Week 8 Flashcards
(69 cards)
generic memory/semantic memory (or the memory for meaning)
memory for facts, concepts, and meanings that is context free and not associated with a particular point in time
knowledge of the name of the capital city of Canada, the meaning of words you know, and what the characteristics of a cat are
not all knowledge is based on meaning. Indeed, you can know how to say ”antidisestablishmentarianism” (once the longest word in the English language) and spell it without having the slightest clue what it means; this shows that you can have knowledge without meaning even when that knowledge is for a word. - why we will use generic memory
better represents the scope of information stored and retrieved for the purposes of everyday life. Generic memory is a type of explicit memory.
concept
a general idea derived or inferred from specific instances or occurrences
Concepts can often be labelled with words but concepts themselves aren’t words; they are abstractions
Generic memory is comprised of
concepts.
Unlike episodic and autobiographical memory, generic memory is
context free
where and when the concept was learned is not linked to the generic memory.
evidence supports the notion that generic memory is distinct from episodic memory
1 - people who acquire retrograde amnesia due to damage to the medial temporal lobe exhibit loss of episodic memories going back several years, but only exhibit loss of generic memories acquired shortly before the onset of amnesia
—-episodic memories are linked to the medial temporal lobe while generic memories are not.
2 - people who develop semantic dementia due to damage in the anterior frontal lobe exhibit generic memory loss while their episodic memory remains intact
—–suggests that the anterior frontal lobe is linked to generic memories but not to episodic memories
The Hierarchical-Network Model
Quillian was developing a computer program to read text and needed the program to “know” facts in order to “understand” text.
conserve space on the hard drive, so he wrote a program that arranged concepts efficiently in a hierarchy.
hierarchical-network model-
a model of generic memory in which concepts are organized from the most general at the top to the most specific at the bottom, with facts about a concept attached to the highest level of the hierarchy to which they apply
clear predictions about how long it will take for a person to verify different types of sentences using a sentence-verification task.
travelling through the hierarchy takes time, that more travelling through more levels takes more time, and that it will take about the same amount of time to move around the network no matter what level you start at
sentence-verification task
a paradigm in which participants are presented with statements such as “A bird can fly” or “A fish eats rocks” and must indicate as quickly and accurately as possible whether the sentence is true or false
The Hierarchical-Network Model - problems
1 - the hierarchical model does not really explain how participants are able to respond to a sentence as false; for example, consider the sentence “Salmon have fur.” Participants quickly respond that this sentence is false; however, the model seems to predict that participants would need to consult the entire hierarchy before establishing a property was absent.
2- the hierarchical model predicts that sentence verification will be related to a concept’s hierarchy; however, Carol Conrad found that the frequency of a property, not it’s level on the hierarchy, predicted verification times.
3- the hierarchical model also predicts that the further a person has to “move” through a hierarchy, the slower she or he will be, but Lance Rips, Edward Shoben, and Edward Smith demonstrated that participants are faster to verify “A pig is an animal” than “A pig is a mammal,” even though a strictly hierarchical model would predict the opposite (as “mammal” is a subset of “animal”
Finally, Michael Posner and Steve Keele showed that typical members of a category are recognized more quickly than atypical category members, a phenomenon known as the typicality effect.
The Spreading-Activation Model
spreading activation model
×
a model of the relationship among concepts in memory based on the ideas that related concepts are connected in a network, and that concepts that share more properties will have more links between them
Activation is assumed to be limited, so when there are more paths, less activation goes to each path
When a concept has been activated by another concept, less additional activation is needed for that concept to reach a decision criterion.
decision criterion
a threshold of activation that must be met in order for an item to be retrieved from generic memory
spreading-activation model can account for the findings that were problematic for the hierarchical-network model
1- the model can explain why responses to false statements such as “A salmon has fur” are made relatively quickly—the model assumes that no activation will spread from salmon to fur
2- the model can account for the fact that higher-frequency features such as “A salmon is a fish” are recognized more quickly than lower-frequency features such as “A salmon has skin” by assuming that connections between concepts and high-frequency features are closer than connections between concepts and low-frequency features.
3- the model can account for findings that seem to contradict a hierarchical structure, such as participants verifying “A pig is an animal” more quickly than “A pig is a mammal,” if it is assumed that there is a closer connection between pig and animal than pig and mammal.
Finally, spreading activation can account for the typicality effect by assuming that typical members of a category are more closely associated to a category than less typical members.
Spreading-activation models were also designed to account for
associative priming, a phenomenon where a person recognizes an item more quickly when it follows a related concept than when it does not.
Associative-priming effects support the spreading-activation model because they are consistent with the premise that activation spreads from a word to related nodes adding activation, thus reducing the amount of additional activation needed to reach the decision criterion, predicting that less activation (and time) will be needed to recognize “nurse” when it is preceded by “doctor.”
“A robin is a bird” is recognized more quickly than “A robin is an animal” because the concepts of “robin” and “bird” share a closer association in the network than “robin” and “animal” do. Literally thousands of experiments have demonstrated associative priming.
spreading-activation model is also supported by research showing greater associative priming for words that are closely related semantically than for those that are less closely related semantically.
Rosa Sáchez-Casas and colleagues assembled a list of 80 words from a variety of categories and then linked each of these target words with four possible primes: the word itself (e.g., donkey–DONKEY), a prime that was very similar in meaning to the original word (e.g., horse–DONKEY), a prime from the same semantic category but that was less closely related (e.g., bear–DONKEY) and a prime that was unrelated (e.g., thimble–DONKEY).
Participants completed a lexical decision task during which they had to indicate as quickly and accurately as possible, using the keyboard, if a letter string that was presented was a word such as DONKEY or a nonword such as HERLOP. Each participant saw each target word only once.
participants made a lexical decision for a target word (e.g., DONKEY) significantly faster when the prime was very closely related (e.g., horse–DONKEY) than when the prime was just closely related (e.g., bear–DONKEY).
onducted an additional experiment with the same word stimuli, but in this experiment participants saw the prime and target together (e.g., horse–DONKEY) and had to indicate if the word appearing in uppercase letters (e.g., DONKEY) was a concrete noun or not by pressing a key on the keyboard as quickly and accurately as possible. Once again, the participants responded more quickly when the prime and target were very closely related (e.g., horse–DONKEY) than when they were just closely related (e.g., bear–DONKEY)
weaknesses with the spreading-activation account
1- it is very difficult to disprove the spreading-activation model; how do you show that spreading activation is not taking place?
2- second weakness of the spreading-activation model is that it cannot explain the mediated priming effect and therefore can’t give a good account of memory retrieval. Mediated priming is the finding that concepts that aren’t directly linked to one another can still sometimes prime one another.
————- McNamara (1992) showed that the word mane can prime the word stripe even though the association between those words, according to the spreading-activation model, is quite remote (mane-lion-tiger-stripe). Mediated priming is problematic because in order for activation to spread from mane to stripe, thousands and thousands of other words would need to be activated in the system, which predicts that most words should prime most other words if there is even a remote association between them; but research has shown this is not the case.
——————McNamara (1992) argues that each word has about 20 other words associated with it. Thus, McNamara argues, if mane were activated, 20 other words would also be activated, including lion.
—————for lion to activate tiger, 20 words associated with lion would need to be activated along with 20 words for each of the other words activated by mane; thus at this point 420 words would be active.
————— If an average person has a vocabulary of 30,000 words, this would mean that one-third of that vocabulary would be activated by mane, which McNamara (1992) argues would be highly inefficient. In addition, McNamara (1992) argues that, according to spreading activation, mane should not just prime stripe but also all words that are removed from mane by four or fewer degrees of association, such as scissors (mane-hair-haircut-scissors); however, mane is not likely to prime scissors. In fact, priming is a relatively rare occurrence. Even when participants are trained to associate two words through repeated exposure, priming effects aren’t usually observed until about five weeks after learning.
3- predicts that priming effects will be very short in duration; after additional stimuli have been presented, new concepts will be active, and old items shouldn’t produce priming. However, Joordens and Becker (1997) showed that semantic priming could occur over as many as eight intervening items.
4- model assumes a single fixed representation for each concept in a network. Thus, the model predicts that a person reading the sentence “Amy painted the egg” and “Amy ate the egg” would activate the same nodes related to “egg.” However, processing of the first sentence would likely focus on the exterior of the egg, while processing of the second sentence would focus on the interior of the egg, which are different concepts
5-represents concepts as individual nodes, implying localized representations in generic memory; however, there is substantial evidence in the literature dating back to the work of Lashley in the 1950s to suggest that generic memories are represented by patterns of activation distributed across neurons
compound-cue model
a model that assumes recognition of an item is based on a measure of familiarity determined by the compounded effects of a number of cues including context, cues that result from sharing associations with items that have recently been presented, and cues from associations in memory
According to the compound-cue model, there are three types of cues: the context within which the memory was learned (called context cues),
the items that were present when the item was learned (called inter-item cues),
and a sense of familiarity represented by an item being a cue for itself (called self–self cues)
Generic memory is more likely to be accessed when more cues for an item are in
short-term memory; cues have a compounding effect
compound-cue model can account for some of the phenomena that could not be explained by the spreading-activation model.
can account for mediated priming. The compound-cue model assumes that mane and stripe have similar associations in memory, such as large cat, claws, danger, and cubs and that these associations are similar enough to significantly affect the familiarity of stripe when it follows the word mane, resulting in priming.
explain why priming can sometimes occur when there are several items between the prime and target if it is assumed that the prime enters short-term memory and stays there for several seconds.
the compound-cue model can explain why “Amy painted the egg” and “Amy ate the egg” bring to mind very different aspects of the concept “egg” if it is assumed painted and egg work together as inter-item cues for the concept of an eggshell, while ate and egg would work together as inter-item cues for the concept of an egg’s edible interior
the compound-cue model is consistent with the notion that concepts may be represented by a pattern of activation rather than by a single node.
two lines of evidence suggest that various features of a concept are also integrated into a single supramodal concept
First, there is the fact that people can identify similarities in concepts that have very different perceptual or action-related features. For example, a hosta and a cactus have very different perceptual features but are both easily linked together as plants;
———-concepts don’t rely entirely on perceptual or motor representations.
Second, individuals with semantic dementia, which is caused by degeneration in the anterior temporal lobes, often lose knowledge of multiple features of a given concept; for example, they may not be able to identify a lemon by sight or by smell. This suggests that the frontotemporal area of the brain is involved in integrating multiple features of a given concept.
hub-and-spoke model of generic memory
a model by Pobric et al. (2010), which proposes that six types of modality-specific representations all meet at a central hub in the anterior temporal lobe
used repetitive transcranial magnetic stimulation (rTMS) to test the hub-and-spoke model. When a brain region is stimulated with rTMS, it loses its functionality, and thus rTMS can be used to determine which specific brain regions are necessary for the completion of specific tasks.
Stimulating the ATL disrupted naming of all living and nonliving objects, supporting the notion that the ATL is linked to supramodal conceptual representations. Stimulating the IPL, on the other hand, only disrupted naming of nonliving objects that could be manipulated by hand, supporting the notion that concepts are represented in modality-specific regions of the brain as well.
One piece of evidence supporting the notion of a “hub” where the modality-specific features of concepts are integrated is that _
we can recognize that certain items belong to the same concepts even though they do not share any perceptual similarities
generic memories are initially episodic memories stored in the
hippocampus.
The first time you handled the coin, you may have noticed the distinct profile of a loon embossed on one side. Soon after, your parent may have asked you, “Why do you think this coin is called a loonie?” At this point you may have referred back to your episodic memory of seeing the loon on the coin and answered, “Because there is a loon on it.” Over the coming months you would encounter the coin many more times
two-stage model of memory suggests that, over time, features that overlap across episodic memories stored in the hippocampus are transferred to the neocortex - happened with loonie knowledge
Dagenbach, Horst, and Carr (1990) used associative priming to test the time course of the transfer of information from episodic to generic memory.
argued that associative priming is only observed for items that are part of generic memory; episodic memories do not produce priming.
learning novel nonword–word pairs such as DRUPE–CHERRY. After five weeks of intense study, DRUPE facilitated processing CHERRY.
the priming effect indicated that DRUPE had become part of generic memory and that, by extension, it takes at least five weeks for episodic information to become a generic memory.
Patterns are extracted from episodic information automatically. For example, you have probably never been told what makes a specific piece of furniture a desk
Anthony Wagner and his colleagues used fMRI data to study the brain regions involved in the retrieval of generic memories
presented participants undergoing an fMRI with a cue word (such as candle) above multiple target words. The target words either had a strong association (such as flame) or a weak association (such as halo) with the cue. Participants were asked to indicate which target best matched the cue
ctivation in the left inferior prefrontal cortex (LIPC) was significantly greater in the weak associative strength condition than the strong associative strength condition and concluded that the left inferior prefrontal cortex is involved with controlled searches of generic memory, such as those involved with determining whether the word candle is more similar to the word flame or the word halo
argue that some information in generic memory is retrieved automatically, such as the fact that candles produce flames. Other information, such as whether candles are more similar to a person’s memory for the concept “flame” or the concept “halo,” requires effort to retrieve, and effortful, controlled retrieval involves the LIPC.
