1 / 17

Sentence Memory: A Constructive Versus Interpretive Approach

Sentence Memory: A Constructive Versus Interpretive Approach. Bransford, J.D., Barclay, J.R., & Franks, J.J. “Under normal circumstances a listener’s memory for sentences may be inaccurate at the word-for-word recall, but accurate at the level of semantic paraphrase.” (Sachs, 1967).

chet
Download Presentation

Sentence Memory: A Constructive Versus Interpretive Approach

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Sentence Memory: A Constructive Versus Interpretive Approach Bransford, J.D., Barclay, J.R., & Franks, J.J.

  2. “Under normal circumstances a listener’s memory for sentences may be inaccurate at the word-for-word recall, but accurate at the level of semantic paraphrase.” (Sachs, 1967)

  3. Theories Interpretive:The deep structure of a sentence represents the input of the semantic component of the grammar and is assumed to provide a sufficient characterization of what is stored Constructive: Sentences are viewed as information which can be used to construct semantic descriptions of relations -- sentences may contain more information than is represented in the linguistic input

  4. In other words… Interpretive Theory: Information is contained in the structure itself Constructive Theory: Relations may be inferred that may not have been specified (1) Three turtles rested beside a log, and a fish swam beneath them. *Includes information that the fish swam beneath the turtles… (2) Three turtles rested on a log, and a fish swam beneath them. *The same information is given here, but from this sentence, we can infer that the fish also swam beneath the log, even though it was not explicitly stated.

  5. Different Predictions… Predictions about the probability that a subject will think they heard a sentence before: Interpretive: Subjects store only the linguistic information underlying the input sentence. So, in hearing either sentence (1) or (2), subjects will be equally as likely to detect a pronoun change in sentence (3). (3) Three turtles rested (beside/on) a floating log and a fish swam beneath it. Constructive: Subjects are assumed to construct wholistic semantic descriptions of situations. Therefore, subjects hearing sentence (1) should still reject sentence (3), since it doesn’t fit with the actual input or the complete semantic description constructed. Subjects hearing sentence (2), however, should be quite likely to think they heard sentence (3) because it fits with the complete semantic description

  6. Experiment 1 Purpose: To contrast the interpretive and constructive theories by comparing recognition memory for sentences which have identical deep structures but differ in the semantic descriptions suggested. Procedure: • Acquisition: Subjects were instructed to listen carefully to a set of sentences that would be read to them (they would be answering questions about the sentences later) • Recognition: After a 3 minute break, subjects were asked to indicate which exact sentences they had heard, and which they had not • Subjects were also asked to rate their confidence in each response

  7. Experiment 1 - Results Recognition mean was computed for each of 6 categories of sentences: 1) Potential inference sentences presented during acquisition (OLD-PI) 2) Potential inference sentences not presented during acquisition (NEW-PI) 3) Noninference sentences presented during acquisition (OLD-NI) 4) Noninference sentences not presented during acquisition (NEW-NI) 5) Filler sentences produced during acquisition (OLD-F) 6) Filler sentences not produced during acquisition (NEW-F)

  8. Experiment 1 - Results • No difference was found between OLD-PI and NEW-PI • Significant difference found between OLD-NI and NEW-NI • NI sentences are not consistent with the same mental model, so it is easier for subjects to remember the difference between them • PI sentences are consistent with the same mental model, so it is harder for subjects to remember the difference between them • The results indicate that subjects were indeed using a mental model or representation to remember these sentences, which supports the constructive theory

  9. Discussion Questions • In Exp. 1, after hearing all those sentences, woudn’t the subjects be less likely to remember the ones they’d heard first and more likely to remember the later ones, regardless of Old/New PI and NI differences? • During the 3 minutes breaks before the recognition task, what were the subjects allowed to do? Were they told to sit quietly and think about the sentences and their meanings, or could the chat with the experimentor if they wanted? • I'm not sure how the data in table are to be interpreted. As for F sentence type, OLDS (2.19), NEWS (-4.15), does it mean subjects recognized the NEWS sentences more confidently than the OLDS one?

  10. Experiment 2 Purpose: To hone in on the part of the sentence that subjects were confusing Procedure: • Acquisition: Same as Experiment 1 • Recall: Subjects were read sentence props and asked to recall the rest of the sentence

  11. Experiment 2 - Results • Sentence recall scores were computed without regard for accuracy of recall of the last pronoun • Pronoun recall scores were computed for sentences that were otherwise correctly recalled • % correct pronoun recall = 57% • In 10 out of 14 sentence frames, PI sentences had a lower probability of accurate pronoun recall than NI sentences • Showed that memory differences between PI and NI sentences were specific to memory of the specific pronoun form • Results of Experiment 1 were replicated, supporting the constructive theory

  12. Discussion Question • Might subject's responses in Bransford et al's experiment 2 be a partially a function of their ability to construct spatial relation structures? Individuals differ greatly in terms of spatial relation abilites, and there are generally consistent gender differences as well.

  13. Experiment 3 Purpose: To further test the results of Experiments 1 & 2 Procedure: • Acquisition: Subjects listened to descriptive passages • Ex) There is a tree with a box beside it , and a chair is on top of the box. The box is to the right of the tree. The tree is green and extremely tall. • Recognition: Subjects were given a sheet with blocks of sentences - from each block they had to indicate which sentence they had actually heard Recognition set A: The box is to the right of the tree. The chair is to the right of the tree. The box is to the left of the tree. The chair is to the left of the tree. Recognition set B: The tree is to the left of the box. The tree is to the left of the chair. The tree is to the right of the box. The tree is to the right of the chair.

  14. Experiment 3 - Results • Recognition set A: 71% of responses consonant with the complete semantic description suggested by the passages • Recognition set B: 70% of responses were situation preserving • If subjects were simply storing information about where objects were described, they should have been very confused by set B - but they were not • Again, this is evidence for the constructive theory

  15. Discussion Questions: • Can you go over experiment 3? I’m a little confused by the stimuli they used. Later in the results they say that the OLDS were recognized more often, but a sentence later they say subjects did not pick the OLD, which seems contradictory. Can you explain this? • In Exp. 3, the subjects who received recognition set B did not actually see any of the sentences during the recognition portion of the experiment that they had heard during the acquisition portion. Wouldn’t this have become obvious to those subjects and influenced their responses?

  16. Overall Conclusions • Recognition is primarily a function of the complete semantic descriptions constructed rather than a function of just the information specified by the linguistic input strings • All experiments support the constructive theory of semantics

  17. Discussion Questions • Are people who have something like “photographic” memory different from those who remember ideas they have read or heard by building a mental model, rather than remembering exactly what the text looked like or the sound sounded like? • How much of a difference is there between the intuitive leaps that the subjects of this study are making with respect to inferring information not found in the linguistic input, and the intuitive leaps made by the subjects of the study by Bransford and Franks (1971) mentioned in Haviland and Clark (1974), in which subjects believed they had heard more details in a scene than they had actually been given? • There are a number of disorders that affect people's ability to comprehend language. Is it also the case that these disorders affect their ability to build mental models and commit certain things to memory?

More Related