Recent experimental literature has investigated across-scale variation in scalar implicature calculation, probing why lexical scales differ from each other in their likelihood of being strengthened (e.g., old → not ancient vs. smart → not brilliant). But in existing studies of this scalar diversity, less attention has been paid to potential variation introduced by the carrier sentences that scales occur in. In this paper, we carry out a systematic investigation of the role of sentential context on scalar diversity, focusing on scales formed by two gradable adjectives. We find within-scale variation: different subject nouns (e.g. The employee is smart v. The scientist is smart) have a significant effect on how robustly a scalar implicature arises. We then explore the relationship between a noun’s prior likelihood of exhibiting the stronger adjectival property (e.g. brilliance) and the rate of implicature calculation, and find that they are negatively correlated. We also test whether a previously identified factor in scalar diversity, adjectival threshold distance between the weaker (smart) and stronger (brilliant) adjective, is sensitive to the subject noun manipulation, but do not find evidence for this. In addition to their theoretical import, our findings also highlight the methodological importance of controlling carrier sentences.

@article{aparicio_scalar_2025,
 title={Scalar implicature rates vary within and across adjectival scales},
 author={Aparicio, Helena and Ronai, Eszter},
 journal={Journal of Semantics},
 year={2025},
 month={mar}
}

A central endeavor in psycholinguistic research has been to determine the processing profile of syntactically ambiguous strings. Previous work investigating syntactic attachment ambiguities has shown that discarding a locally grammatically available, but globally failing, parse is costly. However, little is known about how comprehenders cope with semantic parsing ambiguities. Using the case study of scopally ambiguous definite descriptions such as the rabbit in the big hat, we examine whether comparable penalties arise for non-lexical semantic ambiguities. In a series of reference resolution tasks, we find dispreference for strings that are globally defined but fail to refer under alternative semantic parses, compared to strings where all readings successfully refer to the same individual. Crucially, this effect is only detectable when the alternative failing reading gives rise to a referential garden path, where a dynamic constraint evaluation process temporarily settles on a unique referent before eventually failing. We conclude that failing alternative readings cause dispreference for a definite description, but only when the failing interpretation constitutes a red herring.

@article{aparicio_beware_2025,
 title={Beware of referential garden paths! The dangerous allure of semantic parses that succeed locally but globally fail},
 author={Aparicio, Helena and Levy, Roger and Coppock, Elizabeth},
 journal={Glossa Psycholinguistics},
 volume={4},
 year={2025},
 month={jan}
}

Speakers often choose to utter imprecise sentences that, albeit felicitous,are, strictly speaking, false (e.g., using ‘This bottle is empty’ to describe a bottle witha bit of water in it). The acceptability of an imprecise utterance hinges on the standardof precision (SoP), a discourse parameter that governs how much imprecision is toler-ated in a context. Previous theoretical accounts (e.g., Lewis 1979, Klecha 2018) haveargued that metalinguistic denials that target the assertability of an imprecise utterance(e.g., ‘No, this bottle is not empty!’) more or less force accommodation to a higher SoP.The present study investigates the nature of this accommodation process. In particular,we ask whether metalinguistic disagreements result in an automatic update of the SoP.In two acceptability judgment experiments, we show that imprecise utterances are notdeemed unacceptable when embedded in a disagreement dialogue. Our findings in-stead suggest that metalinguistic denials act as a request to raise the SoP and that anypotential updates ought to be signaled overtly in subsequent conversational moves.

@article{wu_disagreements_2025,
 title={Disagreements do not automatically raise the standard of precision},
 author={Wu, Yifan and Aparicio, Helena},
 journal={Experiments in Linguistic Meaning},
 volume={3},
 pages={435--446},
 year={2025},
 month={jan},
 url={https://journals.linguisticsociety.org/proceedings/index.php/ELM/article/view/5835}
}

Disagreements are speech acts used by interlocutors to challenge previous assertions. When disagreements express subjective views, they can often be perceived as faultless. However, it is unclear whether accepting a disagreement as faultless causes comprehenders to update their own semantic representations of the predicate targeted by the disagreement. Using the vague quantifiers extit{many} and extit{few} as a case study, we find in two adaptation studies that participants shifted their meaning representations of the quantifiers after being exposed to disagreements that on average were more likely to be perceived as faultless. The adaptation strengthened the participants' baseline preferences, suggesting that even when a disagreement is judged to be faultless, there exists a perceived asymmetry in the plausibility of the two viewpoints under discussion.

@article{pecsok_how_2024,
 title={How can they both be right?: Faultless disagreement and semantic adaptation},
 author={Pecsok, Emily and Aparicio, Helena},
 journal={Proceedings of the Annual Meeting of the Cognitive Science Society},
 volume={46},
 year={2024}
}

People derive contrastive inferences when interpreting adjectives (e.g., inferring that 'the short pencil' is being contrasted with a longer one). However, classic eye-tracking studies revealed contrastive inferences with scalar and material adjectives, but not with color adjectives. This was explained as a difference in listeners' informativity expectations, since color adjectives are often used descriptively (hence not warranting a contrastive interpretation). Here we hypothesized that, beyond these pragmatic factors, perceptual factors (i.e., the relative perceptibility of color, material and scalar contrast) and semantic factors (i.e., the difference between gradable and non-gradable properties) also affect the real-time derivation of contrastive inferences. We tested these predictions in three languages with prenominal modification (English, Hindi, and Hungarian) and found that people derive contrastive inferences for color and scalar adjectives, but not for material adjectives. In addition, the processing of scalar adjectives was more context dependent than that of color and material adjectives, confirming that pragmatic, perceptual and semantic factors affect the derivation of contrastive inferences.

@article{ronderos_perceptual_2024,
 title={Perceptual, Semantic, and Pragmatic Factors Affect the Derivation of Contrastive Inferences},
 author={Ronderos, Camilo R. and Aparicio, Helena and Long, Madeleine and Shukla, Vishakha and Jara-Ettinger, Julian and Rubio-Fernandez, Paula},
 journal={Open Mind},
 volume={8},
 pages={1213--1227},
 year={2024},
 month={oct}
}

Recent experimental literature has investigated across-scale variation in scalar implicature calculation: lexical scales significantly differ from each other in how likely they are to be strengthened (e.g., old → not ancient vs. smart → not brilliant). But in existing studies of this scalar diversity, not enough attention has been paid to potential variation introduced by the carrier sentences that scales occur in. In this paper, we carry out the first systematic investigation of the role of sentential context on scalar diversity. Focusing on scales formed by two gradable adjectives, we manipulate the comparison class, specifically whether a noun is likely to have the property described by the scalar adjective (e.g., brilliant employee vs. brilliant scientist). Our results show within-scale variation: a significant effect of comparison class on the likelihood of scalar implicature calculation. We explain this result in terms of the adjectival threshold distance between the weaker (smart) and stronger (brilliant) adjective, conditioned on the comparison class (employee vs. scientist). Our findings also highlight the methodological importance of controlling carrier sentences.

@inproceedings{aparicio2023scalar,
 title={Scalar implicature rates vary within and across adjectival scales}, 
author={Aparicio, Helena and Ronai, Eszter},
 booktitle={Semantics and Linguistic Theory},
 pages={110--130},
 year={2023}
 }

Treebanks have traditionally included only text and were derived from written sources such as newspapers or the web. We introduce the Aligned Multimodal Movie Treebank, an English language treebank derived from naturalistic dialog in Hollywood movies which includes the source video and audio, transcriptions with word-level alignment to the audio stream, as well as part of speech tags and dependency parses in the Universal Dependencies formalism. AMMT consists of 31,264 sentences and 218,090 words, that will be the 3rd largest UD English treebank, and the only multimodal treebank in UD. To help with the web-based annotation effort, we also introduce the Efficient Audio Alignment Annotator (EAAA), a companion tool that enables annotators to speed-up significantly the annotation process.

@inproceedings{yaari-etal-2022-aligned,
 title={The Aligned Multimodal Movie Treebank: An audio, video, dependency-parse treebank},
 author={Yaari, Adam and DeWitt, Jan and Hu, Henry and Stankovits, Bennett and Felshin, Sue and Berzak, Yevgeni and Aparicio, Helena and Katz, Boris and Cases, Ignacio and Barbu, Andrei},
 booktitle={Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing},
 month={dec},
 year = {2022},
 address = {Abu Dhabi, United Arab Emirates},
 publisher = {Association for Computational Linguistics},
 url = {https://aclanthology.org/2022.emnlp-main.648},
 pages={9531—9539}
}

This paper makes the novel observation that definite comparatives, such as the bigger circle, impose restrictions on the cardinality of the comparison class (CC) against which their truth conditions are evaluated. We show that the corpus frequency counts of definite comparatives sharply drop when the comparison class used for their interpretation is formed by more than two individuals. Two alternative theories of these distributional facts are considered and tested experimentally through an acceptability judgment task. According to the first theory, the 2-Individuals Theory, definite comparatives presuppose that the CC is of cardinality 2; under the second theory, the 2-Degrees Theory, the meaning of the comparative is evaluated against a granularity γ that maps the individuals in the CC to degrees in the relevant adjectival scale, and definite comparatives presuppose that the set of the degrees resulting from this mapping is of cardinality 2. Our results show that definite comparative descriptions are most frequent and felicitous when evaluated against comparison classes with two individuals, but also that acceptability drops off with higher cardinalities in a gradient manner that is sensitive to granularity. Taken together, these findings argue against the 2-Individuals theory of definite comparatives and lend support to the 2-Degrees theory.

@inproceedings{aparicio2022granularity,
 title={Granularity in the Semantics of Comparison},
 author={Aparicio, Helena and Chen, Curtis and Levy, Roger and Coppock, Elizabeth},
 booktitle={Semantics and Linguistic Theory},
 volume={31},
 pages={550--569},
 year={2022}
}

Abstract Referential Effects of Contrast (RECs) involving reference resolution of adjectivally modified NPs (e.g. the tall glass) have been attributed to pragmatic reasoning based on the informativity of modification (Sedivy et al. 1999; Sedivy 2003, 2004, a.o.). Recently, it has been claimed that informativity alone cannot account for all the attested interactions between adjectival meaning and context and that factors related to efficiency in the search of a referent also play an important role (Rubio-Fernández 2016). Building on Aparicio et al. (2015), this paper demonstrates that perceived informativity plays an important role in RECs, but lexical semantic properties of different adjective classes are also relevant. We present results from a Visual World eye-tracking study which shows that adjective classes differ in whether they introduce RECs, and results from an offline judgment task which show that this difference correlates to some extent with the perceived informativity of members of these classes. Color adjectives, relative adjectives and maximum standard absolute adjectives were rated as overinformative when used as modifiers in the absence of contrast, and gave rise to RECs; minimum standard absolute adjectives were not rated as overinformative when used as modifiers in the absence of contrast, and did not give rise to RECs. Taken together, our results show that perceived informativity plays an important role in RECs. We also discuss additional differences between the adjective classes which suggest that differences in lexical semantics can further contribute to differences in RECs.

@incollection{aparicio2018perceived,
 title={Perceived informativity and referential effects of contrast in adjectivally modified NPs},
 author={Aparicio, Helena and Kennedy, Christopher and Xiang, Ming},
 booktitle={The semantics of gradability, vagueness, and scale structure},
 pages={199--220},
 year={2018},
 publisher={Springer}
 }

Both relative adjectives (RAs) like ‘big’ and absolute adjectives (AAs) like ‘empty’ are sensitive to the context: in the former case, the context determines how much size is required to count as big; in the latter, the context determines how much deviation from total emptiness is allowed to count as empty. Whereas it is generally agreed that the role of context with RAs is to fix the value of a threshold variable, the status of absolute adjective thresholds, and therefore the role of context in their interpretation, remains an object of debate. Some researchers have argued that all gradable adjectives have context-sensitive threshold variables that are assigned values by the same mechanisms (Lassiter & Goodman 2013). Others have claimed that AAs have fixed, endpoint-oriented meanings and that sensitivity to context arises from pragmatic reasoning about imprecision (Kennedy 2007; Syrett, Kennedy & Lidz 2009; van Rooij 2011; Burnett 2014; Qing & Franke 2014). In an eye-tracking Visual World experiment, we investigate RAs and AAs used as restrictive modifiers. We find that target identification is significantly faster for both types of adjectives when the visual context supports a restrictive interpretation of the predicate, although this effect is considerably delayed in the case of AAs. We conclude that for RAs, the target facilitation effect is driven by the lexical semantics of the predicate itself. However, it is argued that the extra processing cost observed with AAs results from pragmatic reasoning about imprecision.

@inproceedings{aparicio2016processing,
 title={Processing gradable adjectives in context: A visual world study},
 author={Aparicio, Helena and Xiang, Ming and Kennedy, Christopher},
 booktitle={Semantics and Linguistic Theory},
 volume={25},
 pages={413--432},
 year={2016}
 }

@article{aparicio2014syntax,
 title={The Syntax of Ellipsis Resolution: Eye-tracking Evidence from $\varphi$-Feature Mismatches},
 author={Aparicio, Helena and Franich, Katie and Xiang, Ming},
 year={2014}
 }

Subset comparatives (Grant 2013) are amount comparatives in which there exists a set membership relation between the target and the standard of comparison. This paper argues that subset comparatives should be treated as regular phrasal comparatives with an added presuppositional component. More specifically, subset comparatives presuppose that: a) the standard has the property denoted by the target; and b) the standard has the property denoted by the matrix predicate. In the account developed below, the presuppositions of subset comparatives result from the compositional principles independently required to interpret those phrasal comparatives in which the standard is syntactically contained inside the target. Presuppositions are usually taken to be licensed by certain lexical items (presupposition triggers). However, subset comparatives show that presuppositions can also arise as a result of semantic composition. This finding suggests that the grammar possesses more than one way of licensing these inferences. Further research will have to determine how productive this latter strategy is in natural languages.

@inproceedings{terrasa2014compositional,
 title={A compositional analysis for subset comparatives},
 author={Terrasa, Helena Aparicio},
 booktitle={Proceedings of Sinn und Bedeutung},
 volume={18},
 pages={24--41},
 year={2014}
}