Publications

How does linguistic negation develop in early child language? Prior research has suggested that abstract and context-general negation develops from concrete and context-specific communicative functions such as rejection, prohibition, or nonexistence in fixed and ordered stages. The evidence for the emergence of these functions in stages is mixed, however, leaving the possibility that negation starts as an abstract concept that can serve multiple specific functions from the beginning, and that the development of the different functions starts more or less simultaneously depending on the early communicative environment. Leveraging automatic annotations of large-scale child speech corpora in English and growth-curve modeling, we examine children’s production of seven negative constructions that tend to convey communicative functions previously discussed in the literature. We also investigate children’s discourse-level negative responses (saying no) to parents’ utterances with the same constructions as a proxy for children’s comprehension. We do not find strong evidence for population-level stages in children’s development of negation. Instead, the results of our growth-curve modeling suggest that for our measures of comprehension and production, children’s ability to negate different constructions likely emerges around 18–22 months of age. Our results complement and confirm recent findings in experimental studies on children’s comprehension of negation.

Disjunction in natural language alternates between an inclusive reading (A or B or Both) and an exclusive  reading (A or B but not Both). Traditional accounts of this ambiguity focus on scalar implicature as the source of disjunction exclusivity, a process whereby Gricean reasoning over Horn scales strengthens the baseline inclu-sive  reading  to  an  implied  exclusive  reading  (Grice,  1978;  Horn,  1972;  Gazdar, 1980). Despite nearly all theories acknowledging that other factors likely play a role in  the  generation  of  exclusivity  implications,  non-implicature  factors  have  received comparatively  little  attention. Across  four  experiments  we  tested  two  such  non  im-plicature factors, prior compatibility and syntactic category, finding that both play a role in speaker interpretations of disjunctive sentences. Additionally, by drawing our stimuli in the first two experiments from the prior literature, we found evidence that previous  research  on  disjunction,  while  accurately  identifying  the  key  role  of  scalar implicatures, may be overestimating the effect size thereof due to a failure to control for non-implicature factors.

How can modern neural networks like language models be useful to the field of language acquisition, and more broadly cognitive science, if they are not a priori designed to be cognitive models? As developments towards natural language understanding and generation have improved leaps and bounds, with models like GPT-4, the question of how they can inform our understanding of human language acquisition has re-emerged. As such, it is critical to examine how in practice linking hypotheses between models and human learners can be safely established. To address these questions, we propose a model taxonomy, including four modelling approaches, each having differing goals, from exploratory hypothesis generation to hypothesis differentiation and testing. We show how the goals of these approaches align with the overarching goals of science and linguistics by connecting our taxonomy to the realist versus instrumentalist approaches in philosophy of science. We survey recent work having adopted each of our modelling approaches and address the importance of computational modelling in language acquisition studies.

How can modern neural networks like language models be useful to the field of language acquisition, and more broadly cognitive science, if they are not a priori designed to be cognitive models? As developments towards natural language understanding and generation have improved leaps and bounds, with models like GPT-4, the question of how they can inform our understanding of human language acquisition has re-emerged. As such, it is critical to examine how in practice linking hypotheses between models and human learners can be safely established. To address these questions, we propose a model taxonomy, including four modelling approaches, each having differing goals, from exploratory hypothesis generation to hypothesis differentiation and testing. We show how the goals of these approaches align with the overarching goals of science and linguistics by connecting our taxonomy to the realist versus instrumentalist approaches in philosophy of science. We survey recent work having adopted each of our modelling approaches and address the importance of computational modelling in language acquisition studies.

In English binomial expressions, ‚"bread and butter"‚ is preferred over ‚"butter and bread". Morgan & Levy (2015) show that for these types of expressions, frequently used phrases tend to have stronger, more extreme preferences. In contrast, there is roughly an equal preference for ‚"bishops and rooks‚" versus ‚"rooks and bishops‚" a much less common pairing. This paper extends this research to the concept of Mandarin elastic word length, a phenomenon in which most Mandarin words have long and short forms.  We find evidence for frequency-dependent regularization in the elastic length of Noun-Noun compounds in Chinese, demonstrating that frequency-dependent regularization extends to structures with more than two alternations and to languages other than English.

Veltman’s test semantics and developments thereof reject the canon about semantic contents and attitude ascriptions in favor of dynamic alternatives. According to these theories the semantic content of a sentence is not a proposition, but a context change potential (CCP). Similarly, beliefs are not taken to be relations between agents and propositions, but agents and CCPs. These deviations from the canon come at the cost of an elegant explanation about the correctness of belief. Standardly, it is taken that the content of a belief is correct just in case the content of that belief is true. The proponent of the test semantics cannot appeal to this explanation since they hold that certain contents, namely epistemically modalized contents, do not express propositions, and are neither true nor false. After motivating the need for an account of the correctness of belief within the dynamic framework, I argue that the few extant accounts in the literature are unsatisfactory. A novel account is proposed that avoids the problems of competing views. 

Previous research has hypothesized default interpretive biases for three types of ambiguities with English logical words and, or, and not. First, disjunction (A or B) is hypothesized to be biased towards an exclusive interpretation in upward-entailing environments and an inclusive interpretation in downward-entailing environments (Levinson 2000, Chierchia 2004, Breheny et al. 2005). A negated disjunction (not A or B) is claimed to be biased towards a “neither-nor” interpretation (i.e. wide scope negation: ¬[A ∨ B]) and a negated conjunction is said to be biased towards an “either-not” interpretation (i.e. wide-scope negation: ¬[A ∧ B]) (Szabolcsi 2002, Szabolcsi & Haddican 2004). We tested these hypotheses within the same experimental paradigm with 149 English-speaking participants and found disjunction to be biased towards an inclusive interpretation across three different entailment environments: episodic declaratives, questions, and conditional antecedents. Our results also confirmed that English negated disjunction is biased towards a “neither-nor” (wide scope negation) interpretation but the results did not show an “either-not” bias (wide scope negation) for negated conjunction.

What are the constraints, cues, and mechanisms that help learners create successful word-meaning mappings? This study takes up linguistic disjunction and looks at cues and mechanisms that can help children learn the meaning of or. We first used a large corpus of parent-child interactions to collect statistics on or uses. Children started producing or between 18-30 months and by 42 months, their rate of production reached a plateau. Second, we annotated for the interpretation of disjunction in child-directed speech. Parents used or mostly as exclusive disjunction, typically accompanied by rise-fall intonation and logically inconsistent disjuncts. But when these two cues were absent, disjunction was generally not exclusive. Our computational modeling suggests that an ideal learner could successfully interpret an English disjunction (as exclusive or not) by mapping forms to meanings after partitioning the input according to the intonational and logical cues available in child-directed speech.

Previous research has hypothesized default interpretive biases for three types of ambiguities with English logical words and, or, and not. First, disjunction (A or B) is hypothesized to be biased towards an exclusive interpretation in upward-entailing environments and an inclusive interpretation in downward-entailing environments (Levinson 2000, Chierchia 2004, Breheny et al. 2005). A negated disjunction (not A or B) is claimed to be biased towards a “neither-nor” interpretation (i.e. wide scope negation: ¬[A ∨ B]) and a negated conjunction is said to be biased towards an “either-not” interpretation (i.e. wide-scope negation: ¬[A ∧ B]) (Szabolcsi 2002, Szabolcsi & Haddican 2004). We tested these hypotheses within the same experimental paradigm with 149 English-speaking participants and found disjunction to be biased towards an inclusive interpretation across three different entailment environments: episodic declaratives, questions, and conditional antecedents. Our results also confirmed that English negated disjunction is biased towards a “neither-nor” (wide scope negation) interpretation but the results did not show an “either-not” bias (wide scope negation) for negated conjunction.

How does abstract linguistic negation develop in early child language? Previous research has suggested that abstract negation develops in stages and from more concrete communicative functions such as rejection, prohibition, or non-existence. The evidence for the emergence of these functions in stages is mixed, however, leaving the possibility that negation is an abstract concept from the beginning that can serve multiple specific functions depending on early communicative environment. Leveraging automatic annotations of large-scale child speech corpora in English, we examine the production trajectories of seven negative constructions that tend to convey communicative functions previously discussed in the literature. The results demonstrate the emergence and gradual increase of these constructions in child speech within the age range of 18-36 months. Production mostly remains stable, regular, and close to parents’ levels after this age range. These findings are consistent with two hypotheses: first, that negation starts as an abstract concept that can serve multiple functions from the beginning; and second, that negation develops in stages from specific communicative functions but this development is early and quick, leaving our corpus methods incapable of detecting them from the available corpus data.

Previous research has proposed several stages for children’s production of negative morphemes. For example, Cameron-Faulkner, Lieven, and Theakston (2007) proposed that English negative morphemes appear with a no>not>n’t order in children’s speech. Klima and Bellugi (1966) proposed that negation first appears outside the sentence and later moves inside between the subject and the verb. They also proposed that can’t and don’t are learned as unanalyzed wholes before their positive auxiliary variants. However, comprehension studies have not provided evidence for such stages yet (Austin et al. 2014; Feiman et al. 2017; Reuter, Feiman, and Snedeker 2018). This discrepancy can be explained in two ways. First, the lack of evidence may be due to limitations in comprehension studies. Second, the proposed stages may be limited to production and not generalizable to comprehension. This paper presents two exploratory corpus studies that support the second possibility. The results suggest that some previous stage hypotheses do not hold generally and may be limited to a few children. Furthermore, stages that do hold across children may be limited to production only.

Which quantificational forces do languages encode lexically? When a language features multiple quantificational scales (e.g. determiner and adverbial quantification), does the pattern of lexicalization of quantificational forces we discover for one scale correlate with those of other scales? We use English as a first test case for examining these questions, adapting the basic ideas of Lewis (1975) into the hypothesis that English lexical quantifiers unrelated to cardinal numbers or definite descriptions, determiner and adverbial alike, have one of six quantificational forces. To begin to test this claim empirically, we elicited speaker interpretations of a range of quantifiers in a web-based study. Dividing participants into an adverbial condition and a determiner condition, we gave a context specifying a 100-day period and asked participants to judge the quantificational force of quantified sentences denoting an individual’s daily activities during this period. We found evidence of cross-scale correspondences but fewer quantificational forces than expected. These results provide preliminary evidence for parts of our hypothesis but suggest a need for future research that covers more lexical items, languages, and quantificational scales. 

Readability assessment aims to automatically classify text by the level appropriate for learning readers. Traditional approaches to this task utilize a variety of linguistically motivated features paired with simple machine learning models. More recent methods have improved performance by discarding these features and utilizing deep learning models. However, it is unknown whether augmenting deep learning models with linguistically motivated features would improve performance further. This paper combines these two approaches with the goal of improving overall model performance and addressing this question. Evaluating on two large readability corpora, we find that, given sufficient training data, augmenting deep learning models with linguistically motivated features does not improve state-of-the-art performance. Our results provide preliminary evidence for the hypothesis that the state-of-the-art deep learning models represent linguistic features of the text related to readability. Future research on the nature of representations formed in these models can shed light on the learned features and their relations to linguistically motivated ones hypothesized in traditional approaches.

Altough it is widely acknowledged that Tehrani Persian (often broadly labeled as Persian) has no dedicated marker of definiteness, the nominal suffix "-e" has been analyzed as a colloquial definiteness marker. Here I show that "-e" can mark bare nominals to ensure a definite interpretation, but it can also appear on indefinites marked by the indefinite determiner "ye". I show that indefinites marked by "-e" are scopally inert. To unify the effect of "-e" on definites and indefinites, I propose that "-e" introduces a uniqueness implication on the nominal it modifies. More specifically, N-e denotes a singleton set of objects. On a bare nominal, this uniqueness implication ensures a definite interpretation. On an indefinite, it restricts the domain of quantification to a singleton, making the indefinite scopally inert. I present a compositional account of definite and indefinite constructions with "-e" in Tehrani Persian.

The past 15 years have seen increasing experimental investigations of core pragmatic questions in the ever more active and lively field of experimental pragmatics. Within experimental pragmatics, many of the core questions have relied on the operationalization of the theoretical notion of “implicature rate.” Implicature rate based results have informed the work on acquisition, online processing, and scalar diversity, inter alia. Implicature rate has typically been quantified as the proportion of “pragmatic” judgments in two-alternative forced choice truth value judgment tasks. Despite its theoretical importance, this linking hypothesis from implicature rate to behavioral responses has never been extensively tested. Here we show that two factors dramatically affect the “implicature rate” inferred from truth value judgment tasks: (a) the number of responses provided to participants; and (b) the linking hypothesis about what constitutes a “pragmatic” judgment. We argue that it is time for the field of experimental pragmatics to engage more seriously with its foundational assumptions about how theoretical notions map onto behaviorally measurable quantities, and present a sketch of an alternative linking hypothesis that derives behavior in truth value judgment tasks from probabilistic utterance expectations.

The development of the ubiquitous logical connectives "and" and "or" provides a window into the role of semantics and pragmatics in children's linguistic development. Previous research suggested that adults and children might differ in their interpretation of "or" in two ways. First, unlike adults, children might interpret "or" as "and". Second, children might interpret "or" as inclusive disjunction while adults interpret it as exclusive. We report experimental studies that probe interpretations of "and" and "or" in adults and children using truth value judgements and children's spontaneous linguistic feedback. Both measures showed that four-year-olds do not interpret "or" as "and". Children's truth judgments suggested that they did not derive exclusivity implicatures. Yet their corrective feedback showed signs of sensitivity to the implicature, suggesting that the truth judgement task could have underestimated children's pragmatic competence. More generally, four-year-olds' interpretation of logical connectives may not be as different from adults as previously supposed.

This paper investigates three indefinite constructions in Persian: simple (ye-NP), complex (ye-NP-i), and antidefinite (NP-i). It shows that simple indefinites with the determiner ye carry an at-issue existence implication (|NP| ≥ 1), similar to their English counterpart with the indefinite determiner a(n). Complex indefinites with both ye and -i introduce an antisingleton implication (|NP| > 1), similar to Spanish indefinites with algún. Antidefinites with only the clitic -i are a novel category which trigger a projective non-uniqueness implication (|NP| ̸= 1). The paper provides a compositional account in which the antisingleton implication of complex indefinites is derived from the existence implication of the determiner "ye" and the non-uniqueness implication of the clitic "-i". The predictions of this account as well as the pragmatic effects of complex indefinites such as ignorance, indifference, free choice, and domain widening implications are discussed.