Memory access mechanisms such as cue-based retrieval have come to dominate theories of the processing of linguistic dependencies such as subject-verb agreement. One phenomenon that has been regarded as demonstrating the role of such mechanisms is the grammaticality asymmetry in agreement attraction, which is the observation that nouns other than the grammatical controller of agreement can influence the computation of subject-verb agreement in ungrammatical, but not grammatical, sentences. This asymmetry is most often accounted for via the dynamics of retrieval interference. We challenge this interpretation, arguing that the asymmetry largely reflects response bias. Three forced-choice judgment experiments show that neutralizing response bias results in a decrease in the size of the grammaticality asymmetry, or its elimination altogether. Together with the response time patterns in these experiments, this result favors an account that attributes attraction effects to a continuous and equivocal representation of number, rather than to the dynamics of retrieval interference. We implement a model of grammaticality judgments that links a continuous representation of number to the rate of evidence accumulation in a diffusion process. This model accounts for the presence or absence of the grammaticality asymmetry through shifts in the decisional starting point (i.e. response bias), and highlights the importance of monitoring for response bias effects in judgment tasks.

Prediction has been proposed as an overarching principle that explains human information processing in language and beyond. To what degree can processing difficulty in syntactically complex sentences - one of the major concerns of psycholinguistics - be explained by predictability, as estimated using computational language models? A precise, quantitative test of this question requires a much larger scale data collection effort than has been done in the past. We present the Syntactic Ambiguity Processing Benchmark, a dataset of self-paced reading times from 2000 participants, who read a diverse set of complex English sentences. This dataset makes it possible to measure processing difficulty associated with individual syntactic constructions, and even individual sentences, precisely enough to rigorously test the predictions of computational models of language comprehension. We find that the predictions of language models with two different architectures sharply diverge from the reading time data, dramatically underpredicting processing difficulty, failing to predict relative difficulty among different syntactic ambiguous constructions, and only partially explaining item-wise variability. These findings suggest that prediction is most likely insufficient on its own to explain human syntactic processing.

No abstract

No abstract

Salinas de los Nueve Cerros is a precolumbian Maya city located at the base of the highlands in the lowlands of west-central Guatemala. It is the only Classic-period center in the southern Maya Lowlands that based its economy on the production of an important raw material for export: salt. Because of its economic role and its location along a major trade route, Salinas de los Nueve Cerros had a particularly long history of occupation. The site has evidence of a large sedentary population starting during the Middle Preclassic (by ca. 800 B.C.) that continued several hundred years beyond the Classic collapse, before finally being abandoned ca. A.D. 1200. The salt source was located in the center of the city. This presents a rare opportunity to test the degree of elite control over the production of a non-elite resource. Production during the Classic period does appear to have been tightly controlled by elites, as evinced by the presence of multiple administrative structures and elite tombs throughout the salt-working zone, located less than 100 mfrom the site epicenter.