Number feature distortion modulates cue-based retrieval in reading

In sentence comprehension, what are the cognitive constraints that determine number agreement computation? Two broad classes of theoretical proposals are: (i) Representation distortion accounts, which assume feature on subject noun gets overwritten probabilistically by a non-subject noun, leading to non-veridical memory trace noun; and (ii) The cue-based retrieval account, general account dependency completion processes assumes features remain intact, processing difficulty is only function completion. However, both these model fail for full spectrum patterns observed in published studies. Using 17 benchmark datasets from four languages, we implement seven computational models: three variants representation distortion, two models, hybrid models representation-distortion retrieval. Quantitative comparison shows best fit achieved (specifically, percolation) retrieval; numerically, second-best quantitative was distortion-based attraction grammaticality bias during reading. More broadly, work furnishes comprehensive evidence support idea theory, aims be completion, needs incorporate process.