive meaning. (1) a. Niemand rent niet. N-body NEG run. “Nobody doesn’t run” → “Everybody runs” b. Jan belt niet niemand Jan calls NEG n-body “Jan doesn’t call nobody” → “Jan calls somebody” However, this is not the case in all languages. For example, despite involving two negative elements (the negative marker “ne” and the negative indefinite “niko(ga)”), a Serbian sentence like (2-a) only contains one semantic negation. In other words, the two negative elements in both (2-a) and (2-b) do not each contribute a semantic negation, but rather convey a negative meaning together, yielding what is known as a negative concord interpretation. 2) a. Niko ne trci N-word NEG run “Nobody run” b. Milan nikoga ne zove. Milan n-body NEG calls “Milan calls nobody” Words like “niko”, which introduce negative force but participate in negative concord relations, are often referred to as “n-words” (Laka 1990), to differentiate them from words like Dutch “niemand” (which stand on their own). The contrast between (1) and (2) illustrates the existence of cross-linguistic variation wrt the interpretation of negative elements. A natural question to ask is what determines the distribution and behaviour of double negation (DN) and negative concord (NC) languages. That is, where does the difference between languages like Dutch and languages like Serbian come from. Accounts of the difference between DN and NC languages have often relied on the existence of other linguistic properties that correlate with the kinds of interpretations that sentences like (1)/(2) can get. One observation that has played an important role in current theories of negative concord was made by Jespersen in 1917. Jespersen noted that whether a language is double negation or negative concord correlates with the phonological and syntactic nature of the negative marker. That is, whether the negative marker is an adverb or a particle/affix (see Zanuttini 1997 for tests to tease these apart). More specifically, Jespersen argued that languages which have only a negative adverb always exhibit double negation, while languages which have only a negative particle/affix always exhibit negative concord. While this generalization, as it is, is too strong (there are negative concord languages that only have negative adverbs; e.g., Quebecois, Deprez 1997), it can still be accurately reformulated in an unidirectional way: a language may lack negative concord if and only if it has a negative adverb (and not a negative particle or affix) (Zeijlstra 2004). The Jespersen-Zeijlstra generalization is important for current theories of negation because it suggests that the contrast between DN and NC languages is partially due to a difference in the syntactic status of adverbs and particles, which would in turn have consequences on how they compose with negative indefinites. For instance, according to Z., only negative particles/affixes can license n-words, so only languages that have those should be able to have negative concord interpretations. The exact strength of this generalization, however, remains unclear: Jespersen’s original generalization had to be weakened as more typological data became available, leaving open the possibility that even a uni-directional interpretation may be too strong. In this project, we investigate the Jespersen-Zeijlstra generalization by testing whether learners are sensitive to the correlation between type of negative marker and being NC or DN. Are English-speakers more likely to treat a language as negative concord if the negative marker is an affix than if it’s an adverb? What about treating it as double negation? Here, we will use an artificial language learning experiment to test whether English-speaking participants find it easier to learn a DN language if the marker is an adverb than if it is a an affix. Participants will be taught a miniature language, including verbs, nouns, negative markers, and negative indefinites. Depending on the condition, learners will be taught a negative marker that is either an affix or an adverb, and will be taught that the meaning of sentences with two negative elements (i.e., a negative indefinite and a negative marker) is either DN or NC. Participants are then tested on how well they learn these sentences.

From a logical point of view, one would expect every negative element to contribute a negation in the semantics. This is the case for languages like Dutch: the combination of two negative elements (the negative marker “niet” and the negative indefinite “niemand”) in both (1-a) and (1-b) give rise to a double negation interpretation, resulting in a positive meaning. (1) a. Niemand rent niet. N-body NEG run. “Nobody doesn’t run” → “Everybody runs” b. Jan belt niet niemand Jan calls NEG n-body “Jan doesn’t call nobody” → “Jan calls somebody” However, this is not the case in all languages. For example, despite involving two negative elements (the negative marker “ne” and the negative indefinite “niko(ga)”), a Serbian sentence like (2-a) only contains one semantic negation. In other words, the two negative elements in both (2-a) and (2-b) do not each contribute a semantic negation, but rather convey a negative meaning together, yielding what is known as a negative concord interpretation. 2) a. Niko ne trci N-word NEG run “Nobody run” b. Milan nikoga ne zove. Milan n-body NEG calls “Milan calls nobody” Words like “niko”, which introduce negative force but participate in negative concord relations, are often referred to as “n-words” (Laka 1990), to differentiate them from words like Dutch “niemand” (which stand on their own). The contrast between (1) and (2) illustrates the existence of cross-linguistic variation wrt the interpretation of negative elements. A natural question to ask is what determines the distribution and behaviour of double negation (DN) and negative concord (NC) languages. That is, where does the difference between languages like Dutch and languages like Serbian come from. Accounts of the difference between DN and NC languages have often relied on the existence of other linguistic properties that correlate with the kinds of interpretations that sentences like (1)/(2) can get. One observation that has played an important role in current theories of negative concord was made by Jespersen in 1917. Jespersen noted that whether a language is double negation or negative concord correlates with the phonological and syntactic nature of the negative marker. That is, whether the negative marker is an adverb or a particle/affix (see Zanuttini 1997 for tests to tease these apart). More specifically, Jespersen argued that languages which have only a negative adverb always exhibit double negation, while languages which have only a negative particle/affix always exhibit negative concord. While this generalization, as it is, is too strong (there are negative concord languages that only have negative adverbs; e.g., Quebecois, Deprez 1997), it can still be accurately reformulated in an unidirectional way: a language may lack negative concord if and only if it has a negative adverb (and not a negative particle or affix) (Zeijlstra 2004). The Jespersen-Zeijlstra generalization is important for current theories of negation because it suggests that the contrast between DN and NC languages is partially due to a difference in the syntactic status of adverbs and particles, which would in turn have consequences on how they compose with negative indefinites. For instance, according to Z., only negative particles/affixes can license n-words, so only languages that have those should be able to have negative concord interpretations. The exact strength of this generalization, however, remains unclear: Jespersen’s original generalization had to be weakened as more typological data became available, leaving open the possibility that even a uni-directional interpretation may be too strong. In this project, we investigate the Jespersen-Zeijlstra generalization by testing whether learners are sensitive to the correlation between type of negative marker and being NC or DN. Are English-speakers more likely to treat a language as negative concord if the negative marker is an affix than if it’s an adverb? What about treating it as double negation? Here, we will use an artificial language learning experiment to test whether English-speaking participants find it easier to learn a DN language if the marker is an adverb than if it is a an affix. Participants will be taught a miniature language, including verbs, nouns, negative markers, and negative indefinites. Depending on the condition, learners will be taught a negative marker that is either an affix or an adverb, and will be taught that the meaning of sentences with two negative elements (i.e., a negative indefinite and a negative marker) is either DN or NC. Participants are then tested on how well they learn these sentences.

Many of the word languages employ agreement marking (a systematic covariance between the form of at least two elements) even though it adds complexity and redundancy to language systems. This project aims to test a hypothesis that agreement marking might be functional for language learning. In addition, this project further tests a hypothesis that alliterative agreement (a more repetitive type of agreement marking) should benefit learning the most.

Negative Polarity Items (NPIs) are characterized by a polarity-sensitive use, restricted to downward entailing (DE) environments. However, in many languages, the very same lexical items also have positive uses that appear in upward entailing (UE) environments. For example, under negation, the NPI any has an existential meaning (“I didn’t talk to anybody” = “I didn’t talk to a single person”). But certain UE environments allow any to appear with a universal (or free choice) meaning (“I talked to anyone was interested” = “I talked to everyone who was interested.”) Similar positive uses can be found for other NPIs in English (ever, yet, anymore) and other languages (e.g. French encore). Ladusaw (1979) observes that the positive and negative uses of NPIs are often systematically related: they are logical duals. If an NPI is licensed by (and scoping under) negation, its positive counterpart carries the meaning the word would need to receive if it were interpreted as scoping above the negation, in order to derive the same sentential meaning. For any, existential force under negation becomes universal force (since ¬∃ = ∀¬). This observation offers a potential diachronic explanation of the systematic ambiguity. When the syntactic distribution of a logical item is restricted so that it always appears in the presence of negation, the meaning of the item is ambiguous between two denotations depending on its scope relative to negation: A¬ or ¬B. The typological data above can then be explained by the hypothesis that when the use of an NPI is extended to new, UE environments, the most attractive interpretation is the wide-scope dual meaning (A). Using an artificial language learning paradigm, we test how learners generalize the meaning of NPIs when they appear in positive environments. We teach English speaking participants an artificial miniature language which includes a negative marker *em* (‘not’) and a degree modifier *tup*, roughly equivalent to English ‘at all’. During training, participants are exposed to sentences in which tup is restricted to negative sentences (i.e. tup never occurs without em). At test, participants are asked to interpret sentences where the quantifier appears on its own, without negation. We evaluate whether learners are more likely to assign a universal or an existential interpretation to this sentence.

Some clause-embedding predicates like English "know" and "guess" can take both declarative and interrogative clausal complements. It has been hypothesised in the formal semantic literature that the meanings of these predicates when they take a declarative complement and when they take an interrogative complement are constrained in systematic ways. In this project, we use an artificial language learning paradigm to investigate whether learners are sensitive to this constraint on the meaning of clause-embedding predicates. In particular, we test whether learners’ interpretation of sentences of the form [Jo Vs whether p] is conditioned by their interpretation of sentences of the form [Jo Vs that p].

Languages vary on how they instantiate negative dependencies. In particular, sentences involving two negative elements can be assigned either a double negation reading, where each negative element contributes an independent semantic negation, as in the Dutch example in (1), or a negative concord reading, where the two negative elements convey a negative meaning together, as in the Serbian sentence in (2). (1) Niemand rent niet. N-body NEG run. “Nobody doesn’t run” → “Everybody runs” (2) Niko ne trci N-word NEG run “Nobody run” The contrast between (1) and (2) is typically used to illustrate the existence of cross-linguistic variation with respect to the interpretation of sentences involving multiple negative words: languages are thus categorized as either negative concord or double negation (Zeijlstra 2004). However, it has been recently shown that negative concord and double negation readings might also coexist within the same language (Blanchette and Lukyanenko 2019; Déprez et al. 2015; Espinal et al. 2016). In these cases, sentences with two negative words are in principle ambiguous between double negation and negative concord reading. Crucially, the interpretation of sentences containing two negative elements seems to correlate with the distributional and phonological nature of the negative elements, both across and within languages. For example, it has been noticed that languages that express sentential negation through a phonologically weak particle or affix are always negative concord (e.g., Serbian). This is known as Jespersen’s Generalization. While this generalization is stated across languages, there is evidence to suggest that a similar tendency might also hold within languages. For example, Blanchette et al (2018) show that English speakers are more likely to put emphasis on negative words when they want to convey a double negation interpretation as compared to a negative concord one. A similar pattern is found diachronically: negative markers within a language often get weakened, which results in gradual loss of their negative semantic value, in turn leading to negative concord interpretations. When this happens, the marker is often replaced or strengthened by an additional word, which serves as an emphatic negation (Kiparski and Condovardi 2016). In previous work, we have investigated the influential hypothesis that the cross-linguistic variation between DN and NC is partly determined by the phonological and syntactic nature of its negative marker (Jespersen, 1917; Zeijlstra 2004). In this study, we aim to explore how this variation is treated within a language. Specifically, we test whether learners condition their interpretation of sentences with two negative elements on properties of the negative marker. Here, we teach English speaking participants an artificial miniature language which includes one negative indefinite and two different negative markers: one that is phonologically weak/less salient, and one that is phonologically strong/more salient. These two markers vary in their position, their prosody, and their position in the sentence. We test whether learners are more likely to assign a double negation reading to those sentences that include the strong/more salient negative marker (referred to as emphatic throughout) than those that include the weak/less salient one (referred to as plain throughout).

Pre-registration for extrapolation experiment on first-person bipartitions

Data and analysis scripts for Guo, Z.-C., & Smiljanic, R. (2021). "Speaking clearly improves speech segmentation by statistical learning under optimal listening conditions," Laboratory Phonology: Journal of the Association for Laboratory Phonology, 12(1), 14. doi:10.5334/labphon.310

This is a follow-up from the project described here: https://osf.io/uk4fv The main difference is that the desing here is between-subjects and not within-subjects.

Many of the world languages employ agreement marking (a systematic covariance between the form of at least two elements) even though it adds complexity and redundancy to language systems. This project aims to test a hypothesis that agreement marking might be functional for language learning. In addition, this project further tests a hypothesis that alliterative agreement (a more repetitive type of agreement marking) should benefit learning the most.

Syntax famously consists of abstract hierarchical representations, essentially instructions for combining words into larger units like sentences. Less famously, most theories of syntax also assume a higher level of abstract representation. Representations at this level comprise instructions for creating the hierarchical representations used to create sentences. To date, however there is no experimental evidence for this additional level of abstraction. Here, we explain why the existence of such representations would imply that, under certain circumstances, speakers should be able to produce structures they have never been exposed to, and we test this prediction directly. We ask: Given the right type of input, can speakers learn a syntactic structure without direct exposure? In particular, different types of relative clauses have different surface word orders. These may be represented in two ways: with many individual representations or one general representation. If the latter, then learning one type of relative clause amounts to learning all types. We teach participants a novel grammar for only some relative clause types (e.g., just subject relative clauses) and test their knowledge of other types (e.g., object relative clauses). Across experiments, participants consistently produced untrained types, implicating the existence of this higher level of abstract syntactic knowledge.

General vowel harmony and disharmony rules have comparable formal complexity but differ dramatically in typological frequency and phonetic motivation. Previous studies found no difference in learning between vowel harmony and disharmony; this putative equivalence has been used to discount the view that learners are influenced by substantive learning biases. In the current study, we use a more nuanced test to show that there is a clear difference in learning between vowel harmony and disharmony: learners readily infer a vowel harmony pattern, but not a disharmony pattern. The findings suggest that vowel disharmony is in fact strongly disfavored during learning.

The present study explores learning phonological alternations that contain exceptions. Participants were exposed to a back/round vowel harmony pattern in which a regular suffix obeyed a vowel harmony rule, varying between /e/ and /o/ depending on the back/round phonetic features of the stem, and a non-alternating suffix that was always /o/ regardless of the features of the stem vowel. Participants in Experiment 1 learned the behavior of both suffixes, but correct performance for the non-alternating suffix was higher when the suffix happened to be in harmony with the stem. Participants in Experiment 2 were exposed to the non-alternating affix in harmonic contexts only, and continued to show a bias towards harmony. Experiment 3 replicated Experiment 2 with minimal training on disharmonic cases of the non-alternating morpheme. However, participants were less likely to learn the alternating affix without exposure to morphological stem, stem + suffix alternations in Experiment 4, suggesting a bias towards morphophonological alternations in learning vowel harmony patterns.

Neste texto, resenha-se a conferência da pesquisadora Dr.ª Natalia Levshina intitulada Eficiência comunicativa, teoria da informação e os limites da mente humana proferida no evento Abralin ao vivo – Linguists online, no dia 06 de julho de 2020. A conferencista, por meio de sua fala, busca problematizar a questão da eficiência comunicativa e sua relação com a teoria da informação e a pragmática, utilizando exemplos linguísticos tanto da fala humana quanto de línguas artificiais. Ao final de sua arguição, a pesquisadora discorre sobre desafios e aponta possíveis soluções para o desenvolvimento da teoria da eficiência comunicativa utilizando para tanto experimentos com a aprendizagem de línguas artificiais.

Several studies found cross‐linguistic structural priming with various language combinations. Here, we investigated the role of two important domains of language variation: case marking and word order, for transitive and ditransitive structures. We varied these features in an artificial language learning paradigm, using three different artificial language versions in a between‐subjects design. Priming was assessed between Dutch (no overt case marking, SVO word order) and a) an SVO order version, b) a case marking version, and c) an SOV order version. Similar within‐language and cross‐linguistic priming was found in all versions for transitives, indicating that cross‐linguistic structural priming was not hindered. In contrast, for ditransitives we found similar within‐language priming for all versions, but no cross‐linguistic priming. The finding that cross‐linguistic priming is possible between languages that vary in morphological marking or word order, is compatible with studies showing cross‐linguistic priming between natural languages that differ on these dimensions.

How exactly do learners generalise in the face of ambiguous data? While there has been a substantial amount of research studying the biases that learners employ, there has been very little work on what sorts of biases are employed in the face of data that is ambiguous between phonological generalisations with different degrees of complexity. In this article, we present the results from three artificial language learning experiments that suggest that, at least for phonotactic sequence patterns, learners are able to keep track of multiple generalisations related to the same segmental co-occurrences; however, the generalisations they learn are only the simplest ones consistent with the data.

A central question in language acquisition is how children master sentence types that they have seldom, if ever, heard. Here we report the findings of a pre-registered, randomised, single-blind intervention study designed to test the prediction that, for one such sentence type, complex questions (e.g., Is the crocodile who’s hot eating?), children could combine schemas learned, on the basis of the input, for complex noun phrases (the [THING] who’s [PROPERTY]) and simple questions (Is [THING] [ACTION]ing?) to yield a complex-question schema (Is [the [THING] who’s [PROPERTY]] ACTIONing?). Children aged 4;2 to 6;8 (M = 5;6, SD = 7.7 months) were trained on simple questions (e.g., Is the bird cleaning?) and either (Experimental group, N = 61) complex noun phrases (e.g., the bird who’s sad) or (Control group, N = 61) matched simple noun phrases (e.g., the sad bird). In general, the two groups did not differ on their ability to produce novel complex questions at test. However, the Experimental group did show (a) some evidence of generalising a particular complex NP schema (the [THING] who’s [PROPERTY] as opposed to the [THING] that’s [PROPERTY]) from training to test, (b) a lower rate of auxiliary-doubling errors (e.g., *Is the crocodile who’s hot is eating?), and (c) a greater ability to produce complex questions on the first test trial. We end by suggesting some different methods – specifically artificial language learning and syntactic priming – that could potentially be used to better test the present account.

The prevalence of redundancy in the world languages has long puzzled language researchers. It is especially surprising in light of the growing evidence on speakers' tendency to avoid redundant elements in production (omitting or reducing more predictable elements). Here, we propose that redundancy can be functional for learning. In particular, we argue that redundant cues can facilitate learning, even when they make the language system more complicated. This prediction is further motivated by the Linguistic Niche Hypothesis (Lupyan & Dale, 2010), which suggests that morphological complexity can arise due to the advantage redundancy might confer for child learners. We test these hypotheses in an artificial language learning study with children and adults, where either word order alone or both word order and case marking serve as cues for thematic assignment in a novel construction. We predict, and find, that children learning the redundant language learn to produce it, and show better comprehension of the novel thematic assignment than children learning the non-redundant language, despite having to learn an additional morpheme. Children in both conditions were similarly accurate in producing the novel word order, suggesting redundancy might have a differential effect on comprehension and production. Adults did not show better learning in the redundant condition, most likely because they were at ceiling in both conditions. We discuss implications for theories of language learning and language change.

Many languages exhibit differential object marking (DOM), where only certain types of grammatical objects are marked with morphological case. Traditionally, it has been claimed that DOM arises as a way to prevent ambiguity by marking objects that might otherwise be mistaken for subjects (e.g., animate objects). While some recent experimental work supports this account (Fedzechkina et al., 2012), research on language typology suggests at least one alternative hypothesis. In particular, DOM may instead arise as a way of marking objects that are atypical from the point of view of information structure. According to this account, rather than being marked to avoid ambiguity, objects are marked when they are given (already familiar in the discourse) rather than new. Here, we experimentally investigate this hypothesis using two artificial language learning experiments. We find that information structure impacts participants’ object-marking, but in an indirect way: atypical information structure leads to a change of word order, which then triggers increased object marking. Interestingly, this staged process of change is compatible with documented cases of DOM emergence (Iemmolo, 2013). We argue that this process is driven by two cognitive tendencies. First, a tendency to place discourse given information before new information, and second, a tendency to mark non-canonical word order. Taken together, our findings provide corroborating evidence for the role of information structure in the emergence of DOM systems.

Experience with native-language prosody encourages language-specific strategies for speech segmentation. Conflicting findings from previous research suggest that these strategies may not be abstracted away from the acoustic manifestation of prosodic features in the native speech. Using the artificial language learning paradigm, the current study explores this possibility in connection with listeners of a lexical tone language called Taiwanese Southern Min (TSM). In TSM, the only rising lexical tone occurs almost only on the final syllable of the language’s tone sandhi domain and is phonetically associated with final lengthening. Based on these observations, Experiment I examined what constituted a sufficient finality cue for use by TSM listeners to support segmentation: (a) final fundamental frequency (F0) rise only; or (b) final F0 rise conjoined with final lengthening. The results showed that segmentation was inhibited by the former cue but facilitated by the latter. Experiment II showed that the facilitation cannot be attributed entirely to final lengthening, as a null effect was found when final lengthening was the sole prosodic cue to segmentation. It is thus assumed that acoustic details as fine-grained as the lengthening of the rising tone are involved in the modulation of the segmentation strategy whereby TSM listeners perceive F0 rise as signaling finality. The inhibitory effect of final F0 rise alone found in Experiment I motivated Experiment III, which revealed that initial F0 rise in the absence of lengthening cues improved TSM listeners’ segmentation. It is speculated that such use of initial F0 rise might reflect a cross-linguistic segmentation solution.

This study uses an artificial language learning experiment and computational modelling to test Kiparsky's claims about Maximal Utilisation and Transparency biases in phonological acquisition. A Maximal Utilisation bias would prefer phonological patterns in which all rules are maximally utilised, and a Transparency bias would prefer patterns that are not opaque. Results from the experiment suggest that these biases affect the learnability of specific parts of a language, with Maximal Utilisation affecting the acquisition of individual rules, and Transparency affecting the acquisition of rule orderings. Two models were used to simulate the experiment: an expectation-driven Harmonic Serialism learner and a sequence-to-sequence neural network. The results from these simulations show that both models’ learning is affected by these biases, suggesting that the biases emerge from the learning process rather than any explicit structure built into the model.

This article reviews work that has employed artificial languages to investigate the learning and processing of additional language grammar in bilinguals, with a focus on morphosyntactic processing in sentence contexts. The article first discusses research that has utilized artificial languages to elucidate two central issues in research on bilingual third language learning and processing: the role of prior language-learning experience and cross-linguistic transfer from the native and second languages to the third. Then, research that has compared bilingual third language to monolingual second language grammar processing is discussed, with specific consideration of hypothesized bilingual advantages at language learning. Finally, future directions in artificial language learning research on bilingual morphosyntactic processing are considered.

There is substantial debate around the nature of letter position coding in reading. Research on a variety of Indo-European languages suggests uncertainty in position coding; for example, readers pe...

Languages differ regarding the depth, structure, and syntactic domains of recursive structures. Even within a single language, some structures allow infinite self-embedding while others are more restricted. For example, English allows infinite free embedding of the prenominal genitive -s, whereas the postnominal genitive of is largely restricted to only one level and to a limited set of items. Therefore, while the ability for recursion is considered as a crucial part of the language faculty, speakers need to learn from experience which specific structures allow free embedding and which do not. One effort to account for the mechanism that underlies this learning process, the distributional learning proposal, suggests that the recursion of a structure (e.g. X1’s-X2) is licensed if the X1 position and the X2 position are productively substitutable in the input. A series of corpus studies have confirmed the availability of such distributional cues in child directed speech. The present study further tests the distributional learning proposal with an artificial language learning experiment. We found that, as predicted, participants exposed to productive input were more likely to accept unattested strings at both one and two-embedding levels than participants exposed to unproductive input. Therefore, our results suggest that speakers can indeed use distributional information at one level to learn whether or not a structure is freely recursive.

Phonological alternations often happen to conform to phonotactic regularities, from which a single mechanism for phonotactics and alternations has been claimed. We note, however, that empirical evidence supporting the link between phonotactics and alternations comes only from English native speakers whose first language (L1) does exhibit phonotactically motivated alternation patterns. This article examines whether the link between phonotactics and alternations is universally available. To do so, we test learning of phonotactics and alternations with Cantonese native speakers, whose L1 provides no evidence for or against the link. We address learning of a vowel harmony pattern through the use of three artificial languages; one with a harmony pattern both within and across stems, another with a harmony pattern only across stems; and the other with a disharmony pattern within stems but harmony across stems. Learners successfully acquired harmony phonotactics according to input patterns, but they showed no difference in learning alternation patterns across the three languages. Our results suggest that the link between phonotactics and alternations might be language-specific: Only upon receiving L1 evidence, learners can use a unified mechanism to encode phonotactics and alternations.

The idea that universal representations of hierarchical structure constrain patterns of linear order is a central to many linguistic theories. In this paper we use Artificial Language Learning techniques to experimentally probe this claim. Specifically, we investigate how a hypothesized hierarchy of φ-features impacts the linearization of person and number affixes by (English-speaking) learners in the lab.

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.

Artificial language learning methods—in which learners are taught miniature constructed languages in a controlled laboratory setting—have become a valuable experimental tool for research on language development. These methods offer a complement to natural language acquisition data, allowing researchers to control both the input to learning and the learning environment. A large proportion of artificial language learning studies has aimed to understand the mechanisms of learning in infants. This review focuses instead on investigations into the nature of early linguistic representations and how they are influenced by both the structure of the input and the cognitive features of the learner. Looking not only at young infants but also at children beyond infancy, we discuss evidence for early abstraction, conditions on generalization, the acquisition of grammatical categories and dependencies, and recent work connecting the cognitive biases of learners to language typology. We end by outlining important areas for future research.

Morphological paradigms differ widely across languages: some feature relatively few contrasts, and others, dozens. Under the view that languages are under pressure to be learnable and that the distribution of languages in the world reflects biases in language learning, this diversity is surprising – how could paradigms which apparently differ so markedly be similarly learnable? Recent work on morphological complexity has argued that even very large paradigms are designed such that they are easy to learn and use. Specifically, Ackerman & Malouf, 2013 propose an information-theoretic measure, i-complexity, which captures the extent to which forms in one part of a paradigm predict forms elsewhere in the paradigm, and contrast this measure with e-complexity, which captures the number of distinctions made by the language and the different ways to mark each grammatical function. They show that languages which differ widely in e-complexity exhibit similar i-complexity; in other words, morphological paradigms with many contrasts reduce the learnability challenge for learners by having predictive relationships between forms. Here, we test whether i-complexity predicts the learnability of inflectional paradigms using both recurrent neural networks and human participants trained on an artificial language. Furthermore, we compare the effect of i-complexity on learning with that of e-complexity. We find that in RNNs both i- and e-complexity have an effect on learning: paradigms with lower i- and e-complexity are easier to learn, although the effect of e-complexity is larger. However, for human learners, we find only weak evidence (if any) that low i-complexity paradigms are easier to learn; in contrast, low e-complexity is clearly beneficial for learning. This suggests that i-complexity might have relatively little influence on the learnability of inflectional paradigms, with other factors, such as the e-complexity having a greater effect. These results suggest that appealing to i-complexity does not fully resolve the paradox of cross-linguistic variation in morphological systems.

Listeners tend to perceive longer vowels as word-final and include the aspiration of a syllable-onset consonant in metalinguistically judging the duration of the following vowel. It may be hypothesized that longer aspiration is interpreted as extra vowel duration and hence, a word-finality cue. Alternatively, if aspiration is perceived as part of the onset, longer aspiration should be interpreted as word-initial, consistent with previous findings on consonant lengthening [White, Mattys, Stefansdottir, and Jones (2015). J. Acoust. Soc. Am. 138(2), 1214–1220]. An artificial language learning experiment with Taiwanese Southern Min listeners showed that lengthening aspiration in word-initial but not word-final syllables improved speech segmentation, supporting the second but not the first hypothesis.

There is strong scientific consensus that emphasizing print-to-sound relationships is critical when learning to read alphabetic languages. Nevertheless, reading instruction varies across English-speaking countries, from intensive phonic training to multicuing environments that teach sound- and meaning-based strategies. We sought to understand the behavioral and neural consequences of these differences in relative emphasis. We taught 24 English-speaking adults to read 2 sets of 24 novel words (e.g., /buv/, /sig/), written in 2 different unfamiliar orthographies. Following pretraining on oral vocabulary, participants learned to read the novel words over 8 days. Training in 1 language was biased toward print-to-sound mappings while training in the other language was biased toward print-to-meaning mappings. Results showed striking benefits of print-sound training on reading aloud, generalization, and comprehension of single words. Univariate analyses of fMRI data collected at the end of training showed that print-meaning relative to print-sound relative training increased neural effort in dorsal pathway regions involved in reading aloud. Conversely, activity in ventral pathway brain regions involved in reading comprehension was no different following print-meaning versus print-sound training. Multivariate analyses validated our artificial language approach, showing high similarity between the spatial distribution of fMRI activity during artificial and English word reading. Our results suggest that early literacy education should focus on the systematicities present in print-to-sound relationships in alphabetic languages, rather than teaching meaning-based strategies, in order to enhance both reading aloud and comprehension of written words. (PsycINFO Database Record

This article investigates the influence of contextual pressures on the evolution of overspecification, i.e. the degree to which communicatively irrelevant meaning dimensions are specified, in an iterated learning setup. To this end, we combine two lines of research: In artificial language learning studies, it has been shown that (miniature) languages adapt to their contexts of use. In experimental pragmatics, it has been shown that referential overspecification in natural language is more likely to occur in contexts in which the communicatively relevant feature dimensions are harder to discern. We test whether similar functional pressures can promote the cumulative growth of referential overspecification in iterated artificial language learning. Participants were trained on an artificial language which they then used to refer to objects. The output of each participant was used as input for the next participant. The initial language was designed such that it did not show any overspecification, but it allowed for overspecification to emerge in 16 out of 32 usage contexts. Between conditions, we manipulated the referential context in which the target items appear, so that the relative visuospatial complexity of the scene would make the communicatively relevant feature dimensions more difficult to discern in one of them. The artificial languages became overspecified more quickly and to a significantly higher degree in this condition, indicating that the trend toward overspecification was stronger in these contexts, as suggested by experimental pragmatics research. These results add further support to the hypothesis that linguistic conventions can be partly determined by usage context and shows that experimental pragmatics can be fruitfully combined with artificial language learning to offer valuable insights into the mechanisms involved in the evolution of linguistic phenomena.

Previous work has demonstrated that during spoken word recognition, listeners can use a variety of cues to anticipate an upcoming sound before the sound is encountered. However, this vein of research has largely focused on local phenomena that hold between adjacent sounds. In order to fill this gap, we combine the Visual World Paradigm with an Artificial Language Learning methodology to investigate whether knowledge of a long-distance pattern of sibilant harmony can be utilized during spoken word recognition. The hypothesis was that participants trained on sibilant harmony could more quickly identify a target word from among a set of competitors when that target contained a prefix which had undergone regressive sibilant harmony. Participants tended to behave as expected for the subset of items that they saw during training, but the effect did not reach statistical significance and did not extend to novel items. This suggests that participants did not learn the rule of sibilant harmony and may have been memorizing which base went with which alternant. Failure to learn the pattern may have been due to certain aspects of the design, which will be addressed in future iterations of the experiment.

Traditionally, learning is assumed to take place with exposure to simpler elements first followed by exposure to elements with increasing levels of difficulty. Recent reports suggest that exposure to complex elements leads to more widespread changes. However, whether learning via exposure to complex or to simple elements is more beneficial is a matter of ongoing debate. In the current study, using behavioral and electrophysiological measures, we aimed at understanding this by comparing subjects trained with complex speech sounds with those trained with simple speech sounds in a 5-day pseudoword-picture training paradigm. We found that though the subjects learned both complex and simple speech sounds to similar degrees, subjects who were trained with complex stimuli demonstrated more generalizations to novel complex and simple stimuli, whereas those trained with simple stimuli exhibited generalization only to simple but not to complex stimuli (Experiment 1). Along with behavioral measures, using mismatch negativity, we found that training with complex stimuli can lead to more extensive neural changes for both complex and simple stimuli as compared with training with simple stimuli (Experiment 2). In artificial language learning, learning with complex stimuli appears to be more effective than training with simple stimuli as far as generalization is concerned. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Many languages restrict their lexicons by OCP-Place, a phonotactic constraint against co-occurrences of consonants with shared [place] (e.g., McCarthy, 1986). While many previous studies have suggested that listeners have knowledge of OCP-Place and use this for speech processing, it is less clear whether they make reference to an abstract representation of this constraint. In Dutch, OCP-Place gradiently restricts non-adjacent consonant co-occurrences in the lexicon. Focusing on labial-vowel-labial co-occurrences, we found that there are, however, exceptions from the general effect of OCP-Labial: (A) co-occurrences of identical labials are systematically less restricted than co-occurrences of homorganic labials, and (B) some specific pairs (e.g., /pVp/, /bVv/) occur more often than expected. Setting out to study whether exceptions such as (A) and (B) had an effect on processing, the current study presents an artificial language learning experiment and a reanalysis of Boll-Avetisyan and Kager's (2014) speech segmentation data. Results indicate that Dutch listeners can use both knowledge of phonotactic detail and an abstract constraint OCP-Labial as a cue for speech segmentation. We suggest that whether detailed or abstract representations are drawn on depends on the complexity of processing demands.

Non-adjacent dependencies are ubiquitous in language, but difficult to learn in artificial language experiments in the lab. Previous research suggests that non-adjacent dependencies are more learnable given structural support in the input - for instance, in the presence of high variability between dependent items. However, not all non-adjacent dependencies occur in supportive contexts. How are such regularities learned? One possibility is that learning one set of non-adjacent dependencies can highlight similar structures in subsequent input, facilitating the acquisition of new non-adjacent dependencies that are otherwise difficult to learn. In three experiments, we show that prior exposure to learnable non-adjacent dependencies - i.e., dependencies presented in a learning context that has been shown to facilitate discovery - improves learning of novel non-adjacent regularities that are typically not detected. These findings demonstrate how the discovery of complex linguistic structures can build on past learning in supportive contexts.

The right edge of the word is a known domain for processes like phonological devoicing. This has been argued to be the effect of analogy from higher prosodic domains, rather than an in situ motivated change (Hock 1999, Hualde and Eager 2016). Phonetic word-level phenomena of final lengthening and final devoicing have been found to occur natively word-finally (Lunden 2006, 2017, Nakai et al. 2009) despite claims that they have no natural phonetic pressure originating in this position (Hock 1999). We present the results of artificial language learning studies that seek to answer the question of whether phonetic-level cues to the word-final position can aid in language parsing. If they do, it provides evidence that listeners can make use of word-level phonetic phenomena, which, together with studies that have found them to be present, speaks to their inherent presence at the word level. We find that adult listeners are better able to recognize the words they heard in a speech stream, and better able to reject words that they did not hear, when final lengthening was present at the right edge of the word. Final devoicing was not found to give the same boost to parsing.