Effects of Surprisal and Locality on Danish Sentence Processing: An Eye-Tracking Investigation

An eye-tracking experiment in Danish investigates two dominant accounts of sentence processing: locality-based theories that predict a processing advantage for sentences where the distance between the major syntactic heads is minimized, and the surprisal theory which predicts that processing time increases with big changes in the relative entropy of possible parses, sometimes leading to anti-locality effects. We consider both lexicalised surprisal, expressed in conditional trigram probabilities, and syntactic surprisal expressed in the manipulation of the expectedness of the second NP in Danish constructions with two postverbal NP-objects. An eye-tracking experiment showed a clear advantage for local syntactic relations, with only a marginal effect of lexicalised surprisal and no effect of syntactic surprisal. We conclude that surprisal has a relatively marginal effect, which may be clearest for verbs in verb-final languages, while locality is a robust predictor of sentence processing.     

Uncertainty and Expectation in Sentence Processing: Evidence From Subcategorization Distributions

There is now considerable evidence that human sentence processing is expectation based: As people read a sentence, they use their statistical experience with their language to generate predictions about upcoming syntactic structure. This study examines how sentence processing is affected by readers' uncertainty about those expectations. In a self‐paced reading study, we use lexical subcategorization distributions to factorially manipulate both the strength of expectations and the uncertainty about them. We compare two types of uncertainty: uncertainty about the verb's complement, reflecting the next prediction step; and uncertainty about the full sentence, reflecting an unbounded number of prediction steps. We find that uncertainty about the full structure, but not about the next step, was a significant predictor of processing difficulty: Greater reduction in uncertainty was correlated with increased reading times (RTs). We additionally replicated previously observed effects of expectation violation (surprisal), orthogonal to the effect of uncertainty. This suggests that both surprisal and uncertainty affect human RTs. We discuss the consequences for theories of sentence comprehension.     

句子理解中抑制不同性质干扰信息的眼动

该研究采用眼动追踪技术和分心阅读范式考察了句子理解中的抑制过程,以及外在干扰信息的性质如何影响阅读中的抑制效率。研究结果发现语义相关干扰词比语义无关干扰词的阅读时间、注视时间更长,注视次数更多。语义无关干扰词比无意义字符串的注视时间更长、注视次数更多。该结果说明阅读过程中干扰信息越有意义,与目标句语义联系越紧密,则越容易被激活,抑制效率也就越低,符合Hasher等人提出的一般认知过程中抑制加工的原理。     

The time-course of lexical and structural processes in sentence comprehension

Online sentence comprehension involves multiple types of cognitive processes: lexical processes such as lexical access, which call on the user's knowledge of the meaning of words in the language, and structural processes such as the integration of incoming words into an emerging representation. In this article, we investigate the temporal dynamics of lexical access and syntactic integration. Unlike much previous work that has relied on temporary ambiguity to investigate this question, we manipulate the frequency of the verb in unambiguous structures involving a well-studied syntactic complexity manipulation (subject- vs. object-extracted clefts). The results demonstrate that for high-frequency verbs, the difficulty of reading a more structurally complex object-extracted cleft structure relative to a less structurally complex subject-extracted cleft structure is largely experienced in the cleft region, whereas for low-frequency verbs this difficulty is largely experienced in the postcleft region. We interpret these results as evidence that some stages of structural processing follow lexical processing. Furthermore, we find evidence that structural processing may be delayed if lexical processing is costly, and that the delay is proportional to the difficulty of the lexical process. Implications of these results for contemporary accounts of sentence comprehension are discussed.     

Probability and surprisal in auditory comprehension of morphologically complex words

Two auditory lexical decision experiments document for morphologically complex words two points at which the probability of a target word given the evidence shifts dramatically. The first point is reached when morphologically unrelated competitors are no longer compatible with the evidence. Adapting terminology from Marslen-Wilson (1984), we refer to this as the word's initial uniqueness point (UP1). The second point is the complex uniqueness point (CUP) introduced by Balling and Baayen (2008), at which morphologically related competitors become incompatible with the input. Later initial as well as complex uniqueness points predict longer response latencies. We argue that the effects of these uniqueness points arise due to the large surprisal (Levy, 2008) carried by the phonemes at these uniqueness points, and provide independent evidence that how cumulative surprisal builds up in the course of the word co-determines response latencies. The presence of effects of surprisal, both at the initial uniqueness point of complex words, and cumulatively throughout the word, challenges the Shortlist B model of Norris and McQueen (2008), and suggests that a Bayesian approach to auditory comprehension requires complementation from information theory in order to do justice to the cognitive cost of updating probability distributions over lexical candidates.     

Single-Stage Prediction Models Do Not Explain the Magnitude of Syntactic Disambiguation Difficulty

The disambiguation of a syntactically ambiguous sentence in favor of a less preferred parse can lead to slower reading at the disambiguation point. This phenomenon, referred to as a garden-path effect, has motivated models in which readers initially maintain only a subset of the possible parses of the sentence, and subsequently require time-consuming reanalysis to reconstruct a discarded parse. A more recent proposal argues that the garden-path effect can be reduced to surprisal arising in a fully parallel parser: words consistent with the initially dispreferred but ultimately correct parse are simply less predictable than those consistent with the incorrect parse. Since predictability has pervasive effects in reading far beyond garden-path sentences, this account, which dispenses with reanalysis mechanisms, is more parsimonious. Crucially, it predicts a linear effect of surprisal: the garden-path effect is expected to be proportional to the difference in word surprisal between the ultimately correct and ultimately incorrect interpretations. To test this prediction, we used recurrent neural network language models to estimate word-by-word surprisal for three temporarily ambiguous constructions. We then estimated the slowdown attributed to each bit of surprisal from human self-paced reading times, and used that quantity to predict syntactic disambiguation difficulty. Surprisal successfully predicted the existence of garden-path effects, but drastically underpredicted their magnitude, and failed to predict their relative severity across constructions. We conclude that a full explanation of syntactic disambiguation difficulty may require recovery mechanisms beyond predictability.     

Lossy-Context Surprisal: An Information-Theoretic Model of Memory Effects in Sentence Processing

A key component of research on human sentence processing is to characterize the processing difficulty associated with the comprehension of words in context. Models that explain and predict this difficulty can be broadly divided into two kinds, expectation-based and memory-based. In this work, we present a new model of incremental sentence processing difficulty that unifies and extends key features of both kinds of models. Our model, lossy-context surprisal, holds that the processing difficulty at a word in context is proportional to the surprisal of the word given a lossy memory representation of the context—that is, a memory representation that does not contain complete information about previous words. We show that this model provides an intuitive explanation for an outstanding puzzle involving interactions of memory and expectations: language-dependent structural forgetting, where the effects of memory on sentence processing appear to be moderated by language statistics. Furthermore, we demonstrate that dependency locality effects, a signature prediction of memory-based theories, can be derived from lossy-context surprisal as a special case of a novel, more general principle called information locality.     

How strongly do word reading times and lexical decision times correlate? Combining data from eye movement corpora and megastudies

We assess the amount of shared variance between three measures of visual word recognition latencies: eye movement latencies, lexical decision times, and naming times. After partialling out the effects of word frequency and word length, two well-documented predictors of word recognition latencies, we see that 7–44% of the variance is uniquely shared between lexical decision times and naming times, depending on the frequency range of the words used. A similar analysis of eye movement latencies shows that the percentage of variance they uniquely share either with lexical decision times or with naming times is much lower. It is 5–17% for gaze durations and lexical decision times in studies with target words presented in neutral sentences, but drops to 0.2% for corpus studies in which eye movements to all words are analysed. Correlations between gaze durations and naming latencies are lower still. These findings suggest that processing times in isolated word processing and continuous text reading are affected by specific task demands and presentation format, and that lexical decision times and naming times are not very informative in predicting eye movement latencies in text reading once the effect of word frequency and word length are taken into account. The difference between controlled experiments and natural reading suggests that reading strategies and stimulus materials may determine the degree to which the immediacy-of-processing assumption and the eye–mind assumption apply. Fixation times are more likely to exclusively reflect the lexical processing of the currently fixated word in controlled studies with unpredictable target words rather than in natural reading of sentences or texts.     

Parsing as a Cue-Based Retrieval Model

This paper develops a novel psycholinguistic parser and tests it against experimental and corpus reading data. The parser builds on the recent research into memory structures, which argues that memory retrieval is content-addressable and cue-based. It is shown that the theory of cue-based memory systems can be combined with transition-based parsing to produce a parser that, when combined with the cognitive architecture ACT-R, can model reading and predict online behavioral measures (reading times and regressions). The parser's modeling capacities are tested against self-paced reading experimental data (Grodner & Gibson, 2005), eye-tracking experimental data (Staub, 2011), and a self-paced reading corpus (Futrell et al., 2018).     

Threats may be negative promises (but warnings are more than negative tips)

In everyday situations conditional promises, threats, tips, and warnings are commonplace. Previous research has reported disruption to eye movements during reading when conditional promises are produced by someone who does not have control over the conditional outcome event, but no such disruption for the processing of conditional tips. In the present paper, we examine how readers process conditional threats and warnings. We compare one account which views conditional threats and warnings simply as promises and tips with negative outcomes, with an alternative account which highlights their broader pragmatic differences. In an eye-tracking experiment we find evidence suggesting that, in processing terms, while threats operate like negative promises, warnings are more than negative tips.     

Memory Versus Expectation: Processing Relative Clauses in a Flexible Word Order Language

Memory limitations and probabilistic expectations are two key factors that have been posited to play a role in the incremental processing of natural language. Relative clauses (RCs) have long served as a key proving ground for such theories of language processing. Across three self-paced reading experiments, we test the online comprehension of Hungarian subject- and object-extracted RCs (SRCs and ORCs, respectively). We capitalize on the syntactic properties of Hungarian that allow for a variety of word orders within RCs, which helps us to delineate the processing costs associated with memory demand and violated expectations. Results showed a processing cost at the RC verb for structures that have longer verb-argument distances, despite those structures being more frequent in the corpus. These findings thus support theories that attribute processing difficulty to memory limitations, rather than theories that attribute difficulty to less expected structures.     

Basic processes in reading aloud and colour naming: Towards a better understanding of the role of spatial attention

Whether or not lexical access from print requires spatial attention has been debated intensively for the last 30 years. Studies involving colour naming generally find evidence that “unattended” words are processed. In contrast, reading-based experiments do not find evidence of distractor processing. One theory ascribes the discrepancy to weaker attentional demands for colour identification. If colour naming does not capture all of a subject's attention, the remaining attentional resources can be deployed to process the distractor word. The present study combined exogenous spatial cueing with colour naming and reading aloud separately and found that colour naming is less sensitive to the validity of a spatial cue than is reading words aloud. Based on these results, we argue that colour naming studies do not effectively control attention so that no conclusions about unattended distractor processing can be drawn from them. Thus we reiterate the consistent conclusion drawn from reading aloud and lexical decision studies: There is no word identification without (spatial) attention.     

The effect of word position on eye-movements in sentence and paragraph reading

The present study explores the role of the word position-in-text in sentence and paragraph reading. Three eye-movement data sets based on the reading of Dutch and German unrelated sentences reveal a sizeable, replicable increase in reading times over several words at the beginning and the end of sentences. The data from the paragraph-based English-language Dundee corpus replicate the pattern and also indicate that the increase in inspection times is driven by the visual boundaries of the text organized in lines, rather than by syntactic sentence boundaries. We argue that this effect is independent of several established lexical, contextual, and oculomotor predictors of eye-movement behaviour. We also provide evidence that the effect of word position-in-text has two independent components: a start-up effect, arguably caused by a strategic oculomotor programme of saccade planning over the line of text, and a wrap-up effect, originating in cognitive processes of comprehension and semantic integration.     

句法加工中的预测现象及实验证据

句法加工是读者通过整合输入的言语信息来理解句子的过程。句法加工的存在,使得读者能够提前预测文本的信息,提高阅读效率。由于阅读离不开信息整合,研究者提出两类不同的句法理论:句法预测的模块局域性理论和句法预测的互动分析性观点。这两类理论的争论点在于句法加工是单独模块化的自上而下加工,还是自下而上-自上而下的互动分析加工过程。其区别表现为句法预测的作用——前者认为句法预测对句子加工有抑制作用,后者认为句法预测对句子加工起促进作用。来自眼动、ERP等研究的数据佐证了句法预测的存在。未来研究应围绕其差异进行深入探讨,以揭示句法加工的实质。     

Deep dyslexia, imageability, and ease of predication

A development of existing theories of deep dyslexia is outlined. It proposes that the effects of imageability upon the ease of reading of words by deep dyslexics occur as a result of variation in the ease with which individual words summon semantic predicates, on the basis of which reading responses can be made. Ease-of-predication scores are obtained for a corpus of nouns and found to be, as hypothesized, closely related to imageability scores. It is shown that the other major characteristics of deep dyslexia can also be accounted for by this proposed mechanism. These include the well-established effects of syntactic category (in particular, of the distinction between content and function words) upon reading. Further evidence is provided that this effect may be attributed to variation in ease of predication.     

Processing relative clauses in Chinese

This paper reports results from a self-paced reading study in Chinese that demonstrates that object-extracted relative clause structures are less complex than corresponding subject-extracted structures. These results contrast with results from processing other Subject-Verb-Object languages like English, in which object-extracted structures are more complex than subject-extracted structures. A key word-order difference between Chinese and other Subject-Verb-Object languages is that Chinese relative clauses precede their head nouns. Because of this word order difference, the results follow from a resource-based theory of sentence complexity, according to which there is a storage cost associated with predicting syntactic heads in order to form a grammatical sentence. The results are also consistent with a theory according to which people have less difficulty processing embedded clauses whose word order matches the word order in main clauses. Some corpus analyses of Chinese texts provide results that constrain the classes of possible frequency-based theories. Critically, these results demonstrate that there is nothing intrinsically easy about extracting from subject position: depending on the word order in the main clause and in a relative clause, extraction from object position can be easier to process in some circumstances.     

Eye movements reveal the on-line computation of lexical probabilities during reading

Skilled readers are able to derive meaning from a stream of visual input with remarkable efficiency. In this article, we present the first evidence that statistical information latent in the linguistic environment can contribute to an account of reading behavior. In two eye-tracking studies, we demonstrate that the transitional probabilities between words have a measurable influence on fixation durations, and using a simple Bayesian statistical model, we show that lexical probabilities derived by combining transitional probability with the prior probability of a word's occurrence provide the most parsimonious account of the eye movement data. We suggest that the brain is able to draw upon statistical information in order to rapidly estimate the lexical probabilities of upcoming words: a computationally inexpensive mechanism that may underlie proficient reading.     

Effects of word frequency and visual complexity on eye movements of young and older Chinese readers

Research using alphabetic languages shows that, compared to young adults, older adults employ a risky reading strategy in which they are more likely to guess word identities and skip words to compensate for their slower processing of text. However, little is known about how ageing affects reading behaviour for naturally unspaced, logographic languages like Chinese. Accordingly, to assess the generality of age-related changes in reading strategy across different writing systems we undertook an eye movement investigation of adult age differences in Chinese reading. Participants read sentences containing a target word (a single Chinese character) that had a high or low frequency of usage and was constructed from either few or many character strokes, and so either visually simple or complex. Frequency and complexity produced similar patterns of influence for both age groups on skipping rates and fixation times for target words. Both groups therefore demonstrated sensitivity to these manipulations. But compared to the young adults, the older adults made more and longer fixations and more forward and backward eye movements overall. They also fixated the target words for longer, especially when these were visually complex. Crucially, the older adults skipped words less and made shorter progressive saccades. Therefore, in contrast with findings for alphabetic languages, older Chinese readers appear to use a careful reading strategy according to which they move their eyes cautiously along lines of text and skip words infrequently. We propose they use this more careful reading strategy to compensate for increased difficulty processing word boundaries in Chinese.     

Parafoveal processing in reading

The present review summarizes research investigating how words are identified parafoveally (and foveally) in reading. Parafoveal and foveal processing are compared when no other concurrent task is required (e.g., in single-word recognition tasks) and when both are required simultaneously (e.g., during reading). We first review methodologies used to study parafoveal processing (e.g., corpus analyses and experimental manipulations, including gaze-contingent display change experiments such as the boundary, moving window, moving mask, and fast priming paradigms). We then turn to a discussion of the levels of representation at which words are processed (e.g., orthographic, phonological, morphological, lexical, syntactic, and semantic). Next, we review relevant research regarding parafoveal processing, summarizing the extent to which words are processed at each of those levels of representation. We then review some of the most controversial aspects of parafoveal processing, as they relate to reading: (1) word skipping, (2) parafoveal-on-foveal effects, and (3) n + 1 and n + 2 preview benefit effects. Finally, we summarize two of the most advanced models of eye movements during reading and how they address foveal and parafoveal processing.