Age differences in resolving anaphoric expressions during reading

ABSTRACT

One crucial component of reading comprehension is the ability to bind current information to earlier text, which is often accomplished via anaphoric expressions (e.g., pronouns referring to previous nouns). Processing time for anaphors that violate expectations (e.g., ‘The firefighter burned herself while rescuing victims from the building’) provide a window into how the semantic representation of the referent is instantiated and retained up to the anaphor. We present data from three eye-tracking experiments examining older and younger adults' reading patterns for passages containing such local expectancy violations. Younger adults quickly registered and resolved the expectancy violation at the point at which it first occurred (as measured by increased gaze duration on the anaphor), regardless of whether sentences were read in isolation or embedded in a discourse context. Older adults, however, immediately noticed the violation only when sentences were embedded in discourse context, suggesting that they relied more on situational grounding to instantiate the referent. For neither young nor old did prior disambiguation within the context (e.g., stating the firefighter was a woman) reduce the effect of the local violation on early processing. For older readers, however, prior disambiguation facilitated anaphor resolution by reducing reprocessing. These results suggest that (a) anaphor resolution unfolds serially, such that prior disambiguating context does not ‘inoculate’ against local activation of salient (but contextually inappropriate) features, and that (b) older readers use the situational grounding of discourse context to support earlier access to the antecedent, and are more likely to reprocess the context for anaphor resolution.     

How do we process novel conceptual combinations in context?

Most theories of how people interpret novel noun-noun combinations (e.g., ghost forest) do not sufficiently address the role of context, which is surprising given that novel combinations are generally encountered in discourse contexts. We evaluated three hypotheses for the role of context in interpretation. The generation hypothesis states that interpretation is based on sense generation both in and out of context. The anaphor resolution hypothesis states that interpretation initially involves identifying a referent from the prior discourse context, and meaning generation occurs only if anaphoric processing fails. The dual-process hypothesis states that sense generation and anaphor resolution operate in parallel, and both may influence initial interpretation. In two experiments, we examined reading times on novel combinations when the dominant meaning (most likely interpretation) or subordinate meaning (infrequent interpretation) preceded or followed the combination in text. Results supported the dual-process hypothesis.     

How do we process novel conceptual combinations in context?

Most theories of how people interpret novel noun–noun combinations (e.g., ghost forest) do not sufficiently address the role of context, which is surprising given that novel combinations are generally encountered in discourse contexts. We evaluated three hypotheses for the role of context in interpretation. The generation hypothesis states that interpretation is based on sense generation both in and out of context. The anaphor resolution hypothesis states that interpretation initially involves identifying a referent from the prior discourse context, and meaning generation occurs only if anaphoric processing fails. The dual-process hypothesis states that sense generation and anaphor resolution operate in parallel, and both may influence initial interpretation. In two experiments, we examined reading times on novel combinations when the dominant meaning (most likely interpretation) or subordinate meaning (infrequent interpretation) preceded or followed the combination in text. Results supported the dual-process hypothesis.     

The Interaction of Contextual and Syntactic Information in the Processing of Turkish Anaphors

In contrast with languages where anaphors can be classified into pronouns and reflexives, Turkish has a tripartite system that consists of the anaphors o, kendi, and kendisi. The syntactic literature on these anaphors has proposed that whereas o behaves like a pronoun and kendi behaves like a reflexive, kendisi has a more flexible behavior and it can function as both a pronoun and a reflexive. Using acceptability judgments and a self-paced reading task, we examined how Turkish anaphors are processed in isolated sentences and within larger discourse contexts. We manipulated contextual information by creating passages where the context favored a local, long-distance or extra-sentential referent prior to the appearance of the anaphor. We measured the effect of the context on participants’ reading times and their end-of-trial coreference assignments. Our results suggest that contextual information affects the interpretive possibilities associated with an anaphor, but that the influence of context depends on the degree to which the anaphor is syntactically constrained.     

Understanding anaphors in story dialogue

Previous research has identified two processes that play an important role in anaphor resolution. An activation process increases the accessibility of an anaphor’s referent; a suppression process diminishes the accessibility of its nonreferents. In this study, we examined how these processes operate when reference to two story characters shifts rapidly, as it does in story dialogue. Dialogue raises interesting questions about how the antecedent of an anaphor becomes the most activated entity in the reader’s discourse model. Do readers suppress an anaphor’s nonreferent even though that same entity is likely to be the referent of a subsequent anaphor? Are activation and suppression processes triggered by the anaphor itself or by cues that signal a change of speaker? We found that an anaphor’s antecedent is activated differently in dialogue than it is elsewhere in a narrative. Our results suggest that readers use knowledge about the structure of dialogue to anticipate the referent of an upcoming anaphor.     

Event boundaries and anaphoric reference

The current study explored the finding that parsing a narrative into separate events impairs anaphor resolution. According to the Event Horizon Model, when a narrative event boundary is encountered, a new event model is created. Information associated with the prior event model is removed from working memory. So long as the event model containing the anaphor referent is currently being processed, this information should still be available when there is no narrative event boundary, even if reading has been disrupted by a working-memory-clearing distractor task. In those cases, readers may reactivate their prior event model, and anaphor resolution would not be affected. Alternatively, comprehension may not be as event oriented as this account suggests. Instead, any disruption of the contents of working memory during comprehension, event related or not, may be sufficient to disrupt anaphor resolution. In this case, reading comprehension would be more strongly guided by other, more basic language processing mechanisms and the event structure of the described events would play a more minor role. In the current experiments, participants were given stories to read in which we included, between the anaphor and its referent, either the presence of a narrative event boundary (Experiment 1) or a narrative event boundary along with a working-memory-clearing distractor task (Experiment 2). The results showed that anaphor resolution was affected by narrative event boundaries but not by a working-memory-clearing distractor task. This is interpreted as being consistent with the Event Horizon Model of event cognition.     

Anaphor resolution as a function of spatial distance and priming: exploring the spatial distance effect in situation models

Anaphor resolution has been found to depend on the spatial distance between the reader's focus of attention and the location of the anaphor referent in a spatially organized situation model (spatial distance effect; Rinck & Bower, 1995). This effect implies that a) the situation model is spatially organized and b) spatial distance has a stronger effect on the resolution of anaphoric reference than the text priming the anaphor referent. In three experiments, adult participants read 12 short narratives about protagonists moving around a building. Mentionning the location of the anaphor referent in text prior to the anaphoric sentence facilitated anaphor resolution. Decreased spatial distance consistently facilitated anaphor resolution, even when priming the anaphor referent affected anaphor resolution more strongly than spatial distance. Results are discussed with regard to the interpretation and reliability of the spatial distance effect and the interaction of different representational levels in the context of multi-level theories of text comprehension.     

Violating stereotypes: Eye movements and comprehension processes when text conflicts with world knowledge

We investigated the effect of discourse context on the access of word meaning during reading. Target words were role names (e.g., electrician) for which there was a gender stereotype (e.g., electricians are stereotypically male). Target sentences contained a reflexive pronoun that referred to the role name (e.g., The electrician taught herself ...). Participants read these target sentences with or without paragraph context while their eye movements were monitored. In the absence of discourse context and in neutral discourse contexts, fixation times on the reflexive pronoun and immediately following the pronoun were inflated when the pronoun specified a gender that mismatched the stereotype, indicating that the gender stereotype was activated upon encountering the role name. When prior discourse context indicated the gender of the role-named character, this mismatch effect was eliminated. The mismatch effect indicates that gender stereotypes are automatically activated in the absence of disambiguating information. The lack of an effect when gender has previously been specified is consistent with the lexical reinterpretation model proposed by Hess, Foss, and Carroll (1995).     

语篇背景在语义整合中的作用

通过眼动实验考察了语篇背景对不同类型句子语义整合过程的影响。句子类型有3个水平：(1)宾语违反了动词的选择性限制; (2)宾语符合选择性限制, 但是违反了世界知识; (3)无违反。语篇背景有2个水平：(1)中性背景; (2)校正背景。在中性语篇背景下, 句子(1) (2)是不合适的; 而在校正背景下, 三种句子都是合理的。24名大学生被试自然阅读的眼动数据表明, 违反动词选择性限制的目标词上的首次注视时间、回视路径时间和总注视时间, 在中性背景下比在校正背景下长, 注视点个数也更多; 而在非选择性限制违反条件下, 语篇背景的效应只在总注视时间和注视点个数两个指标上显著。这些结果表明, 当句子中的宾语违反了动词的选择性限制时, 语篇背景影响了宾语的早期和晚期加工。而当宾语只违反世界知识时, 语篇背景的作用体现在晚期阶段。语篇背景对当前信息的整合是动态的过程, 并不是只能在固定某个阶段起作用。     

Slow and steady: sustained effects of lexico-semantic associations can mediate referential impairments in schizophrenia

The present study investigated the contribution of lexico-semantic associations to impairments in establishing reference in schizophrenia. We examined event-related potentials as schizophrenia patients and healthy, demographically matched controls read five-sentence scenarios. Sentence 4 introduced a noun that referred back to three possible referents introduced in Sentences 1–3. These referents were contextually appropriate, contextually inappropriate but lexico-semantically associated, and contextually inappropriate and lexico-semantically nonassociated. In order to determine whether participants had correctly linked the anaphor to its referent, the final sentence reintroduced each referent, and participants indicated whether the last two sentences referred to the same entity. Results indicated that between 300 and 400 ms, patients, like healthy controls, used discourse context to link the noun with its preceding referent. However, between 400 and 500 ms, neural activity in patients was modulated only by lexico-semantic associations, rather than by discourse context. Moreover, patients were also more likely than controls to incorrectly link the noun with contextually inappropriate but lexico-semantically associated referents. These results suggest that at least some types of referential impairments may be driven by sustained activation of contextually inappropriate lexico-semantic associations.     

藏-汉读者汉语歧义句加工的歧义效应及语境促进效应

采用眼动技术,通过两个实验,考察藏-汉读者阅读汉语歧义句时的歧义效应以及语境在句子歧义消解中的作用。结果发现:藏-汉读者在阅读汉语歧义句时存在歧义效应; 语境对歧义消解起促进作用,表现为,当语境置于歧义句之前或语境意义偏向歧义句主要意思时,藏-汉读者对歧义句的加工更容易。并且,歧义效应和语境促进效应贯穿于早期阶段到晚期阶段的整个句子加工过程。     

常识性知识和语篇语境对代词指认的影响

本研究围绕中文代词理解中, 读者的常识性知识和语篇语境的作用以及作用时程这一问题展开。实验1首先探讨在中文阅读理解中, 职业性别倾向这种常识性信息是否能对代词的指认产生早期影响。存在一致和冲突(如警卫-他; 警卫-她)两个实验条件, 结果在凝视时间、重读时间和总阅读时间上都发现了职业性别倾向所引发的性别冲突效应。实验2在实验1基础上增加一个代词(如：警卫-他-他; 警卫-她-她), 考察更正后的信息, 即文本语境内容是否可以覆盖常识的作用对代词加工产生影响。结果发现语篇语境确实可以覆盖常识性信息的作用, 对代词加工产生早期影响。但是, 职业性别倾向这种常识性知识仍然在代词加工的后期阶段发挥作用。考虑到代词所形成的语境较为含蓄, 实验3中采用更明确的方式来界定职业名称的性别, 例如男艺人、爸爸等, 然后再出现一个代词, 代词的性别始终和先前的性别描述一致, 而与职业性别倾向冲突, 这样也存在一致和冲突两个条件(例如：保姆-妻子-她, 警卫-妻子-她), 结果发现只有更正后的语境信息对代词加工产生影响, 职业性别倾向不再发挥作用。说明在中文这种高语境依赖性的语言文字中, 语境可以覆盖常识性知识对代词加工产生早期作用。但是, 语境作用的持续性问题会受到语篇语境中性别信息明确程度的影响。     

How Readers Spontaneously Interpret Man-Suffix Words: Evidence from Eye Movements

This study investigated whether readers are more likely to assign a male referent to man-suffix terms (e.g. chairman) than to gender-neutral alternatives (e.g., chairperson) during reading, and whether this bias differs as a function of age. Younger and older adults' eye movements were monitored while reading passages containing phrases such as "The chairman/chairperson familiarized herself with..." On-line eye fixation data provided strong evidence that man-suffix words were more likely to evoke the expectation of a male referent in both age groups. Younger readers demonstrated inflated processing times when first encountering herself after chairman relative to chairperson, and they tended to make more regressive fixations to chairman. Older readers did not show the effect when initially encountering herself, but they spent disproportionately longer looking back to chairman and herself. The study provides empirical support for copy-editing policies that mandate the use of explicitly gender-neutral suffix terms in place of man-suffix terms.     

话题结构对语篇情绪更新的影响

读者在阅读理解的过程中, 能够推断出故事主角的情绪状态, 并能随着语篇中的情绪变化做出情绪更新。在加工情绪信息的同时, 读者也加工语篇结构方面的信息, 二者皆有可能随着语篇的展开而发生变化。本研究采用自定步速阅读的方法, 分别在外显和内隐两种情绪加工任务下, 考察话题结构对语篇情绪更新的影响。结果发现, 在实验1的外显情绪判断任务下, 话题结构未显示出对语篇情绪更新的作用; 在实验2和实验3的内隐情绪理解任务下, 话题延续时, 情绪转换句的阅读时间长于情绪延续句, 此时情绪更新需要额外的加工时间; 而话题转换时, 二者没有显著差异, 说明此时读者在新结构下建立当前句的情绪表征, 并不受到先前情绪的影响。     

Coherence-driven resolution of referential ambiguity: A computational model

We present a computational model that provides a unified account of inference, coherence, and disambiguation. It simulates how the build-up of coherence in text leads to the knowledge-based resolution of referential ambiguity. Possible interpretations of an ambiguity are represented by centers of gravity in a high-dimensional space. The unresolved ambiguity forms a vector in the same space. This vector is attracted by the centers of gravity, while also being affected by context information and world knowledge. When the vector reaches one of the centers of gravity, the ambiguity is resolved to the corresponding interpretation. The model accounts for reading time and error rate data from experiments on ambiguous pronoun resolution and explains the effects of context informativeness, anaphor type, and processing depth. It shows how implicit causality can have an early effect during reading. A novel prediction is that ambiguities can remain unresolved if there is insufficient disambiguating information.     

Anticipating explanations in relative clause processing

We show that comprehenders’ expectations about upcoming discourse coherence relations influence the resolution of local structural ambiguity. We employ cases in which two clauses share both a syntactic relationship and a discourse relationship, and hence in which syntactic and discourse processing might be expected to interact. An off-line sentence-completion study and an on-line self-paced reading study examined readers’ expectations for high/low relative-clause attachments following implicit-causality and non-implicit causality verbs (John detests/babysits the children of the musician who…). In the off-line study, the widely reported low-attachment preference for English is observed in the non-implicit causality condition, but this preference gives way to more high attachments in the implicit-causality condition in cases in which (i) the verb’s causally implicated referent occupies the high-attachment position and (ii) the relative clause provides an explanation for the event described by the matrix clause (e.g., …who are arrogant and rude). In the on-line study, a similar preference for high attachment emerges in the implicit-causality context—crucially, before the occurrence of any linguistic evidence that the RC does in fact provide an explanation—whereas the low-attachment preference is consistent elsewhere. These findings constitute the first demonstration that expectations about ensuing discourse coherence relationships can elicit full reversals in syntactic attachment preferences, and that these discourse-level expectations can affect on-line disambiguation as rapidly as lexical and morphosyntactic cues.     

Spatial coding in the processing of anaphor by good and poor readers: Evidence from eye movement analyses

An experiment that examined the way in which young readers deployed eye movements while reading sentences and while answering questions containing either a pronominal or noun anaphor is reported. To evaluate the possible causal role played by differences in inspection strategies between readers of above- and below-average reading skill, a third“age control” group of younger children was also tested. This group was matched on absolute reading ability with the less skilled group of older children, and on relative reading ability (i.e. reading quotient) with the more skilled group. Differences in inspection strategy were apparent between the groups of good and poor readers. Good readers launched more selective reinspections, whereas the poorer readers were more inclined to engage in“backtracking” and appeared to make less use of the displayed text. In every case there was a marked similarity in the behaviour of the good readers and the“age controls”. These results suggest that the ability to code the spatial location of words in a sentence, and, where necessary, to use this information to launch accurately targetted selective reinspections of previously read text, plays a crucial role in the development of skilled reading performance.     

Selecting one among many referents in spatial situation models

Five experiments related anaphor resolution to a classic memory variable, namely, interference created by multiple uses of a given object-concept, and by spatial distance of the referent from the reader's focus of attention. Participants memorized a diagram of a building with rooms containing objects, and then read narratives describing characters' activities there. Reading was self-paced word by word. Accessibility was measured by readers' time to understand anaphoric sentences containing a definite noun phrase referring to an object in its room. Spatial distance between the object and the current focus of attention increased reading times for names of the object, the room, and sentence wrap-up. Multiple examples of a target-object increased its reading time only if they were scattered across different rooms. An associative model of memory retrieval during text comprehension was used to interpret the complex pattern of results.     

文本阅读中回指推理发生的整合性因素

探讨回指距离、干扰词的精细描述、干扰词的典型性 对回指推理（先行词通达）的影响。包括2个实验,被试是华南师范大学一二年级本科生164名。采用移动窗口技术,要求被试阅读一定篇数（如,实验1a每个被试阅读15篇）的文章,对不同条件的探测词的反应时进行统计。实验1探讨干扰词精细描述类型（精细描述高和精细描述低）对回指推理的影响。结果表明,干扰的精细描述并非是影响回指推理的主要因素。实验2探讨干扰词和先行词的典型性类型的相对变化对回指推理的影响。结果表明高典型干扰在不同回指距离条件下都会对回指推理产生影响。本研究结果初步证明在影响回指推理的干扰典型性和干扰精细描述的二个因素中,起主要作用的是干扰的典型性