Steyvers & Hemmer (2012): Reconstruction from memory in naturalistic environments

Steyvers & Hemmer (2012): Reconstruction from memory in naturalistic environments

1. Aim

   • To investigate the interaction between episodic and prior knowledge in naturalistic environments
     • They wanted to see how prior knowledge was used to reconstruct memory for photographs of normal everyday settings e.g. hotel

2. Procedure

   • Ppts recruited using a random sample from an experimental ppt pool at the University of California, Irvine

3. Main Experiment

   • Using the same experimental pool, 49 ppts who didn't do prior testing were selected
   • 10 stimulus pics from prior testing were chosen
     • 2 from each scene that had elicited the most objects named in the perception test
       • From these 2 sets of 5 images, 2 from each scene were formed
         • Ppts only viewed 1 image of each scene to avoid carry-over effects
   • Ppts shown the 5 images for either 2 (prior knowledge) or 10 (episodic) seconds, to control for exposure duration.
     • 4 possible time trial orderings and ppts were randomly allocated to one
       • Free recall in their own time
         • Correct recall of inconsistent objects only come from episodic memory
         • Recall of objects missing = semantic
     • 2 seconds = relies on prior knowledge
     • 10 seconds = relies on episodic memory

4. Findings

   • The effect of prior knowledge (semantic)
     • Compared number of objects guessed in expectations test to those actually recalled in 2 experimental conditions (2 and 10 seconds)
       • Accuracy of object guesses based on the expectation test was over 55% from semantic memory
         • The actual recall in both experimental conditions was over 80%
           • In the 2 second condition, 7 objects were recalled on average
           • In the 10 second condition, 9 objects were recalled on average
             • Suggests episodic memory played a significant role in recall
   • Incorrect recall of high frequency objects was 9%
     • Suggests that with naturalistic/ unmanipulated scenes, memory is accurate
       • In unnatural scenes, the error rate increased to 19%
   • Incorrect recall of low frequency objects was 18%

5. Conclusions

   • In recall of naturalistic scenes, prior knowledge drawn from semantic memory can contribute to accurate recall in episodic memory tasks, when such scenes are unmanipulated
     • We draw on general knowledge as good guesses of what is expected to be seen
   • Prior knowledge contributes greatly to the recall of naturalistic environments but not at the expense of accuracy
     • We're more likely to notice novel items
   • Prior knowledge tends to be unrepresentative of everyday event recall; removing high frequency objects induces false memory - makes memory seem unreliable
   • With naturalistic, unmanipulated context, guesses can be effective due to the high frequency objects present
     • This frees up cognitive resources to be better spent focusing on novel and unexpected objects in a scene
       • Recall of consistent and inconsistent objects is benefited using a more ecologically valid approach

6. Evaluation

   • Weaknesses
     • culture bias
       • California
     • University students used
       • age bias
     • Difficult to apply
     • Despite naturalistic scenes
       • Artificial lab experiment
         • Low ecological validity
     • Demand characteristics
     • Individual differences
       • Prior tasks
     • Ethics
       • lack of protection from psychological harm
         • Embarrasment
   • Strengths
     • No gender bias
     • Highly controlled, standardised procedure (photos easily resused)
     • Application: used to help police with eyewitness testimony
     • Lab experiment - good internal validity
     • Different time conditions for each ppt removes bias
       • order effects
     • Ethics
       • right to withdraw
       • consent/ debrief
       • limited harm

7. Prior Testing

   • Part 1
     • 22 ppts required to list objects that they would expect to find in 5 naturalistic scenes (office, kitchen, hotel, urban, dining)
       • Important to assess prior knowledge and what they would expect to find in naturalistic scenes (control group)
     • Enter their responses into a computer for at least 1 minute per scene
   • Part 2
     • A separate group of 25 ppts shown 25 images of the 5 scenes
     • Asked to name all the objects they could see
       • A measure of perception
   • Results
     • Objects were placed into high frequency and low frequency (most and least recalled)
     • Expectations of high frequency objects tended to be iconic objects e.g.TV in a hotel room