jpguevara [at] umass (dot) edu
Field psycholinguist. Experimental syntactician.
NSF SBE Postdoctoral Research Fellow
Prepping another round of visual world experiments for my research trip to the Philippines this December.
Updated: 18 October 2023
My research is an effort to understand human linguistic capacity. How do we coordinate multiple sources of information to comprehend language? How is this coordination affected by the structure of our language? My research draws inspiration from linguistics and cognitive psychology, and can best be described as an intersection of psycholinguistics—the study of how grammatical principles are deployed in real-time comprehension, and how these interact with the cognitive architecture of humans—and linguistic fieldwork—a collaborative enterprise between a linguist and a community of speakers to document and analyze the structure of a particular language.
I focus primarily on the description and analysis of Austronesian languages, a family of some 1200 languages dispersed over Southeast Asia and the Pacific. I specialize in Tagalog and using types of data available to me as a bilingual speaker, fieldworker/experimental syntactician, I produce a fuller, empirically richer picture of Tagalog morphosyntax. This richer picture then sets the stage for my experimental investigations. As a psycholinguist, I am interested in the types of linguistic information that comprehenders attend to and use when processing non-local dependencies. I look at processing asymmetries to see how these classes of information interact.
In my research, I leverage linguistic features of "smaller languages" like Tagalog (and Dabaw Bisaya [db] and Santiago Laxopa Zapotec [slz]) that are typologically un(der)represented in psycholinguistics. I have also been (i) adapting methodologies to make them more deployable in contexts where data collection cannot be conducted in a typical laboratory setting, (ii) testing the robustness of empirical generalizations on a larger scale, and (iii) taking seriously the role of low-level task effects and decision-making when explaining participants' behavior.
This is work in progress and is in collaboration with Brian Dillon.
In English, abstract syntactic principles regulate coreference in reflexives (Chomsky, 1981). There is some debate, however, on the status of these constraints in Austronesian languages. For example, some scholars argue that these languages have semantically-grounded constraints on reflexive interpretation, such as a thematic hierarchy (e.g. Andrews (1985) for Tagalog; Bell (1976) for Cebuano; and Van Valin (1999) for Toba Batak, another related language spoken in Indonesia). On the other hand, some have argued that much like English, these languages have structurally-grounded constraints (Rackowski, 2002; Richards, 2013). The goal of this project is to address this debate by combining fieldwork and large-scale experimentation.
To investigate the role of grammatical constraints on resolving a reflexive’s real-time interpretation, researchers have largely used variants of Sturt’s mismatch paradigm (2003). For example, researchers have examined the extent to which comprehenders momentarily consider DPs not licensed by Principle A (henceforth, distractor) as they process the reflexive. In this paradigm, researchers systematically manipulate the features of the distractor, and measure whether comprehenders are sensitive to the matching features between the distractor and the actual antecedent. Consider: The new executives who oversaw the middle manager(s) doubted themselves... To the extent that themselves is processed differently when the distractor is singular and when it is plural indexes attention to the features of the distractor. This indicates activation during the course of resolving the reflexive’s reference. Any influence exerted by the distractor is referred to as interference effects.
I am investigating the extent to which Tagalog comprehenders attend to syntactic factors (e.g., c-command and locality) and to semantic factors (e.g., thematic relations) during the real-time comprehension of reflexives. To this end, I have conducted a large-scale offline antecedent selection experiment to investigate reflexive licensing. Consistent with Principle A, participants overwhelmingly chose as antecedents DPs that were in the same clause and that c-commanded the reflexive. DPs not licensed by Principle A—either because they did not c-command or they were not in the same clause as the reflexive, or both—had very little impact on participants’ ultimate interpretation. This part of the project has been presented at the 15th Philippine Linguistic Congress and at the 30th Annual Meeting of the Austronesian Formal Linguistics Association.
I have also conducted a self-paced reading experiment using a number-mismatch paradigm to see whether distractors were momentarily considered as antecedents. Participants’ responses to comprehension questions replicated the offline interpretations. However, their reading times suggested that they were momentarily considering the distractor in a way that is not predicted by the models of retrieval espoused by Lewis & Vasishth (2005) or by Dillon et al. (2013). A more articulated version of this research is available here. In another self-paced reading experiment, I have partially replicated these findings. Responses to comprehension questions were again replicated. Their reading times, however, provided no evidence that they were attending to the distractor. While I had collected data from 104 participants, only data from 58 were left after adopting the same exclusion criteria as the earlier study. The results were trending in the same direction, suggesting that this failure to replicate could be due to low power. These inconsistencies echo recent research by Jäger et al. (2017) that suggested that effect sizes of interference effects in reading measures are small and easily require 100+ participants to yield sufficient power to detect.
The unclear empirical picture to date suggests that an alternative experimental methodology would be helpful to investigate interference effects. For this reason, I adopted a fairly underutilized paradigm to measure real-time interference: the visual world paradigm. An additional benefit of this paradigm is that the linking hypothesis between models of retrieval (Lewis & Vasishth, 2005) and the experimental measures is fairly straightforward: visual attention is typically taken to be a probabilistic function of attention (Allopena, Magnuson, & Tanenhaus, 1998). Thus, in VWP, we can directly track the degree to which different DPs are activated and considered during the processing of a reflexive pronoun.
This summer (August to September 2023), I started running a visual word experiment to investigate interference effects, where the distractor is inside a relative clause that is modifying the antecedent. I have collected data from 80 participants, out of 120, and will be finishing data collection this December. Next year (around July to August 2024), I will be running another visual world experiment to test whether the thematic role of the distractor modulates the degree of interference. If comprehenders resolve reflexive interpretation by making reference to thematic prominence à la Andrews (1985), then we expect to see more pronounced interference effects from distractors that are more prominent on the hierarchy (agents) than those that are less prominent (patient).
This is work in progress and is in collaboration with Alessa Farinella (UMass graduate student), and Patricia Asuncion and John Michael De Pano (UP Diliman grad students).
In Tagalog, adjective-noun word order is variable (e.g., magandang lalaki vs. lalaking maganda 'beautiful man'). Both word orders are interchangeable and have no difference in meaning (Schachter & Otanes, 1972; SO72). Shih and Zuraw (2017; SZ17) presented a large-scale corpus analysis of this variation and found that phonological markedness can condition word order. For example, their corpus shows that a sequence of adjacent nasals is disfavored (e.g., itim na peluka 'black wig'); the alternative word order that avoids a sequence of adjacent nasals is favored (e.g., pelukang itim). SZ17 argued that nasal-OCP, a typologically robust phonological markedness constraint that disprefers consecutive nasals, influenced the order.
The extent to which the alternative order is favored may be confounded by the general preference for encliticized form of the linker (-ng) over the free form na, as noted by SO72 and SZ17. We argue that a stronger case could be made if the form of the linker is kept constant: both alternatives use na (e.g., itim na itik vs itik na itim 'black duck') or both use -ng (e.g., gwapong mamamayan vs mamamayang gwapo 'handsome citizen'). Having this precise level of control on the alternatives may be difficult to achieve in corpus studies, however. For this reason, we approach the question of whether phonological well-formedness can condition word order using an experimental approach.
This is in collaboration with Rowena Garcia. This has been accepted for publication with revisions and is slated to appear in Volume 10 of the Annual Review of Linguistics.
Over the last decade, there has been a slow-but-steady accumulation of psycholinguistic work focusing on typologically diverse languages. In this review, we provide an overview of the psycholinguistic research on Philippine languages at the sentence-level. We first discuss the grammatical features of these languages that figure prominently in existing research. We identify four linguistic domains that have received attention from language researchers and give a summary of the empirical terrain. We advance two claims that emerge across these different domains: (i) the agent-first pressure plays a central role in many of the findings; and (ii) the generalization that the patient argument is the syntactically privileged argument cannot be reduced to frequency, but instead is an emergent phenomenon caused by the alignment of competing pressures toward an optimal candidate. We connect these language-specific claims to language-general theories of sentence processing.
This is joint work with Matt Wagers. I gave a talk at CUNY 2020 about this project and here is a link to the video. It starts around 3:05:40.
Subject-gap relative clauses (SRC; the dog that chased the piglet) are said to be easier to process than object-gap relative clauses (ORC; the dog that the piglet chased). This asymmetry extends across languages and methods. Different classes of proposals emphasize distinct aspects of the dependency. Some have emphasized the role of syntactic structure. Others have emphasized the role of working memory, and others, the role of linguistic experience.
In this project, we leverage various features of Tagalog morphosyntax to evaluate the empirical coverage of these classes of proposals. We also examine the types of linguistic information that Tagalog comprehenders attend to that can modulate this asymmetry, like the relative order of the head noun and the RC, and the referential type of the intervening co-argument.
As participants heard a sentence containing a relative clause, their eye movements were recorded. After hearing the sentence, they had to choose the picture that best corresponded to what they had heard, and to rate how confident they were with their response. Analyses of the participants’ eye movement data are still in progress, but their choice data and confidence ratings indicate an asymmetry which can then be neutralized by word order and the pronominality of the co-argument. The asymmetry always persisted in their RTs, however.
The manuscript is currently in preparation.
This is joint work with Matt Wagers. This will be submitted for publication this week.
Voice morphology has been central to many investigations in Austronesian syntax because it interacts with case marking, voice morphology, and A-bar movement. this project revisits the interaction between voice and A-bar movement in Tagalog. The received view is that the extraction restriction is clean-cut: Only the argument cross-referenced by voice is eligible for A-bar movement. That is, when the verb exhibits agent voice (AV), only the agent is extractable. When the verb exhibits patient voice (PV), only the patient is extractable. When the verb exhibits the various flavors of applicative, only the applied argument is extractable.
I conducted a series of acceptability judgment studies and found that speakers indeed gave high ratings when the extracted argument was cross-referenced by voice. These findings are consistent with the received view. However, some speakers also gave surprisingly high ratings when the extracted argument was an agent even when not cross-referenced by voice. I supplemented my judgment studies with computational evidence that modeled the participants’ judgment process using the framework established by Dillon et al (2017). Participants were treating agent-extractions not cross-referenced by voice categorically. These results, taken together with previous descriptions of a potential variation in extraction (Ceña & Nolasco, 2011, 2012), suggest that there may be subtle but categorical distinctions in individual speakers’ grammars, in the vein of Han, Lidz, and Musolino (2007). In other words, there may be at least two grammars of extraction operative in Tagalog: One conservative, one permissive. When the extracted argument is cross-referenced by voice, the linguistic output of these grammars converge. their output diverge only when we look at the speakers’ behavior when the extracted argument is not cross-referenced by voice. I propose that the grammaticalization of a functional pressure in sentence processing, agent-first, is a source of this variation.
This is joint work with Matt Wagers. This has been published in Frontiers in Psychology and here's a link to the article.
When processing filler-gap dependencies (FGDs), comprehenders do not wait until they encounter all of the bottom-up information in the input. Instead, they use various types of linguistic information to predictively posit a gap that would allow the dependency to be resolved. They can use syntactic, lexical, morphological, and prosodic information. In this project, we examine whether Tagalog comprehenders use the language’s voice morphology to guide their incremental interpretations. We hypothesized that voice allows comprehenders to commit to an interpretation upon encountering the verb, since they have information about the event structure at this point in time and by virtue of voice, the thematic role of the filler.
In experiment 1, using an acceptability judgment study, we found that comprehenders differed in how they used the different voices in different FGD-contexts to detect the licitness of extractions. These differences may have consequences for how voice is used in real-time. In experiments 2 and 3, using the stops-making-sense paradigm, comprehenders used voice as a cue to actively associate the filler with the gap. However, in experiment 3, the way in which they used voice varied by type and varied across types of FGDs. We argue that comprehenders were using construction-specific cue validities when processing FGDs. However, they also engaged with other classes of linguistic information, including (but not limited to) information about the structural similarities and the thematic complexity of the dependencies involved, and the relative frequency of the different types of voices in the language. These interactions resulted in processing asymmetries.
How are morphologically complex words represented mentally in the lexicon? Different models of lexical processing make different claims about the nature of these representations. There are three main classes of models that have been proposed. On one end, full-listing models argue for holistic representations. Under this view, holistic representations are used for recognition, and the morphological relationship between words is an epiphenomenon. Constituent morphemes may become activated and may influence post-lexical processing, but morphological information is not crucial at the level of recognition. On the other end, decompositional models maintain that complex words are automatically decomposed into constituent morphemes, a process called morphological decomposition. Under this view, morphological information is leveraged by the parser, and complex words are recognized via the root morpheme. In the middle, hybrid models allow for both holistic and decompositional processing.
I investigated how the lexical representation of related forms in Dabaw Bisaya, an Austronesian language spoken in Davao City, Philippines, is affected by nasal substitution, a morpho-phonological process that obscures the onset of the root. I conducted a lexical decision experiment with auditory masked priming to investigate this, in lieu of the more commonly used visual masked priming technique, since the language has no standardized orthography.
I found no evidence for priming when participants were presented with a morphologically related prime that has undergone nasal substitution. Participants had faster decision times, however, when presented with the same prime or a morphologically related prime whose root remained transparent. These results suggest that models of lexical processing that leverage morphological information during word recognition have better empirical coverage, with the proviso that surface opacity may modulate morphological decomposition. I presented these findings at the Morphological Typology and Linguistic Cognition workshop in Lexington, KY.
This is joint work with Steven Foley, Kelsey Sasaki, Maziar Toosarvandani, and Matt Wagers.
Language comprehenders make predictions about what they will hear next, using parsing principles to resolve ambiguities. One principle, Parse Subject, has been attested in many experiments; however, these have predominantly tested a small handful of subject-initial languages, undersampling the world’s linguistic diversity. We tested the universality of Parse Subject in Santiago Laxopa Zapotec, a verb-initial language, through a series of eye-tracking studies. Speakers inspected illustrations while they listened to recordings and we tracked their eye-movements. The distribution of visual attention over time showed that Parse Subject interpretations were indeed the earliest mid-sentence interpretations entertained by participants. Such interpretations were routinely defeated by subsequent information, indicating that it is a defeasible principle. These findings support the universality of Parse Subject by incorporating data from an under-investigated language whose speakers have not historically contributed to our understanding of the language processing architecture.
Our latest findings was featured by Matt Wagers in his keynote address at CUNY 2020. An earlier version of the study was presented by Steven at the 2019 annual meeting of the LSA.
Teaching dossier available upon request N.B.: At UCSC, unless indicated otherwise
I go by Jed. If you have to use my full name, note that my name and the name of one of the eels in Little Mermaid (Jetsam) don't form a minimal pair. None of my name's vowels get reduced.
I was born in Davao City —the durian capital of the Philippines—but grew up in Cerritos, CA—home of Cerritos Auto Square, the largest auto mall in the world. lol. I speak Dabaw Bisaya, a variety of Cebuano, natively. I also speak Tagalog (or Filipino, depending on your language ideologies) since it was part of the national curriculum when I went to school there. Aside: I don't particularly care if you call it Tagalog or Filipino, but I will fight you if you say they're different languages. Filipino is simply the standardized form of Tagalog.
When I am not doing linguistics, i am most likely cooking with wine, while adding it to the food as well. My favorite ingredient to work with is coconut cream. I use it when I make desserts (like cassava cake, or sticky rice wrapped in banana leaf), kinilaw (a.k.a Filipino ceviche), adobo, etc. I love to cook and feed people because sharing the food I make means sharing my culture. If I'm not busy stuffing myself or cooking to stuff myself and others, I'm probably playing board games and video games, or binging on TV shows and anime.