When High-Capacity Readers Slow Down and Low-Capacity Readers Speed Up: Working Memory and Locality Effects

Abstract

We examined the effects of argument-head distance in SVO and SOV languages (Spanish and German), while taking into account readers' working memory capacity and controlling for expectation (Levy, 2008) and other factors. We predicted only locality effects, that is, a slowdown produced by increased dependency distance (Gibson, 2000; Lewis and Vasishth, 2005). Furthermore, we expected stronger locality effects for readers with low working memory capacity. Contrary to our predictions, low-capacity readers showed faster reading with increased distance, while high-capacity readers showed locality effects. We suggest that while the locality effects are compatible with memory-based explanations, the speedup of low-capacity readers can be explained by an increased probability of retrieval failure. We present a computational model based on ACT-R built under the previous assumptions, which is able to give a qualitative account for the present data and can be tested in future research. Our results suggest that in some cases, interpreting longer RTs as indexing increased processing difficulty and shorter RTs as facilitation may be too simplistic: The same increase in processing difficulty may lead to slowdowns in high-capacity readers and speedups in low-capacity ones. Ignoring individual level capacity differences when investigating locality effects may lead to misleading conclusions.

Dependency Distance Differences across Interpreting Types: Implications for Cognitive Demand

“…We further considered the probable effect of individual interpreter's style. There is a collection of evidence supporting the influence of individual differences in working memory capacity on dependency resolution processes, especially in long-distance dependency resolution (Nicenboim et al, 2015, 2016). Hence, ignoring individual differences may confound the final results.…”

Exploratory and Confirmatory Analyses in Sentence Processing: A Case Study of Number Interference in German

“…However, it may be the case that the inhibitory interference effect is also present for the number feature in grammatical sentences, but it is harder to detect. This could be because either (a) the weight of the number feature as a cue for retrieval might be smaller than other semantic or syntactic cues (for details about cue weights in cue‐based retrieval, see Nicenboim et al., ), (b) facilitatory interference has a different cause than inhibitory interference (for details, see Engelmann, ; Jäger et al., ), or (c) there are other mechanisms at play that may counteract the effect of interference (e.g., feature percolation/movement: Bock & Miller, ; Pearlmutter, Garnsey, & Bock, ; and see Patson & Husband, , for evidence of the faulty number representation of nouns in number interference configurations).…”

“…In this paper, we fit shifted lognormal mixed models, which lie outside the class of generalized linear models. For reading time data, such models can provide more accurate estimates than standard linear mixed models (see Nicenboim, Logacev, Gattei, & Vasishth, ; Rouder, ; Rouder, Tuerlinckx, Speckman, Lu, & Gomez, ). The justification for assuming a shifted lognormal distribution for the reading times instead of a normal distribution (as in linear mixed models) is that residual reading times in self‐paced reading are highly right skewed and have a lower bound greater than zero (i.e., the shift of the distribution).…”

“…While sometimes reading times are log‐ or reciprocal transformed to avoid violations of normality assumptions (Box & Cox, ), these transformations still assume that the reading times are defined by their location (mean) and scale (standard deviation), and they are unshifted. Unshifted distributions for reading times in self‐paced reading are unreasonable, since they do not take into account the fact that there is a minimal amount of time that it takes to read a word and press a button on the keyboard, typically around 150–250 ms and that this may vary across participants (see Nicenboim et al., , for self‐paced reading; and Logan, ; Rouder, ; Rouder et al., , for reaction time tasks). If distributions are shifted and analyzed as unshifted, estimates will be affected, and they may influence conclusions (Rouder, ), especially when dealing with small effect sizes, as it seems to be the case for the number interference effects we will investigate here (Jäger, Engelmann, & Vasishth, ).…”

The Effect of Prominence and Cue Association on Retrieval Processes: A Computational Account

“…This model is based on the general cognitive architecture ACT-R ("Adaptive Control of Thought-Rational," Anderson et al, 2004;Anderson & Lebiere, 1998). Over the last decade, the LV05 model has been widely used as a computational modeling framework by several research groups for investigating a range of empirical phenomena: (a) similaritybased interference effects (Dillon et al, 2013;J€ ager et al, 2015;Kush & Phillips, 2014;Nicenboim, Logacev, Gattei, & Vasishth, 2016;Nicenboim, Vasishth, Engelmann, & Suckow, 2018;Parker & Phillips, 2016Patil, Vasishth, & Lewis, 2016;Vasishth, Bruessow, Lewis, & Drenhaus, 2008); (b) the relative roles of predictive processing and memory effects (Boston, Hale, Vasishth, & Kliegl, 2011); (c) impairments in individuals with aphasia (M€ atzig, Vasishth, Engelmann, Caplan, & Burchert, 2018;Patil, Hanne, Burchert, De Bleser, & Vasishth, 2016); (d) the interaction between oculomotor control and sentence comprehension (Dotlacil, 2018;Engelmann, Vasishth, Engbert, & Kliegl, 2013); and (e) the effect of working memory capacity differences on underspecification ("good-enough" processing) in sentence comprehension (Engelmann, 2016). The model relies on the core assumptions of ACT-R that retrieving an item from memory is affected by activation decay and similarity-based interference.…”