Memory Retrieval; Semantic Memory and Language Comprehension Flashcards
What is a levels of processing theory? What are the 4 assumptions it makes?
Theory that the level the info was processed at will determine later memory for that info.
1. Memory = series of analyses, each deeper than the previous one
2. Deeper level = more durable memory
3. Rehearsal in only important to the extent it deepens the level of processing — Elaborative rehearsal (Type 2)
4. Control over processing in research — Incidental learning procedure (intent to learn is unimportant)
What are the problems with levels of processing theory?
Circularity and focus only on encoding, not retrieval (omission)
What is transfer appropriate processing?
Theory that emphasises the appropriateness of study for the test, not the level of depth -> focus on retrieval
What is the relationship between distinctiveness and organisation in encoded items?
Distinctiveness is for item-specific processing, it makes the item different from the competing items
Organisation is for relational processing
Both are important because they represent different types of processing
What does encoding specificity principle say? What theory of processing does it relate to?
It says that recollection is an interaction btw the properties of encoded event and retrieval info. Such interaction is called ecphory. Relates to transfer appropriate precessing, since retrieval info must be appropriate to the encoded info.
What are context effects? Which principle does it relate to?
Performance gets worse when the context changes from study to test. Relates to encoding specificity principle.
What is a way to avoid/eliminate context effects?
Mentally reinstate the context at study
When are the context effects the strongest?
During recall and for conceptually-driven data
What are state-dependent memory effects?
Similar to context effects. When a state of a person (mood, intoxication) changes from study to test, memory performance suffers
What is “recognition failure of recallable words paradigm?
- Study a list of weak associations (ball — spoon)
- Generate as much associations to one word as possible (bench)
- Asked if spoon was in the study list
- Asked to recall the association for ball —
Most ppl fail to recognise that spoon was in the study list (3.), but can easily recall it with a weak cue -> word is not always the best cue for itself
What is “Brown-Peterson paradigm”?
- 3 random letters
- Count backwards
- Asked to recall the letters
As trials go on, ppl find it hard to recall the letters
What are the 2 explanations of the effect found in Brown-Peterson paradigm?
- Decay in STM -> not supported
- Proactive interference -> retrieval effect
How did they show that release from PI happens at retrieval not encoding?
There was a stimulus category change on trial 4 in Brown-Peterson paradigm. One group was informed about the change before the trial, while other group wasn’t. Both groups experienced release from PI -> even the group that wasn’t informed about the category change -> can’t be due to encoding
What is part-set cuing effect?
Giving a few items at free-recall task, worsens the recall -> disrupts the natural retrieval strategy
What id word-frequency effect?
Low frequency words are better recognised (opposite of recall)
What is the dual-processes theory of recognition?
It states that retrieval is based on 2 independent processes:
1. Familiarity-based recognition — fast, automatic, perception of memory strength, “know”, perirhinal cortex
2. Recollection — slower, attention-demanding, contextual info about memory, “remember”, hippocampus
What is the SAM model?
Search of associative memory — cue-dependent probabilistic search theory of retrieval -> operated within retrieval structure based on associative network
What happens if short-term store in SAM?
Incoming info is assembled in buffer rehearsal system — words join until the buffer size is reached, each new word replaces the old one with probability 1/r
Probe cues that probe LTM are also assembled there
The info about recovered images in evaluated
What type of info is stored in long-term store (SAM)?
Word-Context info — association, time a word spends in rehearsal buffer
Word-Word info —association, time both words spend in a buffer together
What is the order of long-term retrieval in SAM?
- Probe cues (probe set) are assembled in STS
- LTS is probed
- Image is sampled with sampling probability
- Image is recovered with recovery probability
- Info is evaluated and used
What is a retrieval structure in SAM?
Matrix of strengths (context to word image; word to word image)
What is the formula for sampling rule in SAM?
Ps (Wis|Ct) = [S(Ct; Wis)]/[sum S(Ct; Wis)]
What is the formula for recovery rule in SAM?
Pr (Wi|Ct) = 1 - e^-S(Ct; Wis)
What is incrementing in SAM?
Increase in strengths after successful recovery (adds parameter to strength)
What is stopping rule in SAM?
When total attempts to recover fail (after reaching Kmax vaule)
What is a stopping rule for word cues in SAM?
When a certain probe set fails to recover required image after reaching Lmax -> dropped from STS and context alone is used
When does rechecking happen in SAM?
When Kmax is reached, subject uses each recalled word + context until Lmax to perform thorough search if smth was missed
What is forgetting according to SAM?
Retrieval failure
What causes forgetting in the Brown-Peterson paradigm? Is the locus of forgetting in encoding or
retrieval?
Proactive interference due to retrieval effects
Does deep processing always result in better memory performance than shallow processing?
Not always, processing needs to be appropriate for the test
Explain the retrieval-induced forgetting paradigm.
- Subjects study word lists -> cue-target associations
- One association is practiced (strengthened by retrieval)
- Other words associated with the same cue get inhibited (also with independent probes)
It is cue-independent
Finding similarities btw all expemplars increases RIF
Finding similarities btw competitors and integrating category members decreases RIF
Inteval btw retrieval and practice eliminates RIF
Explain the think/no-think paradigm.
- Subjects study list of words -> cue-target associations
- Asked to recall some associations, while for others asked not to recall (not even enter awareness)
- Final test -> “think” words remembered better than baseline; “no-think” words remebered worse than baseline (also with independent probes)
Cue-independent
According to the logic of Anderson and colleagues, the results obtained with independent probes cannot be accounted for by non-inhibitory theories. Explain.
If it wasn’t bc of inhibition, targets wouldn’t be recalled less with independent probes (no cross-category inhibitions), which is not the case! Item is weakened by retrieval of another item and stays inhibited for some time
What other explanations of RIF exist (apart from inhibition)?
Interference (target item is trengthened -> overpowers other items)
Links to competing items is damaged
What is the total-control effect in “Think/No-Think” paradogm? What are the subcomponents?
Intention to control retrieval influences later memory
Positive control — facilitation of “think”
Negative control — inhibition of “no-think”
What brain areas are more active during suppression in “Think/No-Think” paradogm?
Lateral PFC, ACC, preSMA, PMDr, IPS
What brain areas are less active during suppression in “Think/No-Think” paradigm?
Hippocampus, forntal polar cortex, amygdala, posterior insula, left parietal cortex, cuneus
What is the pattern of neural activations during encoding of forgotten and remembered items in “Think/No-Think” paradigm?
In “think” condition remembered items lead to more brain activation than forgotten items
In no-think” condition forgotten items lead to more brain activation than remembered items
What is a mental lexicon?
It contains orthographyc, semanic and phinological info about all known words -> mental dictionary
What does it mean when a letter maps onto an abstract letter identity?
It means that the letter can be recognised independently from how it looks (surface representations)
Letter analysis
What is an interactive-activation model of visual word recognition? Structure, word representations and frequency, coding, lexical decision, priming?
This model posits that each word locally represents one unit in an interconnected network — Orthographic Lexicon
It has 3 layers: features -> letters -> orthographic/word, and info can flow both ways (bottom-up and top-down). In each level, units compete and activated units inhibit weaker units.
More frequent words have a higher resting activation -> recognised quicker.
Letter order is coded through slot-based coding — every letter in every slot
This model can explain the word superiority effect (and weakly pseudoword superiority effect) — letters are recognised faster within a word -> top-down influences
And can explain the performance in lexical decision task — if a stimulus is present in the orthographic lexicon, it is a word
IAM also explains the Neighbourhood size effect: High-N words are recognised faster than low-N words -> all get activated due to competition. High-N non-words are rejected slower than low-N non-words -> all Ns get activates and interfere with recognition
In masked form priming, word-primes would inhibit the target -> meaning will interfere and prime the wrong thing. Non-word primes will facilitate the target -> look similar to the target (+ don’t have a meaning that could interfere)
What is a distributed-connectionist model of visual word recognition?
This model postis that words are represented globally in the brain -> a unit for every feature. These units are interconnected in a network. There is no orthographic lexicon.
Order of letters is coded through wickelcoding — triplets of letters
This model can’t explain the performance in the lexical decision task
But can explain word and pseudoword superiority effect.
What is a search model of visual word recognition? Masked form priming?
It postits that to recognise a word, one searches through their mental lexicon in serial with more frequent words being accessed first.
In masked-form priming, a prime that is visually or morphologically similar to the target would facilitate target recognition. While longer addition neighbours (not morphological) would inhibit it
What is the dual-route cascaded (DRC) model of visual word recognition? What is the structure?
This model posits that semantics and phonology also affect word recognition (orthography). The structure of this model is similar to the interactive-activation model, but it has 2 additional pathways (3 in total):
1. Lexical pathway: print -> features -> letters -> words -> phonology -> phoneme -> speech
2. Second lexical pathway (for familiar words): print -> features -> letters -> words -> semantics -> phonology -> phonemes -> speech
3. Non-lexical pathway (for unknown words): print -> features -> letters -> rule-based translation -> phonemes -> speech
Info flows through all 3 pathways, but semantic and phonological info is not necessary for visual word recognition
Rastle (2007) discusses some of the most important empirical findings of the word recognition literature. What are these findings?
Abstract letter representations, word superiority effect, pseudoword superiority effect, masked form priming, neighbourhood effects, morpho-orthographic segmentation, phonological priming
What is a property-verification task?
- Pairs of words for property verification (from different modalities)
- RTs measured
Longer RTs when modality switches for the next word pair
What did Pecher et al. (2003) do in their Experiment 1 and what did they find?
Property verification task with different SOAs (0 and 260ms)
Found slower RTs for modality-switch trials, no effect of SOA
SOA 260 ms faster in general
What did Pecher et al. (2003) do in their Experiment 2 and what did they find?
Here they wanted to rule out the associative priming hypothesis (properties within the same modality are highly associated and, therfore, the first property primes the second, leading to faster RTs). To do so they made a comparison btw highly associated word pairs and not associated word pairs (modality-independent). If the associative priming hypothesis was true, highly associated (modality-independent) words whold lead to faster RTs. THIS WASN’T THE CASE!
They still replicated the experiment 1 tho -> slower RTs are only due to modality switch, not weak associations
What is a picture-sentence verification task?
- Read a sentence with implied object orientation
- See an object with a certain orientation (orientation match/mismatch to the sentence)
- Decide if the object was mentioned in the sentence
Faster RTs when orientation in the picture mathces the orientation in the sentence -> ppl mentally simulate written material
What is special about van Zuijlen et al. (2024) study about color match effect that is different from all the previous studies in the topic?
They controlled for the correlation btw the presence of a color-match and the required response.
In previous studies, if a participant realised that the color of an object in the picture matches the implied colour of an object in the sentence, that would mean that the object was mentioned in the sentence -> this could be a potential strategy participants develop, not automatic mental simulation.
In this study, they added filler trials where the object in the picture WAS NOT mentioned in the sentence but the implied colour in the sentence and the colour of the object in the picture matched. Particitants were still correct in identifying that the object was NOT mentioned in the sentence -> rules out the possibility that subjects develop a strategy.
Subjects were faster and made less errors on colour match trials (when the objects was actually mentioned in the sentence) meaning that colour simulations are automatic during language comprehension
Van Zuijlen et al. (2024) studied sentence comprehension in non-native speakers. Do their results suggest that language understanding of native and non-native speakers are based on qualitatively different types of knowledge? Explain.
No, they said that lexical representations of native and non-native languages are connected to shared conceptual knowledge system.
They demonstrated it by testing non-native english speakers on picture-sentence verification task and showing that the effect size for the colour match was as strong as in a native sample -> automatic color simulations are present in both groups
What are the 3 types of information used to represent conceptual knowledge (Fernandino et al., 2022)?
Taxonomic — hierarchical interconnected categories: superordinate -> basic -> subordinate (amodal)
Experiential — encodes experiences that lead to the formation of a certain concept (grounded cognition theories)
Distributional — statistical patterns of co-occurrence btw concepts (semantics + context)
What was the design of Fernandino et al. (2022) study?
- Paricipants judged familiarity of words under fMRI
- Whole-brain activations were mapped for each concept
- Representational Dissimilarity Matrix (RDM) was made for each concept (fMRI data and model predictions)
- Those neural activations (matrices) were compared to the activations (matrices) predicted by each model -> Representational Similarity Analysis (RSA)
The more similar the model predictions to actual fMRI data, the better is the model at describing concept representations
Which representational code is used by the brain according to Fernandino et al. (2022)? What was the conclusion?
Experienctial Information
Exp48 mathed fMRI data the best (then SM8, then Categorical)
Experiential info is very important for the formation of concepts in humans
