The Effect of Individual Differences in Working Memory Capacity on Sentence Comprehension: An fMRI Study

This study explores the interaction between working memory systems and language processing by examining how differences in working memory capacity (WMC) modulates neural activation levels and functional connectivity during sentence comprehension. The results indicate that two working memory systems may be involved in sentence comprehension, the verbal working memory system and the episodic buffer, but during different phases of the task. A sub-region of the left inferior frontal gyrus (BA 45) was correlated with WMC during the probe and not during sentence reading while the only region to reveal a correlation with WMC during sentence reading was the posterior cingulate/precuneus area, a region linked to event representation. In addition, functional connectivity analysis suggests that there were two distinct networks affected by WMC. The first was a semantic network that included the middle temporal cortex, an anterior region of the inferior frontal gyrus and the inferior parietal region. The second included the posterior cingulate and BA 45 of the inferior frontal gyrus. We propose here that high capacity readers may generate an event representation of the sentence during reading that aids in comprehension and that this event representation involves the processing of the posterior cingulate cortex.     

Sentence comprehension in Alzheimer's disease

A visual sentence‐picture matching task was used to clarify the nature of the comprehension deficit in Alzheimer's Disease (AD). Test sentences varied according to how crucial the processing of syntax was in computing a correct interpretation. Half the sentences could be understood through the comprehension of semantic cues alone; the other half required the processing of syntactic cues. In the first experiment, sentences were removed from view before the test pictures were presented; in the second, sentences and pictures were presented together to decrease demands on working memory. Two factors determined the accuracy of sentence comprehension: whether a syntactic analysis was needed and whether additional working‐memory resources were needed for temporary storage. When storage demands were minimized, patients were able to use semantic cues to guide comprehension. When semantic cues were unavailable, comprehension was impaired, regardless of whether or not there were additional storage requirements, providing evidence for a genuine syntactic deficit that is not due to coexisting semantic impairment or working‐memory dysfunction.     

A case for conflict across multiple domains: Memory and language impairments following damage to ventrolateral prefrontal cortex

Patients with focal lesions to the left inferior frontal gyrus (LIFG; BA 44/45) exhibit difficulty with language production and comprehension tasks, although the nature of their impairments has been somewhat difficult to characterize. No reported cases suggest that these patients are Broca's aphasics in the classic agrammatic sense. Recent case studies, however, do reveal a consistent pattern of deficit regarding their general cognitive processes: They are reliably impaired on tasks in which conflicting representations must be resolved by implementing top-down cognitive control (e.g., Stroop; memory tasks involving proactive interference). In the present study, we ask whether the language production and comprehension impairments displayed by a patient with circumscribed LIFG damage can best be understood within a general conflict resolution deficit account. We focus on one patient in particular—patient I.G.—and discuss the implications for language processing abilities as a consequence of a general cognitive control disorder. We compared I.G. and other frontal patients to age-matched control participants across four experiments. Experiment 1 tested participants' general conflict resolution abilities within a modified working memory paradigm in an attempt to replicate prior case study findings. We then tested language production abilities on tasks of picture naming (Experiment 2) and verbal fluency (Experiment 3), tasks that generated conflict at the semantic and/or conceptual levels. Experiment 4 tested participants' sentence processing and comprehension abilities using both online (eye movement) and offline measures. In this task, participants carried out spoken instructions containing a syntactic ambiguity, in which early interpretation commitments had to be overridden in order to recover an alternative, intended analysis of sentence meaning. Comparisons of I.G.'s performance with frontal and healthy control participants supported the following claim: I.G. suffers from a general conflict resolution impairment, which affects his ability to produce and comprehend language under specific conditions—namely, when semantic, conceptual, and/or syntactic representations compete and must be resolved.     

Involvement of the Motor System in Comprehension of Non-Literal Action Language: A Meta-Analysis Study

Although numerous studies have shown that the sensory-motor system is involved in semantic processing of language stimuli, it is still unclear whether comprehension of abstract concepts is embodied, and whether the involvement of the sensory-motor system is context-dependent. Investigation of how the motor system is activated during comprehension of non-literal action languages can help address these issues. So far several studies have reported brain activations during non-literal action language comprehension, but the findings are highly inconsistent because of different types of non-literal action language stimuli. To clarify how the motor system is involved in comprehension of different types of non-literal languages, the current study conducted quantitative meta-analyses on fMRI findings about comprehension of sentences describing fictive motions, metaphoric actions, and idiomatic actions. Results showed that fictive motion sentences elicited activation in the right parahippocampal gyrus, an area important for spatial processing. For metaphoric actions, the left precentral gyrus (BA 6) was strongly activated, suggesting a link between metaphoric and literal meanings. For idiomatic actions, activity was found in the left inferior frontal gyrus (BA 44/45), highlighting semantic selection and inhibition. No premotor or motor activity was found in idiom condition. These results together suggest that the involvement of the sensory-motor system in abstract concepts processing is flexible, depending on semantic features of the language stimuli and links between abstract and literal meanings.     

Transcranial magnetic stimulation (TMS) applied to left dorsolateral prefrontal cortex disrupts verbal working memory performance in humans

Working memory refers to the temporary maintenance and processing of information and involves executive processes that manipulate the contents of the working memory. The role of the executive function in the human left dorsolateral prefrontal cortex (LDLPFC) was explored using transcranial magnetic stimulation (TMS) after confirming the LDLPFC activation using fMRI. We applied double-pulse TMS having a 100-ms inter-pulse interval to LDLPFC immediately after the subjects finished reading the sentences of the reading span test (RST) task, an efficient measure of verbal working memory, in which dual tasks that include both sentence comprehension and word maintenance are required. Using eight normal participants, we found a significant deterioration of performance, i.e., decreased number of correctly reported words, in RST due to TMS stimulation of LDLPFC. Evidence suggests that transient functional disruption of the LDLPFC impairs performance in the maintenance processing of the RST task.     

Event-related potentials suggest early interaction between syntax and semantics during on-line sentence comprehension

Event-related potentials (ERPs) were used to investigate interaction between syntactic parsing and semantic integration processes during a visual sentence comprehension task. The linguistic stimuli were Finnish five-word sentences containing morphosyntactic and/or semantic violations. Single morphosyntactic violations evoked left anterior negativity (LAN) and P600 components. Single semantic violations elicited a robust N400 effect over the left hemisphere. A later and weaker N400-like response was also observed in the right hemisphere, left-right hemispheric latency difference being 40 ms. Combined morphosyntactic and semantic violations elicited a P600 component and a negative ERP component within the latency range of the LAN and N400 components. Further analysis of these ERP effects provided evidence for early processual interaction between syntax and semantics during on-line sentence comprehension. The hemispheric distribution of the LAN and N400 components was taken to suggest lateralization of initial morphosyntactic parsing and semantic integration processes to the left hemisphere. In contrast, the later syntax-related P600 component was observed as being more pronounced over the posterior areas of the right hemisphere.     

Phonological short-term store and sentence processing

The grammatical and semantic processing of auditorily presented sentences and passages of prose was investigated in a left brain-damaged patient (PV), who has a reduced auditory memory span, interpreted in terms of a selective deficit of the phonological short-term store component of working memory. In the case of short sentences the patient's performance is well within the normal range, whether tested by sentence-picture matching or by the detection of syntactic or semantic anomalies.

She retains an intact capacity to detect semantic anomalies whether tested using short sentences, long sentences, or prose passages. She retains some capacity for detecting syntactic anomalies even in long sentences, provided these are tested under conditions where such mismatches are very frequent; when they are embedded in more varied material, however, her performance deteriorates. Finally, when the syntactic anomaly involves an anaphoric mismatch across sentences, her performance drops to chance level. These results are consistent with the view that the short-term phonological store serves as a “mnemonic window” that facilitates certain aspects of the comprehension of sentential material.     

Revisiting the incremental effects of context on word processing: Evidence from single‐word event‐related brain potentials

The amplitude of the N400—an event‐related potential (ERP) component linked to meaning processing and initial access to semantic memory—is inversely related to the incremental buildup of semantic context over the course of a sentence. We revisited the nature and scope of this incremental context effect, adopting a word‐level linear mixed‐effects modeling approach, with the goal of probing the continuous and incremental effects of semantic and syntactic context on multiple aspects of lexical processing during sentence comprehension (i.e., effects of word frequency and orthographic neighborhood). First, we replicated the classic word‐position effect at the single‐word level: Open‐class words showed reductions in N400 amplitude with increasing word position in semantically congruent sentences only. Importantly, we found that accruing sentence context had separable influences on the effects of frequency and neighborhood on the N400. Word frequency effects were reduced with accumulating semantic context. However, orthographic neighborhood was unaffected by accumulating context, showing robust effects on the N400 across all words, even within congruent sentences. Additionally, we found that N400 amplitudes to closed‐class words were reduced with incrementally constraining syntactic context in sentences that provided only syntactic constraints. Taken together, our findings indicate that modeling word‐level variability in ERPs reveals mechanisms by which different sources of information simultaneously contribute to the unfolding neural dynamics of comprehension.     

Processing of syntactically complex sentences relies on verbal short-term memory: Evidence from a short-term memory patient

The role of the phonological loop in auditory sentence comprehension was examined in a 35-year-old woman with a selective deficit of verbal short-term memory (STM). More specifically, the objective of the experiment was to test whether sentence comprehension is limited by number of propositions, as suggested by Rochon, Waters, and Caplan (2000), or whether it depends on syntactic complexity. In an offline task, severe impairment was present on cleft objects and centre-embedded structures; the deficit on object relatives in right peripheral position was less relevant, and the patient was able to handle sentential coordination easily. In an online task, her processing of centre-embedded structures and object relatives in right peripheral position was significantly slower than that in controls. She was also significantly slower in processing the verb of the object relative in centre-embedded structures. The results obtained do not support the claim that the difficulties encountered by patients with limited working-memory resources are due to the number of propositions in the sentence; they do, however, suggest a direct involvement of the phonological loop in processing syntactically complex sentences.     

Processing of syntactically complex sentences relies on verbal short-term memory: evidence from a short-term memory patient

The role of the phonological loop in auditory sentence comprehension was examined in a 35-year-old woman with a selective deficit of verbal short-term memory (STM). More specifically, the objective of the experiment was to test whether sentence comprehension is limited by number of propositions, as suggested by Rochon, Waters, and Caplan (2000), or whether it depends on syntactic complexity. In an offline task, severe impairment was present on cleft objects and centre-embedded structures; the deficit on object relatives in right peripheral position was less relevant, and the patient was able to handle sentential coordination easily. In an online task, her processing of centre-embedded structures and object relatives in right peripheral position was significantly slower than that in controls. She was also significantly slower in processing the verb of the object relative in centre-embedded structures. The results obtained do not support the claim that the difficulties encountered by patients with limited working-memory resources are due to the number of propositions in the sentence; they do, however, suggest a direct involvement of the phonological loop in processing syntactically complex sentences.     

The neuropsychology of sentence processing: Where do we stand?

In the early 1980s, sentence comprehension deficits were attributed to a loss of syntactic knowledge in agrammatic Broca's aphasics and to a short-term memory deficit in conduction aphasics. Findings in the remainder of the decade called both of these claims into question and presented general difficulties for the group study approach. Results from case studies support the representational independence of syntactic and semantic information but the interaction of these knowledge sources during processing. Working memory is still considered to provide critical constraints on sentence comprehension, but the capacity involved appears to be largely independent of the phonological storage involved in word list recall. Current computational approaches to sentence comprehension provide the means of accounting for the interaction of multiple sources of information and working memory requirements, but have yet to be tested against neuropsychological findings.     

Aphasic disorders of syntactic comprehension and working memory capacity

Abstract

Miyake, Carpenter, and Just (1994) present results of two sentence comprehension experiments using RSVP in normal subjects, which they say simulate important features of aphasic patients' comprehension of syntactic structures. On this basis, they claim that normal and aphasic subjects' performance represents a continuum of limitations in working memory capacity. In this paper, we argue that the Miyake et al. data do not resemble the performance of aphasic patients in crucial respects, and that their results provide evidence for the opposite conclusion…namely, that the processing resource system that underlies syntactic processing and that is affected in aphasia is substantially separate from the one used for other verbal tasks, such as reasoning and problem solving, which is tested In tasks such as the Daneman and Carpenter (1980) reading span task.     

Neural synchronization mediates on-line sentence processing: EEG coherence evidence from filler-gap constructions

This study investigated cognitive and neural processes involved in gap filling during on-line sentence comprehension. Electroencephalogram (EEG) coherences were used to demonstrate that increases in the synchronization of neural activity in different cortical regions occur during gap filling when load in semantic working memory is high due to semantically unrelated words in the filler-gap interval. Sentences could either require gap filling at a verb or not, and the nouns preceding the verb could be either semantically related or unrelated. In the unrelated but not related condition, coherences in the beta band were larger during verb processing for sentences requiring gap filling compared to sentences not requiring gap filling. The coherence changes involved linkages between frontal and posterior temporal-parietal sites in both hemispheres. These results further indicate that semantic working memory is involved in the process of gap filling.     

Early components of event-related potentials related to semantic and syntactic processes in the Japanese language

Brain activities were compared between semantic and syntactic processing in the Japanese language using event-related potentials with a 58-ch EEG system. We previously found that semantic violations elicited N400 and syntactic violations elicited P600 but not early left anterior negativity (ELAN) or left anterior negativity (LAN) using a relatively long stimulus presentation time (1 s). In the present study, we adopted a shorter stimulus presentation time (0.5 s), which might impose a heavier burden on the working memory system, to test the possible relevance of load on the working memory system to ELAN/LAN. A global field power analysis showed an increased potential field strength at the latency of 320 ms in either type, as well as those at the later latencies reflecting N400 (556 ms) and P600 (712 ms). Statistical analyses revealed a significant negative deflection in the right frontal region for the semantic type, whereas no significant deflection in either specified region was obtained for the syntactic type at the latency of 320 ms. The lack of ELAN/LAN suggested that it was not due to the deactivation of the working memory system. Moreover, scalp current density topographies implied that the processing of the verbal stimuli was mediated by distinct areas within the left temporal region, according to its semantic congruency with the preceding context at a latency as early as 320 ms. These findings are in line with the dual-route hypothesis of reading, which suggests that the reading of verbal stimuli semantically incongruent with the preceding context is dominated by phonological processes rather than lexico-semantic.     

MEG study on neural activities associated with syntactic and semantic violations in spoken Korean sentences

We carried out an magnetoencephalography (MEG) study to record cortical responses elicited in the left hemisphere by ending verb phrases, which had syntactic or semantic anomalies, in Korean sentences of subject-object-verb order. Using the high temporal and spatial resolution of MEG, the study was aimed at identifying neural activities that occur during a latency course associated with the syntactic or semantic process in the spoken sentence. Major responses, distinct from the responses to normal sentences, were observed in two latency periods of about 400 and 600 ms following the onset of the verb phrase. Source localization of the grand average fields indicated separate activities in the inferior frontal region and the vicinity of the auditory cortex for the first 400-ms response to the syntactic anomaly. The region around the auditory cortex was indicated for the response to the semantic anomaly in a similar latency. The second 600-ms response indicated activity around the middle temporal gyrus inferior to the auditory cortex for both syntactic and semantic anomalies. The results are discussed with reference to the ERP components established for Indo-European languages, and the possibility of concurrent processing of syntactic and semantic aspects is suggested.     

When task demands induce “asyntactic” comprehension: A study of sentence interpretation in aphasia

Abstract

We investigated sentence processing in two aphasic patients who appeared to have asyntactic comprehension when tested using sentence-picture matching. It was found that neither patient could handle the nonlinguistic cognitive demands of the original task: Specifically, processing two semantically incongruous inputs (sentence plus reverse-role picture) overloaded working memory. Their ability to deal with semantic conflict in the absence of multiple inputs was examined in an interleaved meaning-classification/actor-identification task. The patients rarely accepted misordered sentences like The cheese ate the muse as plausible, but performed poorly when asked to identify the “actors” in such sentences, often selecting the more likely alternative (the mouse). We concluded from this dissociation between tasks that semantic conflict only overtaxed their limited processing capacity when the conflicting options were explicitly available and directly relevant to the decision process. There was, therefore, no adverse effect in meaning classification, where the alternative (lexically-based) sentence reading (1) had to be computed by patients, and (2) bore no direct relevance to the plausibility judgement. By contrast, in actor identification, the patients had to choose between two explicitly available candidates for the actor role, one syntactically based and the other more plausible. Since our patients made errors in identifying actors immediately after correctly classifying sentences with regard to plausibility, we argued that their inaccurate performance under conditions of high processing load was more likely to reflect an inability to perform the necessary decision processes than an impaired capacity to analyse linguistic structure in the face of increased (nonlinguistic) task demands.     

Active maintenance of sentence meaning in working memory: evidence from EEG coherences.

The active maintenance of sentence meaning in working memory was investigated using event-related electroencephalogram (EEG) coherences. Participants read a sentence, retained it for 2.5 s, and then verified a statement about its meaning. The sentences contained either three semantically related nouns or unrelated nouns and started either with a what phrase (WH sentences) or not (non-WH sentences), imposing either a high or low demand on verbal working memory. Comprehension accuracy showed an interaction of semantic relatedness and sentence type due to the presence of a relatedness effect (lower accuracy in the unrelated condition) in WH sentences but not in non-WH sentences. During the post-sentence retention interval, EEG coherences also displayed this interaction of relatedness and sentence type. A semantic relatedness effect was obtained in the WH sentences (high demand) but not in the non-WH sentences (low demand). In addition, compared to a pre-sentence baseline and sentence presentation, coherences increased in the 10-14 Hz band during retention and decreased in the 4-6 Hz band. These coherence changes spanned prefrontal and posterior brain regions, possibly reflecting increased synchronization in projection loops between attention control systems in prefrontal cortex and activated meaning representations in semantic memory in posterior cortex. These findings suggest that short-term retention of the meaning of a sentence involves active maintenance in a capacity-limited working memory, accompanied by a heightened inner direction of attention after sentence presentation.     

Sentence processing is uniquely human

In this article, we will focus on three fundamental issues concerning language processing in the human brain, and update recent advances made by functional neuroimaging and magnetic stimulation studies of language. First, we will provide the first experimental evidence that the neural basis of sentence comprehension is indeed specialized. Specifically, our recent functional magnetic resonance imaging (fMRI) study has clarified that the human left prefrontal cortex (PFC) is more specialized in the syntactic processes of sentence comprehension than other domain-general processes such as short-term memory. Second, the distinction between explicit and implicit syntactic processes will be clarified, based on our fMRI studies that elucidate syntactic specialization in the left PFC. Third, we will advance a hypothesis stating that distinct subregions of the left PFC are recruited for the syntactic integration of lexico-semantic information. The current direction of research in the neuroscience of language is beginning to reveal the uniqueness of the human mind.     

Does action semantic knowledge influence mental simulation in sentence comprehension?

This research investigated whether action semantic knowledge influences mental simulation during sentence comprehension. In Experiment 1, we confirmed that the words of face-related objects include the perceptual knowledge about the actions that bring the object to the face. In Experiment 2, we used an acceptability judgment task and a word-picture verification task to compare the perceptual information that is activated by the comprehension of sentences describing an action using face-related objects near the face (near-sentence) or far from the face (far-sentence). Results showed that participants took a longer time to judge the acceptability of the far-sentence than the near-sentence. Verification times were significantly faster when the actions in the pictures matched the action described in the sentences than when they were mismatched. These findings suggest that action semantic knowledge influences sentence processing, and that perceptual information corresponding to the content of the sentence is activated regardless of the action semantic knowledge at the end of the sentence processing.     

Neural evidence for direct meaning access from orthography in Chinese word reading

A fundamental issue in the study of reading is to understand the processes involved in determining word meaning from print. We used functional magnetic resonance imaging (fMRI) and scanned participants performing lexical decision tasks to discriminate between real Chinese words and non-words, presented either visually or auditorily. For the visual task, two left inferior frontal cortical regions were significantly more activated for non-words than for words, one in BA (Brodmann's area) 44/45 implied in phonological processing, and one in BA47 implied in semantic processing. For the auditory task, stronger neural activity for non-words, relative to words, was only found in BA44/45 but not in BA47. The results were interpreted to suggest that printed words in Chinese can directly activate their semantic representations, independent of an indirect, mediated pathway through phonology. In reference to related imaging studies on English, our finding implies a greater reliance on orthography in Chinese reading.