Levels of representation in the interpretation of anaphoric reference and instrument inference

Age-related differences in the retrieval of phonologically similar words during sentence processing: Evidence from ERPs

We investigated how phonologically similar words are encoded and retrieved from memory during sentence processing across younger and older adults. Critical sentences included two phonologically similar or dissimilar noun phrases (henceforth NPs) followed by a pronoun. We examined brain activity time-locked to the onsets of the second NP, and the pronoun to investigate the encoding and retrieval of the NPs, respectively. Encoding the second NP resulted in smaller N400 amplitudes when the preceding NP was phonologically similar, for both younger and older adults, suggesting age-invariant encoding facilitation with increasing phonological similarity. However, when processing the pronoun, younger adults exhibited greater negativity following phonologically similar NPs, suggesting retrieval difficulty, whereas older adults showed greater negativity for pronouns following dissimilar NPs, suggesting an apparent retrieval facilitation. A post-hoc behavioral experiment suggested that older adults perform shallow processing during retrieval. The results suggest age-related decline in retrieval, but not encoding, of phonological information.

"A cruel king" is not the same as "a king who is cruel": Modifier position affects how words are encoded and retrieved from memory

We examined whether the position of modifiers in English influences how words are encoded and subsequently retrieved from memory. Compared with premodifiers, postmodifiers might confer more perceptual significance to the associated head nouns, are more consistent with the "given-before-new" information structure, and might also be easier to integrate because the head noun is available before the modifications are encountered. In 4 experiments, we investigated whether premodified (the cruel and merciless king), and postmodified (the king who was cruel and merciless) noun phrases (henceforth, NPs) could induce variations in ease of subsequent retrieval. In Experiments 1, 2, and 3, participants used more pronouns (he), as opposed to full descriptions (the king) to refer to postmodified NPs than to unmodified competitors, but pronominal reference to premodified NPs and unmodified competitors did not differ, suggesting that postmodified NPs are more accessible in memory. When the data from all 3 experiments were combined, we also observed significantly more pronominal reference to post- than to premodified NPs, as well as a greater increase in pronominal reference rates between postmodified NPs and unmodified competitors than between premodified NPs and unmodified competitors. In Experiment 4, words following critical pronouns were read faster when the pronouns referred to modified than to unmodified NPs, and also when the pronouns referred to post- rather than premodified NPs. Taken together, our results show enhanced retrieval facilitation for postmodified NPs compared with premodified NPs. These results are the first to demonstrate that the linear position of modifications results in measurable processing cost at a subsequent point. The results have important implications for memory-based theories of language processing, and also for theories assigning a central role for discourse status and information structure during sentence processing. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

What does "it" mean, anyway? Examining the time course of semantic activation in reference resolution

Pronouns serve a critical referential function, yet the cognitive processes engaged during pronoun comprehension remain incompletely understood. One view is that encountering a pronoun leads the comprehender to reactivate the semantic features of its antecedent. We examined this by manipulating the concreteness of a noun antecedent and assessing whether an Event Related Potential (ERP) concreteness effect was elicited at a downstream pronoun. We observed a robust concreteness effect at the noun, but no similar effect at the pronoun. We also examined whether N400 semantic priming from the antecedent would increase on content words shortly following the pronoun, relative to those preceding it. Again, although we observed semantic priming following the noun, it did not increase following the pronoun. Our data suggest that pronouns do not induce the activation of (much) new semantic information in long-term memory, perhaps instead triggering an attentional shift towards their antecedents' extant representations within working memory.

On-Line Reflexive Resolution in Japanese Causative Garden Path Sentences

This study explored the time course in which lexical information in causative verbs influences the identification of the antecedent of a Japanese reflexive pronoun, jibun. Since verb information specifies that the reflexive is bound to the indirect object despite its being ordinarily associated with the subject in Japanese sentences, this requires parsers to revise the syntactic representation they had previously formed. 42 students were required to identify quickly the antecedent when given a marker which was placed immediately after the reflexive, immediately at the end of the sentence following a verb, or 4 sec. after the end of the sentence. Correct identification was almost nil immediately after the reflexive, increasing to 20.3% correct immediately after the end of the sentence and 52.4% correct 4 sec. after the end of the sentence. This finding indicates that the verb-control information is not used immediately when parsing sentences in which the verb occurs at the end of the sentence. Thus, the products of moment-to-moment initial syntactic computation are not immediately mapped onto the final semantic interpretations.

Good-enough linguistic representations and online cognitive equilibrium in language processing

We review previous research showing that representations formed during language processing are sometimes just “good enough” for the task at hand and propose the “online cognitive equilibrium” hypothesis as the driving force behind the formation of good-enough representations in language processing. Based on this view, we assume that the language comprehension system by default prefers to achieve as early as possible and remain as long as possible in a state of cognitive equilibrium where linguistic representations are successfully incorporated with existing knowledge structures (i.e., schemata) so that a meaningful and coherent overall representation is formed, and uncertainty is resolved or at least minimized. We also argue that the online equilibrium hypothesis is consistent with current theories of language processing, which maintain that linguistic representations are formed through a complex interplay between simple heuristics and deep syntactic algorithms and also theories that hold that linguistic representations are often incomplete and lacking in detail. We also propose a model of language processing that makes use of both heuristic and algorithmic processing, is sensitive to online cognitive equilibrium, and, we argue, is capable of explaining the formation of underspecified representations. We review previous findings providing evidence for underspecification in relation to this hypothesis and the associated language processing model and argue that most of these findings are compatible with them.

The Effects of Student and Text Characteristics on the Oral Reading Fluency of Middle-Grade Students

We evaluated the effects of student characteristics (sight word reading efficiency, phonological decoding, verbal knowledge, level of reading ability, grade, gender) and text features (passage difficulty, length, genre, and language and discourse attributes) on the oral reading fluency of a sample of middle-school students in Grades 6-8 (N = 1,794). Students who were struggling (n = 704) and typically developing readers (n = 1,028) were randomly assigned to read five 1-min passages from each of 5 Lexile bands (within student range of 550 Lexiles). A series of multilevel analyses showed that student and text characteristics contributed uniquely to oral reading fluency rates. Student characteristics involving sight word reading efficiency and level of decoding ability accounted for more variability than reader type and verbal knowledge, with small, but statistically significant effects of grade and gender. The most significant text feature was passage difficulty level. Interactions involving student text characteristics, especially attributes involving overall ability level and difficulty of the text, were also apparent. These results support views of the development of oral reading fluency that involve interactions of student and text characteristics and highlight the importance of scaling for passage difficulty level in assessing individual differences in oral reading fluency.

Effects of role typicality on processing person information in German: evidence from an ERP study

The present ERP study investigated how reference resolution is affected by semantic cues to referent gender such as the gender typicality of role names. Participants read general statements about social and occupational groups denoted by a role name (e.g., pilots, florists), which was followed by a co-referring NP with masculine, feminine or neutral lexical gender (e.g., these men/women/people) that could either semantically match, mismatch or be neutral to the role name's gender typicality. The study was run in German. Between 300 and 400 ms (N400) responses to all anaphors were more negative after typically male than after typically female antecedents. Between 500 and 700 ms (P600) role name typicality interacted with the lexical gender of the anaphor. Responses to feminine anaphors were more positive in case of incongruent compared to congruent typicality of the antecedent. The effects suggest that participants anticipate a possible semantic incongruity before the anaphor is actually resolved. They, furthermore show that different cues to referent gender are conceptually integrated after 500 ms following the onset of the anaphoric noun. Results are discussed with regard to two-stage models of reference resolution.

From simple associations to systematic reasoning: A connectionist representation of rules, variables and dynamic bindings using temporal synchrony

Human agents draw a variety of inferences effortlessly, spontaneously, and with remarkable efficiency – as though these inferences were a reflexive response of their cognitive apparatus. Furthermore, these inferences are drawn with reference to a large body of background knowledge. This remarkable human ability seems paradoxical given the complexity of reasoning reported by researchers in artificial intelligence. It also poses a challenge for cognitive science and computational neuroscience: How can a system of simple and slow neuronlike elements represent a large body of systemic knowledge and perform a range of inferences with such speed? We describe a computational model that takes a step toward addressing the cognitive science challenge and resolving the artificial intelligence paradox. We show how a connectionist network can encode millions of facts and rules involvingn-ary predicates and variables and perform a class of inferences in a few hundred milliseconds. Efficient reasoning requires the rapid representation and propagation of dynamic bindings. Our model (which we refer to as SHRUTI) achieves this by representing (1) dynamic bindings as the synchronous firing of appropriate nodes, (2) rules as interconnection patterns that direct the propagation of rhythmic activity, and (3) long-term facts as temporal pattern-matching subnetworks. The model is consistent with recent neurophysiological evidence that synchronous activity occurs in the brain and may play a representational role in neural information processing. The model also makes specific psychologically significant predictions about the nature of reflexive reasoning. It identifies constraints on the form of rules that may participate in such reasoning and relates the capacity of the working memory underlying reflexive reasoning to biological parameters such as the lowest frequency at which nodes can sustain synchronous oscillations and the coarseness of synchronization.

Comprehending Conceptual Anaphors in Spanish

This paper examines the mechanisms involved in the assignment of an antecedent to an anaphoric element. In general, pronouns must match their antecedents at least with respect to number and gender. Sensitivity to such constraints has been shown in several experiments. But Gernsbacher (1991) has also shown that people have no difficulty comprehending a plural pronoun with an antecedent that is grammatically singular but conceptually plural. In the first three experiments, we tested whether such a "conceptual effect" was preserved with zero anaphors in Spanish. (The typical omission of pronouns in subject position in Spanish.) Verbs in a second clause were marked with plural or singular endings. Plural verbs were rated more natural than singular verbs when they followed three types of singular but conceptually plural antecedents (Experiment 1). Clauses containing plural verbs were read faster when they followed one type of singular but conceptually plural antecedents, i.e. collective sets (Experiments 2 and 3). In fact, clauses containing plural verbs were read equally fast when they followed literally singular collective sets or explicitly group nouns. Using pronominal anaphors, these reading time effects were replicated and extended to sentences that contained generic types as antecedents (Experiment 4). The results are discussed in terms of the use of information during the comprehension of anaphors.

Testing the shared resource assumption in theories of text processing

Eight experiments evaluated a core assumption of several theories of text processing, the shared resource assumption, which states that component text processes share limited processing resources. Short texts each contained two critical sentences that together warranted a causal inference. The syntactic structure of the second sentence was either more or less difficult to parse. Results from a lexical decision task suggested that readers formed the causal inferences when the syntactic structure was less difficult to parse but that inferencing was constrained when syntactic structure was more difficult. Follow-up experiments suggested that this interference was not due to inferior output of the syntactic parser nor to the increased demands of difficult syntax interfering with maintenance of information needed to form the inference. The results suggest insufficient resources were available for the operation of inference processes due to the increased demands of syntactic parsing, consistent with the shared resource assumption.

Retrieving text inferences: controlled and automatic influences

Bridging inferences contribute to text coherence by identifying the connections among ideas, whereas elaborative inferences simply specify sensible extrapolations from text. Bridging inferences have been indistinguishable from explicit text ideas on numerous measures, suggesting similar longterm memory (LTM) representations for the two, whereas elaborative inferences are inferior. To evaluate the LTM representations of text ideas, we used the extended process-dissociation procedure (Buchner, Erdfelder, & Vaterrodt-Plunnecke, 1995; Jacoby, 1991) to partition the controlled, recollective contributions to text retrieval from the automatic, familiarity-based contributions. The automatic contribution to the recognition of implied concepts was consistently negligible, an outcome consistent with the absence of perceptual processing of those concepts during the original reading. In addition, the controlled basis of recognition was consistently higher for explicit than for implicit concepts, which suggests a more robust conceptual representation for explicit text ideas (Yonelinas, 2002). These results were interpreted to reflect the asymmetric representation of explicit ideas and inferences (elaborative and even bridging inferences) in the surface, propositional textbase, and situational levels of text representation.

Zero anaphora: transfer of reference tracking strategies from Chinese to English

English contains many nominals that would be absent in Chinese because Chinese makes greater use of "zero anaphora," which is an empty grammatical slot in a sentence standing for a previously mentioned referent. Native Chinese and native English speakers were compared in comprehending modified English passages from a standardized reading test with nominals deleted that would be absent in Chinese. In three experiments, Chinese speakers showed superior comprehension as measured by objective multiple-choice questions in this case, but not when no words were deleted or when nominals were deleted that would not be absent in Chinese. These results imply that native Chinese speakers develop reference tracking strategies that they transfer to comprehending English.

Semantic category discrimination and N400

14 subjects performed a semantic category-discrimination task in which they were presented pairs of words and asked to judge whether these belonged to the same category. Three semantic distance conditions, established for the word pairs on the basis of a survey, were semantically close condition (mammals and birds condition) and two semantically distant conditions (mammals and tools condition and birds and tools condition). The reaction times and the percentage of errors were significantly greatest for the mammals and bird condition. There were no significant differences among the three different conditions in N400 latency for target, but the amplitude of N400 tended to be largest for the mammal and bird condition. Comparison N400 of prime and target for each of the conditions showed that the latency for prime was longer than for target in all three conditions.

How Automatically Do Readers Infer Fictional Characters' Emotional States?

We propose that reading stories, such as a narrative about a character who takes money from a store where his best friend works and who later learns that his best friend has been fired, stimulates readers to activate the knowledge of how the character feels when he finds out that his best friend has been fired from a job for something he did. In other words, we propose that readers infer fictional character's emotional states. In this article, we first review two series of laboratory experiments (Gernsbacher, Goldsmith, & Robertson, 1992; Gernsbacher & Robertson, 1992) that empirically tested this hypothesis by measuring participants' reading times to target sentences that contained emotion words that matched (e.g., guilt) or mismatched (e.g., pride) the implied emotional state. We then present a third series of laboratory experiments that tested how automatically such knowledge is activated by using a divided-attention task (tone-identification, per-sentence memory load, or cumulative memory load) and by comparing target-sentence reading time when the emotional state is explicitly mentioned versus only implicit.