L9 Long Term Memory: Encoding & Retrieval (PSYC2050)

Long Term Memory: Encoding & Retrieval

• Lecture 9 covers long term memory, specifically encoding and retrieval processes.

Last Week's Concepts

• Long-Term Memory (LTM) was defined.
• Semantic versus Episodic memory was covered.
• Characteristics and implications of LTM were discussed.
• Short-Term Memory (STM) also known as Working Memory (WM) was compared to LTM.
• Key findings and the Modal Model were reviewed.
• Tasks for testing LTM such as Recall and Recognition were discussed.
• Issues in signal detection versus bias were examined.
• Implicit versus Explicit LTM tasks were addressed.
• Forgetting curves, decay versus interference, and inhibition in memory were discussed.

Learning Objectives

• Understand encoding, retrieval, and their interactions.
• Review evidence about depth of processing and emotion in memory encoding.
• Understand evidence about retrieval processes and retrieval cues.
• Discuss evidence for the Transfer Appropriate Processing framework versus memory systems.
• Summarize the benefits of repeated testing and suggest applications of memory research to learning at university.

This Week's Focus

• Encoding: Getting material into memory.
• Retrieval: Getting material out again.
• Encoding-retrieval interactions.

Encoding and Retrieval

• Studying words involves identifying them via visual word identification.
• The study task involves encoding the word into memory including contextual information.
• At test, cues are provided by the experimenter, focusing on the study list.
• Self-generated cues can also be used, such as thinking of words rated as pleasant.
• Retrieval processes may differ for recall and recognition tasks.

Encoding

• Encoding is the process of entering material into memory.
• A theory of encoding into LTM given in the modal model says information is maintained by rehearsal in STM until transferred to LTM.
• According to the modal model, memory is largely structural.
• Rehearsal does not necessarily produce good long-term memory.
• There are multiple ways to encode material in both STM and LTM.

Levels of Processing (LOP)

• When studying words, individuals can focus on orthography (spelling), phonology (pronunciation), and meaning (semantics).
• Recall of studied words is better following semantic processing compared to rehearsal (Craik & Lockhart, 1972).
• Rehearsal involves repeating the word aloud or silently, over and over.
• The levels of processing theory describes a continuum of depth, from orthographic to phonological to semantic.

LOP Effect

• Intentional learning instructions compared to three study tasks (Craik, 1977):
  • Upper/lower case (orthographic).
  • Does FROG rhyme with DOG? (phonological).
  • Is it a living thing? (semantic).

Intention to Learn

• Research questions whether intention to learn matters over and above specific study tasks (Hyde & Jenkins, 1973).
• A study task x instruction about later test was conducted:
  • Rate pleasantness (semantic) compared with letter checking (orthographic).
  • Results indicated that intention does not matter.

Evidence for LOP

• Jacoby, Craik & Begg, 1979:
  • Participants imagined an object and evaluated size differences at study, such as horse-goat versus cat-elephant.
  • Results suggest that better memory for words occurs with small size differences.
  • LOP is valid in many semantic situations with deeper semantic analysis with small size differences.

LOP: Elaboration

• Connections provide retrieval cues and paths.
• Craik & Tulving, 1975 study task examines whether CHICKEN fits in a sentence such as:
  • A: The girl cooked the _____.
  • B: The great bird swooped down and carried off the struggling _____.
• Recall of words is better for elaborate condition B where there are more connections to event memories and concepts.

Deep Processing and Learning

• Semantic and elaborate processing benefits student learning.
• Elaborative processing is associated with organization
  • Imposing order enhances memory.
  • Method of loci, other mnemonics.
  • Promotes connections.
• Chunking is applies beyond STM.
• Semantic processing promotes chunking based on meaning/structure, as demonstrated by skilled vs novice chess players and memory of piece locations.
• Understanding promotes better memory
  • Recall is improves when an interpretation is provided during encoding.
  • Provides understanding and promotes connections
  • Interpretation unifies and means fewer elements to remember

Independent Assessments of Depth

• Logical or conceptual approaches have circularity (Baddeley, 1978).
• Deep processing promotes good memory.
• Early attempts to measure direct effects of deep processing that promotes memory were unsuccessful.
• Processing time is not an index of depth (Craik & Tulving, 1975).
  • Deeper processing does not necessarily take longer.
• Processing difficulty does not improve memory and is not the critical factor
  • Difficult but superficial (orthographical) tasks did not improve memory.
  • For example: does the word WITCH match CCVCC? (C=consonant, V=vowel)

Empirical Problems with LOP

• Does meaning or distinctiveness enhance memory?
  • "Has a trunk” identifies elephant better than "contains two letter Es".
  • Semantic processing enhances distinctiveness of memories.
• The question of whether distinctive non semantic processing can produce good memory is important.
• Eysenck & Eysenck 1980
  • Distinctiveness x Semantic coding condition.
  • Recognition test
  • Words with atypical spelling-sound correspondences, e.g., love, glove.

Conditions: Eysenck & Eysenck

• Examined semantic and non-semantic conditions, distinctive and non-distinctive trials.
• An atypical descriptor like "saggy glove" is considered distinctive and semantic.
• A distinctive pronunciation, such as "glove rhymes with stove" is considered distinctive while being non-semantic.
• A typical descriptor like "leather glove" is non-distinctive and semantic.
• Correct pronunciation, such as “glove rhymes with love” is considered non-distinctive non-semantic.
• Within-Ps design - each P gets 4 blocks of study trials, one for each condition.

Results: Eysenck & Eysenck

• Distinctiveness combined with Semantic value is better than distinct pronunciation.

Distinctiveness vs. Connections

• Connections and relationships support retrieval through semantic means
  • Provides retrieval cues and paths
  • Helps access memories of the study phase.
• Distinctiveness discrimination helps to discriminate among retrieved items that are similar or related
  • Among the study-list encounter with an item versus and prior encounters.
  • Especially useful for recognition tests

Other Factors Helping Encoding

• Rhyme works with semantic cues
• The oral tradition (David Rubin)
• Rhyme and meaning work together to provide structure, integration, cues
  • E.g., Baa Baa black sheep, have you any wool?

Emotion and Encoding

• When people asked for clearest childhood memories, they name emotionally-charged events (Rubin & Kozin):
  • accidents, birthdays, etc.
• Cahill & McGaugh study:
  • Participants viewed a slide show of a hospital visit with graphic surgery slides.
  • One group told surgery was real, other that it was faked.
  • The “real” group showed better memory for the emotional material.

What Does Emotion Do?

• Increases arousal, also increasing attention.
• There may be an emotion effect beyond attention.
• Inducing emotion shortly after material will give it a memory benefit.
• Emotion is important in events involving threat, opportunity, pain, love, and friendship.
• Humans have evolved to encode emotional events better because they are important.

Flashbulb Memories

• People report vivid detailed memories from major and consequential disturbing events:
  • World Trade Center Towers, tsunamis (Indonesia and Japan).
• High confidence about “flashbulb” memories
• Many studies conclude people's confidence in “flashbulb” memories is misplaced.
  • Forgetting similar to regular memories.
  • 9/11 memories for distress change over time and align with appraisal of the attack (Levine et al.)

Retrieval Cues

• Forgetting does not necessarily mean memories have been lost.
• Decay may not be a major cause of forgetting, but inaccessibility may be.
• Tip of the tongue effect: knowing the word, but not being able to access it.
• Multiple study and recall attempts with a list of 36 words (Tulving, 1967):
  • On each recall attempt, people remember nearly 4 new words.
  • People forgot an average of 3.9 words from the last test.

Encoding Retrieval Interactions

• For the best results, memory material should be encoded in a way that improves accessibility:
  • Effective retrieval cues should be provided at test.

Encoding-Retrieval Interactions

• Early memory research focused on word lists in recall and recognition tests.
• Later, realized that effectiveness depends upon how the material will be used and how memory will be probed at test.
• Semantic encoding is useful for conceptually based tests such as recall, recognition, and many university exams.

Interactions

• Encoding processes and factors combine
• Stein, 1978, participants studied words with one capital letter
  • Group 1: does the word fit? The _____ rolled down the hill.
  • Group 2: which is the capital letter?
• They were then tested
  • A. Choose study words from distractors. Higher accuracy after sentence study.
  • B. Another group would choose word from Rock, rOck, rocK, rock. Higher accuracy after letter-task study

Stein Results

• Semantic study produces the best results.

Transfer Appropriate Processing (TAP)

• TAP Model: Superiority of semantic study reflects the fact that most memory tests are semantically based (Morris, Bransford & Franks, 1977).
• Compares a standard recognition test with:
  • A phonological test where participants choose words that sounded like the studied words.
  • Encoding is tested asking:
    • Does the word fit in a sentence (semantic) versus Does the word rhyme...?
    • Semantic study is better for recognition but not for phonological test.
• Transfer is best when test processes overlap with processing at study.

State Dependent Learning

• Environmental context is a state-dependent factor in learning.
• Reinstate as many of the study conditions as possible, situation.
• Baddeley & Godden experiment:
  • Divers studied words either 20 feet underwater, or by the water's edge then were asked to recall in water, or on land.

Baddeley & Godden

• Recalling did best in the location where they studied.

Recent Applications of Memory Research: University Learning

• Testing effects provide insight into student learning.
• Tests serve as opportunity for encoding.
  • Indirect test effects: students are driven to study more if they know there is a test.
  • Direct test effects: More is encoded on the tested material.
  • additional encoding of material encountered or retrieved during tests
  • easier retrieval of texts

Roedigger & Karpicke, 2006

• Roedigger & Karpicke reviewed evidence on direct effects of tests on learning & memory.
• Studied in Early study by Spitzer 1939:
  • 6th grade children in 91 Iowa schools studied 600 word articles on peanuts/bamboo.
  • The test was 25 multi-choice questions with varied number/timing
  • There were no additional study opportunities or feedback on tests

Spitzer - Results

• Show early group retained more in the long run.

In The Lab

• Recall of words (Roedigger & Karpicke, 2006).
• List of 40 words to memorize, 20 trials with the list.
• Varied the trails in the following way
  • Standard group (study-test alternation): STST .....
  • Repeated study: SSST .......
  • Repeated test: STTT ........
• They were assessing after one week every minute during the 10 min test.

Multi-Choice Tests

• Can be bad news.
• Possible that if students choose a familiar answer without retrieval, there may be no observed effect (Chan et al., 2006).
• But, well designed MC tests can do well
  • Little et al., 2012, when plausible distractors were given, students think about why each alternative is correct/incorrect
  • Multi-choice tests fostered learning about both correct and incorrect alternatives

Cognitive Psychology's Learning Tips

• A review by Roediger, 2013, in the US Govt report highlights effective and ineffective learning and teaching tips.
• Effective strategies:
  • Distribute practice often for shorter durations.
  • Practice retrieval through self-testing and teacher tests.
  • Interleave practice by mixing problems and topics.
  • Use elaborative interrogation to ask why something is true.
  • Engage in self-explanation.
• Not recommended strategies:
  • Re-reading text, highlighting text as study methods.
  • Summarizing text, generating static images for text-based learning, and devising key-word mnemonics.

Erasing Memories?

• A memory may be reconstructed upon retrieval.
• Beta-adrenergic receptor blockers may be used to disrupt the process (Nader et al, 2013).

Putting It All Together

• Learning
• Attention/Working memory
• Long-term memory

Putting It All Together

• Minimizing surprise, or maximizing the sensory evidence for an agent's existence
• A predictive/Bayesian brain

Contextual Cueing

• Contextual cueing: learning distractor layout and target placement pairings facilitates in the following
  • Attention
  • Implicit (recognition test at chance)
  • Typically 12 different repeats each block
  • Long-term memory

Findings

• No Contextual cueing for patients with hippocampal damage
• Results published Chun & Phelps, 1999