@phdthesis{LissonHernandez2022,
  author    = {Liss{\´o}n Hern{\´a}ndez, Paula J.},
  title     = {Computational models of sentence comprehension in aphasia},
  doi       = {10.25932/publishup-55548},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-555487},
  school      = {Universit{\"a}t Potsdam},
  pages     = {vi, 159},
  year      = {2022},
  abstract  = {It is well-known that individuals with aphasia (IWA) have difficulties understanding sentences that involve non-adjacent dependencies, such as object relative clauses or passives (Caplan, Baker, \& Dehaut, 1985; Caramazza \& Zurif, 1976). A large body of research supports the view that IWA's grammatical system is intact, and that comprehension difficulties in aphasia are caused by a processing deficit, such as a delay in lexical access and/or in syntactic structure building (e.g., Burkhardt, Pi{\~n}ango, \& Wong, 2003; Caplan, Michaud, \& Hufford, 2015; Caplan, Waters, DeDe, Michaud, \& Reddy, 2007; Ferrill, Love, Walenski, \& Shapiro, 2012; Hanne, Burchert, De Bleser, \& Vasishth, 2015; Love, Swinney, Walenski, \& Zurif, 2008). The main goal of this dissertation is to computationally investigate the processing sources of comprehension impairments in sentence processing in aphasia. In this work, prominent theories of processing deficits coming from the aphasia literature are implemented within two cognitive models of sentence processing -the activation-based model (Lewis \& Vasishth, 2005) and the direct-access model (McEl- ree, 2000)-. These models are two different expressions of the cue-based retrieval theory (Lewis, Vasishth, \& Van Dyke, 2006), which posits that sentence processing is the result of a series of iterative retrievals from memory. These two models have been widely used to account for sentence processing in unimpaired populations in multiple languages and linguistic constructions, sometimes interchangeably (Parker, Shvarts- man, \& Van Dyke, 2017). However, Nicenboim and Vasishth (2018) showed that when both models are implemented in the same framework and fitted to the same data, the models yield different results, because the models assume different data- generating processes. Specifically, the models hold different assumptions regarding the retrieval latencies. The second goal of this dissertation is to compare these two models of cue-based retrieval, using data from individuals with aphasia and control participants. We seek to answer the following question: Which retrieval mechanism is more likely to mediate sentence comprehension? We model 4 subsets of existing data: Relative clauses in English and German; and control structures and pronoun resolution in German. The online data come from either self-paced listening experiments, or visual-world eye-tracking experiments. The offline data come from a complementary sentence-picture matching task performed at the end of the trial in both types of experiments. The two competing models of retrieval are implemented in the Bayesian framework, following Nicenboim and Vasishth (2018). In addition, we present a modified version of the direct-acess model that - we argue - is more suitable for individuals with aphasia. This dissertation presents a systematic approach to implement and test verbally- stated theories of comprehension deficits in aphasia within cognitive models of sen- tence processing. The conclusions drawn from this work are that (a) the original direct-access model (as implemented here) cannot account for the full pattern of data from individuals with aphasia because it cannot account for slow misinterpretations; and (b) an activation-based model of retrieval can account for sentence comprehension deficits in individuals with aphasia by assuming a delay in syntactic structure building, and noise in the processing system. The overall pattern of results support an activation-based mechanism of memory retrieval, in which a combination of processing deficits, namely slow syntax and intermittent deficiencies, cause comprehension difficulties in individuals with aphasia.},
  language  = {en}
}
@phdthesis{Laurinavichyute2021,
  author    = {Laurinavichyute, Anna},
  title     = {Similarity-based interference and faulty encoding accounts of sentence processing},
  doi       = {10.25932/publishup-50966},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-509669},
  school      = {Universit{\"a}t Potsdam},
  pages     = {237},
  year      = {2021},
  abstract  = {The goal of this dissertation is to empirically evaluate the predictions of two classes of models applied to language processing: the similarity-based interference models (Lewis \& Vasishth, 2005; McElree, 2000) and the group of smaller-scale accounts that we will refer to as faulty encoding accounts (Eberhard, Cutting, \& Bock, 2005; Bock \& Eberhard, 1993). Both types of accounts make predictions with regard to processing the same class of structures: sentences containing a non-subject (interfering) noun in addition to a subject noun and a verb. Both accounts make the same predictions for processing ungrammatical sentences with a number-mismatching interfering noun, and this prediction finds consistent support in the data. However, the similarity-based interference accounts predict similar effects not only for morphosyntactic, but also for the semantic level of language organization. We verified this prediction in three single-trial online experiments, where we found consistent support for the predictions of the similarity-based interference account. In addition, we report computational simulations further supporting the similarity-based interference accounts. The combined evidence suggests that the faulty encoding accounts are not required to explain comprehension of ill-formed sentences. For the processing of grammatical sentences, the accounts make conflicting predictions, and neither the slowdown predicted by the similarity-based interference account, nor the complementary slowdown predicted by the faulty encoding accounts were systematically observed. The majority of studies found no difference between the compared configurations. We tested one possible explanation for the lack of predicted difference, namely, that both slowdowns are present simultaneously and thus conceal each other. We decreased the amount of similarity-based interference: if the effects were concealing each other, decreasing one of them should allow the other to surface. Surprisingly, throughout three larger-sample single-trial online experiments, we consistently found the slowdown predicted by the faulty encoding accounts, but no effects consistent with the presence of inhibitory interference. The overall pattern of the results observed across all the experiments reported in this dissertation is consistent with previous findings: predictions of the interference accounts for the processing of ungrammatical sentences receive consistent support, but the predictions for the processing of grammatical sentences are not always met. Recent proposals by Nicenboim et al. (2016) and Mertzen et al. (2020) suggest that interference might arise only in people with high working memory capacity or under deep processing mode. Following these proposals, we tested whether interference effects might depend on the depth of processing: we manipulated the complexity of the training materials preceding the grammatical experimental sentences while making no changes to the experimental materials themselves. We found that the slowdown predicted by the faulty encoding accounts disappears in the deep processing mode, but the effects consistent with the predictions of the similarity-based interference account do not arise. Independently of whether similarity-based interference arises under deep processing mode or not, our results suggest that the faulty encoding accounts cannot be dismissed since they make unique predictions with regard to processing grammatical sentences, which are supported by data. At the same time, the support is not unequivocal: the slowdowns are present only in the superficial processing mode, which is not predicted by the faulty encoding accounts. Our results might therefore favor a much simpler system that superficially tracks number features and is distracted by every plural feature.},
  language  = {en}
}
@phdthesis{Engelmann2016,
  author    = {Engelmann, Felix},
  title     = {Toward an integrated model of sentence processing in reading},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-100864},
  school      = {Universit{\"a}t Potsdam},
  pages     = {xiii, 143},
  year      = {2016},
  abstract  = {In experiments investigating sentence processing, eye movement measures such as fixation durations and regression proportions while reading are commonly used to draw conclusions about processing difficulties. However, these measures are the result of an interaction of multiple cognitive levels and processing strategies and thus are only indirect indicators of processing difficulty. In order to properly interpret an eye movement response, one has to understand the underlying principles of adaptive processing such as trade-off mechanisms between reading speed and depth of comprehension that interact with task demands and individual differences. Therefore, it is necessary to establish explicit models of the respective mechanisms as well as their causal relationship with observable behavior. There are models of lexical processing and eye movement control on the one side and models on sentence parsing and memory processes on the other. However, no model so far combines both sides with explicitly defined linking assumptions. In this thesis, a model is developed that integrates oculomotor control with a parsing mechanism and a theory of cue-based memory retrieval. On the basis of previous empirical findings and independently motivated principles, adaptive, resource-preserving mechanisms of underspecification are proposed both on the level of memory access and on the level of syntactic parsing. The thesis first investigates the model of cue-based retrieval in sentence comprehension of Lewis \& Vasishth (2005) with a comprehensive literature review and computational modeling of retrieval interference in dependency processing. The results reveal a great variability in the data that is not explained by the theory. Therefore, two principles, 'distractor prominence' and 'cue confusion', are proposed as an extension to the theory, thus providing a more adequate description of systematic variance in empirical results as a consequence of experimental design, linguistic environment, and individual differences. In the remainder of the thesis, four interfaces between parsing and eye movement control are defined: Time Out, Reanalysis, Underspecification, and Subvocalization. By comparing computationally derived predictions with experimental results from the literature, it is investigated to what extent these four interfaces constitute an appropriate elementary set of assumptions for explaining specific eye movement patterns during sentence processing. Through simulations, it is shown how this system of in itself simple assumptions results in predictions of complex, adaptive behavior. In conclusion, it is argued that, on all levels, the sentence comprehension mechanism seeks a balance between necessary processing effort and reading speed on the basis of experience, task demands, and resource limitations. Theories of linguistic processing therefore need to be explicitly defined and implemented, in particular with respect to linking assumptions between observable behavior and underlying cognitive processes. The comprehensive model developed here integrates multiple levels of sentence processing that hitherto have only been studied in isolation. The model is made publicly available as an expandable framework for future studies of the interactions between parsing, memory access, and eye movement control.},
  language  = {en}
}
@phdthesis{Jaeger2015,
  author    = {J{\"a}ger, Lena Ann},
  title     = {Working memory and prediction in human sentence parsing},
  url       = {http://nbn-resolving.de/urn:nbn:de:kobv:517-opus4-82517},
  school      = {Universit{\"a}t Potsdam},
  pages     = {xi, 144},
  year      = {2015},
  abstract  = {This dissertation investigates the working memory mechanism subserving human sentence processing and its relative contribution to processing difficulty as compared to syntactic prediction. Within the last decades, evidence for a content-addressable memory system underlying human cognition in general has accumulated (e.g., Anderson et al., 2004). In sentence processing research, it has been proposed that this general content-addressable architecture is also used for language processing (e.g., McElree, 2000). Although there is a growing body of evidence from various kinds of linguistic dependencies that is consistent with a general content-addressable memory subserving sentence processing (e.g., McElree et al., 2003; VanDyke2006), the case of reflexive-antecedent dependencies has challenged this view. It has been proposed that in the processing of reflexive-antecedent dependencies, a syntactic-structure based memory access is used rather than cue-based retrieval within a content-addressable framework (e.g., Sturt, 2003). Two eye-tracking experiments on Chinese reflexives were designed to tease apart accounts assuming a syntactic-structure based memory access mechanism from cue-based retrieval (implemented in ACT-R as proposed by Lewis and Vasishth (2005). In both experiments, interference effects were observed from noun phrases which syntactically do not qualify as the reflexive's antecedent but match the animacy requirement the reflexive imposes on its antecedent. These results are interpreted as evidence against a purely syntactic-structure based memory access. However, the exact pattern of effects observed in the data is only partially compatible with the Lewis and Vasishth cue-based parsing model. Therefore, an extension of the Lewis and Vasishth model is proposed. Two principles are added to the original model, namely 'cue confusion' and 'distractor prominence'. Although interference effects are generally interpreted in favor of a content-addressable memory architecture, an alternative explanation for interference effects in reflexive processing has been proposed which, crucially, might reconcile interference effects with a structure-based account. It has been argued that interference effects do not necessarily reflect cue-based retrieval interference in a content-addressable memory but might equally well be accounted for by interference effects which have already occurred at the moment of encoding the antecedent in memory (Dillon, 2011). Three experiments (eye-tracking and self-paced reading) on German reflexives and Swedish possessives were designed to tease apart cue-based retrieval interference from encoding interference. The results of all three experiments suggest that there is no evidence that encoding interference affects the retrieval of a reflexive's antecedent. Taken together, these findings suggest that the processing of reflexives can be explained with the same cue-based retrieval mechanism that has been invoked to explain syntactic dependency resolution in a range of other structures. This supports the view that the language processing system is located within a general cognitive architecture, with a general-purpose content-addressable working memory system operating on linguistic expressions. Finally, two experiments (self-paced reading and eye-tracking) using Chinese relative clauses were conducted to determine the relative contribution to sentence processing difficulty of working-memory processes as compared to syntactic prediction during incremental parsing. Chinese has the cross-linguistically rare property of being a language with subject-verb-object word order and pre-nominal relative clauses. This property leads to opposing predictions of expectation-based accounts and memory-based accounts with respect to the relative processing difficulty of subject vs. object relatives. Previous studies showed contradictory results, which has been attributed to different kinds local ambiguities confounding the materials (Lin and Bever, 2011). The two experiments presented are the first to compare Chinese relatives clauses in syntactically unambiguous contexts. The results of both experiments were consistent with the predictions of the expectation-based account of sentence processing but not with the memory-based account. From these findings, I conclude that any theory of human sentence processing needs to take into account the power of predictive processes unfolding in the human mind.},
  language  = {en}
}