User:Chatuka/sandbox

=Bridging and Elaborative Inferences=

The generation and processing of inferences while reading plays a critical role in comprehension. Linguists make an important distinction between bridging inferences and elaborative (or predictive) inferences. Bridging inferences are required to establish a coherent connection between propositions in a given text. That is, people make such inferences to “bridge” the gaps between sentences, resulting in comprehension. Elaborative inferences establish connections between statements in the text and general pre-existing knowledge (or “knowledge base”). Such inferences provide additional information to the reader, but are not necessarily required for coherence. Whether or not any elaborative inferences are required for comprehension has been a central area of interest to many linguistic and cognitive science researchers.

Types of Inferences
There are several types of bridging and elaborative inferences. Referential bridging inferences specify that a current word is referring to a previous word or element in the passage. Causal bridging inferences are generated as a means of connecting a current explicitly stated action or event with the context of previous text. That is, adequate comprehension requires the reader to determine that a previous event caused the outcome stated in the current sentence. Thus, such inferences are needed for a reader to comprehend narrative text. In contrast, predictive inferences (also known as causal consequence inferences) are elaborative. Predictive inferences are involved in forecasting subsequent events that are likely to occur based on the events or circumstances explicitly described in the final sentences of the text. Thus, such inferences expand on the information explicitly provided by the passage but are not critical for readers’ sufficient understanding.

Inference Theories
Many researchers in the field of cognitive psychology are interested in how inferences are encoded during processing and whether bridging and elaborative inferences are encoded differently. Inferences may be processed “on-line”, or in other words when they are generated, while reading and comprehending text. In this case, inferences are processed automatically. In contrast, “off-line” processing reflects inference generation at a time after which comprehension has occurred. In this case, inferential processing requires relatively more effort and conscious awareness.

Two influential theories of inferential processing attempt to explain the ways in which bridging and elaborative inferences are encoded while reading. Proponents of the minimalist hypothesis assert that only local connections between elements of a text are required for sufficient reading comprehension. In turn, inferences that result in local coherence are compulsory for adequate comprehension of text, and therefore, are spontaneously processed “on-line”. On the other hand, elaborative inferences provide more than is necessary for local text coherence. Thus, such inferences are suggested to be processed “off-line” unless they are based on information that is quickly and easily accessible in memory.

Not unlike the minimalist viewpoint, proponents of the constructivist view assert that only inferences that are necessary for comprehension are made “on-line”. However, the constructivist approach proposes that information that provides a reader with overall text coherence (global coherence) may also be necessary for comprehension. Comprehension entails building (or “constructing”) a situational model of events described in the text. That is, readers create and maintain mental representations of the situations or events occurring in the text to aid comprehension. Therefore, inferences that function to integrate explicit text with the knowledge base (elaborative inferences) are important, along with those that connect local text elements (bridging inferences). Hence, it is proposed that bridging inferences are always made “on-line”, but elaborative inferences may be processed automatically as well.

Research has demonstrated that inferential processing is more consistent with the constructivist approach than the minimalist hypothesis. For example, various elaborative inferences have been found to be normally generated online, such as those regarding the goals and emotions of characters and the story’s overarching theme.

Working Memory
According to Baddeley’s model, information held in working memory (WM) is actively maintained while individuals perform cognitive tasks such as: problem-solving, mathematical computations, and so on. WM stores information for very brief periods of time and can be characterized as having a limited capacity, in terms of the amount of information that can be maintained and processed at any one time.

Just and Carpenter’s (1992) Capacity Constrained Comprehension theory describes the role of individual differences in WM capacity in processing information during language comprehension. Information encountered during reading activates related concepts in long-term memory (LTM). In turn, these related concepts become part of WM where they are operated on by the reader. The activation level of items must be above a certain threshold in order to maintain the items in WM. However, the total activation of WM is limited and varies across individuals. Thus, individuals differ in how much info can be maintained and processed in their WM at a given time. That is, individuals can be characterized as having either a high or low WM span.

Research evidence suggests that WM is implicated in inference generation and processing ability during reading. WM resources are required in order to maintain representations of explicit text while simultaneously searching LTM for relevant prior knowledge. As a result, more resources are required as the complexity of inferences increases. Individuals with a high WM span can maintain more information in WM simultaneously. Consequently, more retrieval cues become available to LTM, which increases the chances of elaboration. Indeed, many studies have generally found that individuals with high WM capacity make both bridging and elaborative inferences during reading, whereas individuals with low WM span are make only bridging inferences during reading. In particular, readers with low WM capacity have consistently been found (a) to be less likely to draw elaborative than bridging inferences and (b) to generally require more inferential processing time, than those with high WM spans. Hence, a benefit of high WM capacity is an ability to draw relatively more inferences with less difficulty compared to individuals with more restricted WM capacities. This suggests that WM capacity becomes more important as the degree of inferential processing required for comprehension increases.

Trait Anxiety
Trait anxiety can be characterized as a disposition to feel anxious and distressed in response to potentially stressful situations. It is a relatively stable personality trait. Thus, individuals high in this trait have a tendency to experience long term feelings of anxiety relative to individuals low in this trait. This relation can be explained by the fact that individuals with high trait anxiety are also more likely to have lower WM capacities relative to those with low trait anxiety. In turn, WM capacity is negatively affected by stress. The experience of anxious feelings from anticipating stressful events can be distracting and interfere with information processing, including the allocation of WM resources to the cognitive task at hand. Thus, people high in trait anxiety may have more difficulty with inferential processing relative to others, because they are more likely to respond to events with feelings of anxiety over time.

Development
Younger children are generally less successful in making elaborative inferences when compared to older children. Specifically, young children have more difficulty when asked to recall the object of a particular sentence if the object was implied rather than explicitly stated in the sentence. A majority of research on this topic suggests that the use of elaborative inferences increases with age.

For instance, Casteel (1993) investigated how younger children, older children and adults process inferences while reading. Participants were instructed to read stories and answer inference-related questions. The findings indicated that elaborative inferences were not as likely to be made compared to those more necessary for comprehension (e.g., bridging inferences). All age groups showed a similar pattern in their responses to bridging inference questions, suggesting that young children could also generate inferences necessary for comprehension fairly easily. However, younger children were much slower at responding to elaborative inference questions compared to older children and adults. Again, this suggests that younger children are less likely to generate inferences that go beyond what is necessary for sufficient comprehension. One interpretation is that this pattern of increased efficiency for inferential processing runs parallel to increases in WM capacity over development.

The Brain
The N400 is a component of an ERP signal that has been implicated in language comprehension. Lower N400 signals correspond with increased likelihood the reader will generate inferences during reading. Moreover, there is evidence that meaningfully related and familiar words result in smaller N400 amplitudes relative to meaningfully unrelated and novel words.

Many researchers have been interested in determining parts of the brain responsible for inferential processing and whether differences arise when generating different types of inferences. For example, in one study that employed a lexical decision task, researchers found that the activation of the cerebral hemispheres during inferential processing varies depending on the degree of contextual constraint (e.g., how strongly the context of the passage implies the inference to be made). Specifically, both hemispheres have been found to become active for both bridging and elaborative inferences that are strongly constrained by the surrounding text. However, processing inferences that are weakly constrained by the context, results in greater activation of the right hemisphere than the left hemisphere for both types of inferences.

Another technique for delineating the specializations of the cerebral hemispheres involves observing the resulting deficits on inferential processing in patients with sustained brain damage. Patients with damage to their right hemisphere differ in how they process and comprehend narratives relative to individuals without brain damage. They answer inference questions less accurately than explicit questions after reading a story, whereas people without brain damage answer both equally well. In addition, damage to the right hemisphere is associated with slower responding to inference-related words compared to unrelated words, whereas people without brain damage show the opposite pattern. This suggests that such patients are unable to activate the relevant semantic information necessary to draw appropriate inferences.