Working Memory Constraints On Syntactic Ambiguity Resolution: The Cognitive Mechanism

How working memory (WM) constrains syntactic ambiguity resolution is one of the core and unresolved questions within the studies of language comprehension. Two hypotheses have emerged. The activation account claims high span readers can maintain multiple representations after their initial constructions, while low span readers maintain only one representation. In contrast, the inhibition account claims high-span readers can inhibit inappropriate or unprefered representation efficiently, while low-span readers have to maintain multiple representations. The present study further explores how WM capacity constrains syntactic ambiguity resolution (Experiment 1) and how it modifies the representation states after a long time (Experiment 2).In Experiment 1, participants were asked to perform a self-paced moving window reading to each sentence in which the sentence-initial noun was ambiguous between object reading and subject reading. In this experiment, the ambiguity was always resolved as the subject reading. The selectional restrictions of a verb for the sentence-initial noun were manipulated to examine the object reading participants pursued. The results showed that low span readers perform the object reading, while high span readers do not, suggesting greater WM capacity permits more efficient inhibition of an inappropriate interpretation.In Experiment 2, we use a comprehension-to-comprehension syntactic priming paradigm to examine how WM capacity modifies the representation states after a relatively long time. Participants read sentences starting with an ambiguous phrase consisting of Verb+N1+de+N2, word-by-word. The phrase could be interpreted either as a modifier-noun structure (MNS) or a narrative-object structure (NOS). The ambiguity is resolved as MNS or NOS in the prime and the target sentences, respectively. The results showed that high span readers show a negative priming effect when they read target sentences following a prime sentence starting with a balanced ambiguous phrase. This finding indicates that inappropriate NOS is inhibited during syntactic ambiguity resolution for high span readers, and that the inhibitory processing state can be prolonged for a while. In contrast, low span readers did not show any priming effect, indicating either they can not inhibit the inappropriate NOS efficiently, or they can not maintain the inhibitory state. We conclude that although the inhibition account can explain the individual differences in sentence processing between high span and low span readers in some cases, the activation account can also explain the individual differences in other cases.