5. Hill-climbing and means-end problem solving are heuristic problem-solving strategies. In the hill-climbing heuristic, you simply choose the alternative that seem to lead most directly towards your goal state. The biggest drawback to this heuristic is that problem solvers must consistently choose the alternative that appears to lead most directly toward the goal. In doing so, they may fail to choose a less direct alternative which may have greater long-term benefits. Means-end analysis can help overcome this problem. In this heuristic, you divide the problem into a number of subproblems (or sub-goals), and then you try to reduce the difference between the initial state and the goal state for each of the subproblems. An example of hill-climbing is to begin studying for a final as close to the actual final itself with the goal of memorizing the most amount of information right before the test. However, you can use means-ends to realize that studying a little bit each night for a week before the test (sub goals) will be more efficient for obtaining the best score possible. Studying a little bit each night may seem like a step back – weeknights are less free for watching TV and relaxing. In the long run, the outcome will be better.

6. Solving by analogy means you find a problem with a smiliar structure for which you know the solution and you translate the solution to the new problem. The term “problem isomorphs” is used to refer to a set of problems that have the same underlying structures and solutions, but different specific details. The specific details are often the surface features – the specific objects and terms used in the problems. The deep structure is the underlying meaning and solution that is common to both problems. For example, we can use the military strategy of “divide and conquer” (used for the troops who need to enter the tower in small groups in order to overthrow the dictator) for the tumor problem (where the ray needs to enter the brain in small rays in order to destroy the tumor). The surface structure is military strategy vs. medical procedure. The deep structure is the idea of “divide and conquer.” People often make the mistake relying solely on the surface structure of the problems and not being able to solve a new problem isomorph that has similar features. People also have trouble solving the same problem in a new setting; they fail to transfer their knowledge.

7. Ill-defined problems are problems where parts of the problem space are unclear and situation assessment is a necessary and crucial part to the problem-solving. The approach of the problem solver is different with ill-defined problems than with well-defined problems (where the goal state is obvious) because the emphasis is on figuring out and representing the problem space rather than searching (which is often trivial). A great ill-defined problem is leading a happy life.

8. Experts and novices reason in different ways. Experts have a greater amount of knowledge in semantic/episodic memory, use bigger chunks, have greater pattern recognition, and their semantic memory is more highly cross-referenced. Experts organize by deep structure, while novices tend to organize by surface structure. Novices also tend to work backward from a goal while experts know the very first step and apply a well-known schema. Experts are different than computers in that their top-down processes allow them to perform well on many different components of problem solving in their area of expertise.

3. A heuristic is a general rule that is usually correct. In problem solving, a heuristic is a strategy in which you ignore some alternatives and explore only those alternatives that are most likely to produce a solution. Examples of the problem solving heuristics are hill-climbing and means-ends. An example of the hill-climbing heuristic is as follows: let’s say I need to clean the house as quickly as possible. I take blind steps to clean the house: I vacuum, dust, start the dishwasher, etc. I vacuum first, but I later realize that it would have been better in the long-run to start this dishwasher first so that I could vacuum while the dishwasher is running. An algorithm is a method that will always produce a solution to the problem, although the process can sometimes be inefficient. One example of an algorithm is exhaustive search, where you try out all possible answers using a specified system. If I had multiple, unidentified packages of different colored flower seeds, I could plant one of every single package and see which one came out red. (Another bad example, but whatever).

4. When using insight problem solving, the problem initially seems impossible to solve, but an alternative approach suddenly bursts into consciousness – immediately realizing that the new solution is correct. Say I have a bunch of flowers that seem impossible to arrange nicely within a vase – and suddenly, out of the blue, I think “Aha! The flower stems are too tall and need to be cut.” Functional fixedness is a problem for insight problems solving – it shows that we tend to use objects and concepts only in their conventional ways. For example, I may need to hang mistletoe at Christmastime but have no tape or tacks. However, there’s a water pipe hanging over my dorm room door, and that will serve just nicely as an anchor for the string attached to my mistletoe. The pipe normally only allows water flow, but now it serves as an anchor. When problem solvers have a mental set, they keep trying the same solution they have used in previous problems, even though the problem could be solved by a different, easier method. In this class, I tend to answer focus questions in their numerical order. The chronology of the focus questions is a little asymmetrical to the chronology of the concepts presented in class, but I answer them in the focus-question-order anyways because that’s what I have always done. Jeremy Jurva is the most awesome hero.

FQ 9 #1-8

1. Conditional reasoning is the logical determination of whether the evidence supports, refutes, or is irrelevant to the given relationship. It usually involves an “if…then” problem. The antecedent is the proposition or statement that comes first; the antecedent is contained in the “if” part of the sentence. The consequent refers to the proposition that follows; it is the consequence and is contained in the “then” part of the sentence. When working on a conditional reasoning task, we can either affirm the sentence (say that it is true) or we can deny part of the sentence (saying that it is false). The easiest kind of conditional reasoning task is affirming the antecedent, and the most difficult is denying the consequent. This may be because conditional reasoning relies upon working memory, and we’d expect the burden on working memory to be especially heavy when some of the propositions contain negative terms or abstract ideas. Also, invalid conclusions can be hard because people are tempted to make an illicit conversion (if p then q = if q then p) or in real life, logically invalid conclusions are often true.

2. In the four-card problem, there is a conditional statement that says “If there is a vowel on one side, there is an even number on the other.” There are then four cards – one with a vowel, one with a consonant, one with an even number, and one with an odd number. The question is: what cards would need to be turned over in order to determine if the rule is valid or invalid? We should flip over the vowel – if there’s an even number on the other side, then the rule is upheld. If we flip over the consonant, there’s no valid conclusion no matter what’s on the other side. If we flip over the even number, this is assuming that q is true and leads to no valid conclusion. We should flip over the odd number and there’s a vowel on the other side, then the rule is broken. If there’s a consonant, then there’s no valid conclusion. The greatest percentages of subjects requesting flips were for the vowel and the even number ( 53% and 43% respectively) because of a confirmation bias: a bias to look for positive evidence (or evidence that supports the rule).

FQ 8 #1-8

1. MAUT is multi-attribute utility theory. It is a normative theory of decision making, meaning that it is meant to make the perfect decision under perfect circumstances. The first step in MAUT is to weight the dimensions, or the attributes of a particular choice. The next step is to rate the utility of each major on each dimension. Then, one should combine all ratings of all dimensions across alternatives (or choices). This involves multiplying the weight times the rating. Then, one should choose the alternative with the highest utility, or the highest overall rating. Utility can also be defined as taking all dimensions into account simultaneously. When using this theory, you have to be willing to make tradeoffs between dimensions – one dimension will often compensate for the other. MAUT cannot be used when trying to protect one dimension. Even though a choice may be highest in that dimension, it still may fall in overall utility to other choices.

2. One error (from the perspective of normative theory) that people make when making decisions is that they ignore the weights of the dimensions. In a study by Golotti, students selecting a major were asked to list the alternatives, dimensions, weights of the dimensions, and ratings of the alternatives on dimension. This (WS1) was then taken away from them, and they were given WS2 in which they were to rate their overall feeling of how good the given alternatives felt. Then subjects were given back WS1, and one group calculated ratings with the weights (weights and ratings summed) while another group calculated without the weights. WS1s were then compared to WS2. If people are using the weights in their head in WS2, then the calculation with the weights of WS1 will be more similar to WS2 than the calculations without the weight in WS1. However, there was no specific difference. This suggests that people are not using weights in the overall ratings (WS2) and ignore weights when they have to make decisions in their heads. The error is that normative models say that if one dimension is more important, it should have a bigger impact on the decision. In this case, it didn’t.
Another error is that people do not consider all dimensions simultaneously. One choice may have the highest utility, but it may have been cut from your list of options because you only considered one dimension at a time and that choice didn’t rate high enough on that particular dimension. You might end up missing – from a compensatory point of view – your perfect choice.

3. Framing effects demonstrate that the outcome of a decision can be influenced by two factors: the background context of the choice and the way in which a question is worded. Framing effects are inconsistent when evaluating utilities. The error, from the normative model, is that the utility should be the same if the facts are the same, but this is not the case with framing effects. Framing effects can be illustrated in two separate problems that have the same deep structure – the amount lost or gained is the same. However, the amount seems bigger when expressed as a loss than when expressed as a gain. For example (list own personal example here).

4. According to the representativeness heuristic, we judge that a sample is likely if it is similar to the population from which the sample was selected. People tend to base their decisions on representativeness rather than actual probability. People also tend to ignore the base rate, or how often the item occurs in the population. When given a description of a man that is likely to be engineer, we will say that (because of the representative heuristic), he is more likely to be an engineer than a lawyer even if the population consists of 30 engineers and 70 lawyers. We also commit the conjunction fallacy – we judge the probability of the conjunction of two events to be greater than the probability of either of its constituent events. The combination of the two constituents makes the description seem more representative and plausible. (Give an example?)

5. According to the availability heuristic, we estimate frequency and probability in terms of how easy it is to think of examples of something. Recency plays a role in this heuristic – memory for items generally declines with the passage of time. Thus, we recall more recent items more accurately, and recent items are therefore more available. For example, if I’m currently overloaded with work from this quarter’s classes, I’m more likely to report that classes in general at UW give way too much homework. Familiarity also plays a role – we estimate higher frequencies of things that are more familiar to us. If I know many divorced individuals, I’m likely to estimate higher nation divorce rates (book example).

6. According to the anchoring and adjustment heuristic, we begin with a first approximation – an anchor – and then we make adjustments to that number on the basis of additional information. People tend to rely too heavily on the anchor, and adjustments are often too small. The anchoring and adjustment heuristic is often based on the availability heuristic because highly available information is likely to serve as an anchor. For example, when we meet someone from a particular group, we are likely to rely on stereotypes in order to create an initial anchor. Then we consider the unique characteristics of that individual and make some adjustments. Sometimes we may not make sufficiently large adjustments away from that initial anchor. We use this heuristic when estimating confidence intervals, or ranges within which we expect a number to fall a certain percentage of the time. We tend to make our confidence intervals way too narrow because, as stated earlier, our adjustments away from the anchor are way too small.

7. One assumption is that a perfect decision is always required. This would be good for, say, cancer treatments. However, an adequate decision is often fine. Normative model says that alternative and dimensions are static. In reality, information is dynamic. Normative model assumes that we have plenty of time to deliberate, which actually, in real life, decisions are made quite quickly (like in the situation of the fire commander). Normative model also assumes that all information is available simultaneously and that alternatives are mutually exclusive. Information is really, however, often incomplete and alternatives are not necessarily mutually exclusive. Normative model also assumes that decisions are self-contained. In real life, decisions are often part of action-feedback loops. Situation assessment is figuring out what the alternatives and what the dimensions are. NM emphasizes choice point – selecting among alternatives. In real life, when people have expertise (such as the fire commander), they only really consider one alternative.

8. Overconfidence means that people’s confidence judgments are higher than they should be, based on their actual performance on a task. People are often overconfident in estimating their future performance and even their assessment about other people. People also commit the planning fallacy – they typically underestimate the amount of time or money needed to complete a project; they also estimate that the task will be relatively easy to complete. This had certainly happened to me – before delving into an English paper, I completely underestimated that time it would take to finish it and was pretty stressed out at the last minute. This is related to decision making because it helps me choose to not allot as much time as I really need. Overconfidence is typically based on the following: 1) people are often unaware that their knowledge is based on very tenuous and uncertain assumptions or unreliable information; 2) examples for confirming our hypotheses is readily available, whereas we resist searching for counterexamples, 3) people have difficulty recalling different hypotheses, 4) when able to recall different hypotheses, people don’t treat them seriously, and 5) a self-fulfilling prophecy operates.

6. According to the anchoring and adjustment heuristic, we begin with a first approximation – an anchor – and then we make adjustments to that number on the basis of additional information. People tend to rely too heavily on the anchor, and adjustments are often too small. The anchoring and adjustment heuristic is often based on the availability heuristic because highly available information is likely to serve as an anchor. For example, when we meet someone from a particular group, we are likely to rely on stereotypes in order to create an initial anchor. Then we consider the unique characteristics of that individual and make some adjustments. Sometimes we may not make sufficiently large adjustments away from that initial anchor. We use this heuristic when estimating confidence intervals, or ranges within which we expect a number to fall a certain percentage of the time. We tend to make our confidence intervals way too narrow because, as stated earlier, our adjustments away from the anchor are way too small.

7. One assumption is that a perfect decision is always required. This would be good for, say, cancer treatments. However, an adequate decision is often fine. Normative model says that alternative and dimensions are static. In reality, information is dynamic. Normative model assumes that we have plenty of time to deliberate, which actually, in real life, decisions are made quite quickly (like in the situation of the fire commander). Normative model also assumes that all information is available simultaneously and that alternatives are mutually exclusive. Information is really, however, often incomplete and alternatives are not necessarily mutually exclusive. Normative model also assumes that decisions are self-contained. In real life, decisions are often part of action-feedback loops. Situation assessment is figuring out what the alternatives and what the dimensions are. NM emphasizes choice point – selecting among alternatives. In real life, when people have expertise (such as the fire commander), they only really consider one alternative.

8. Overconfidence means that people’s confidence judgments are higher than they should be, based on their actual performance on a task. People are often overconfident in estimating their future performance and even their assessment about other people. People also commit the planning fallacy – they typically underestimate the amount of time or money needed to complete a project; they also estimate that the task will be relatively easy to complete. This had certainly happened to me – before delving into an English paper, I completely underestimated that time it would take to finish it and was pretty stressed out at the last minute. This is related to decision making because it helps me choose to not allot as much time as I really need. Overconfidence is typically based on the following: 1) people are often unaware that their knowledge is based on very tenuous and uncertain assumptions or unreliable information; 2) examples for confirming our hypotheses is readily available, whereas we resist searching for counterexamples, 3) people have difficulty recalling different hypotheses, 4) when able to recall different hypotheses, people don’t treat them seriously, and 5) a self-fulfilling prophecy operates.

3. Framing effects demonstrate that the outcome of a decision can be influenced by two factors: the background context of the choice and the way in which a question is worded. Framing effects are inconsistent when evaluating utilities. The error, from the normative model, is that the utility should be the same if the facts are the same, but this is not the case with framing effects. Framing effects can be illustrated in two separate problems that have the same deep structure – the amount lost or gained is the same. However, the amount seems bigger when expressed as a loss than when expressed as a gain. For example (list own personal example here).

4. According to the representativeness heuristic, we judge that a sample is likely if it is similar to the population from which the sample was selected. People tend to base their decisions on representativeness rather than actual probability. People also tend to ignore the base rate, or how often the item occurs in the population. When given a description of a man that is likely to be engineer, we will say that (because of the representative heuristic), he is more likely to be an engineer than a lawyer even if the population consists of 30 engineers and 70 lawyers. We also commit the conjunction fallacy – we judge the probability of the conjunction of two events to be greater than the probability of either of its constituent events. The combination of the two constituents makes the description seem more representative and plausible. (Give an example?)

5. According to the availability heuristic, we estimate frequency and probability in terms of how easy it is to think of examples of something. Recency plays a role in this heuristic – memory for items generally declines with the passage of time. Thus, we recall more recent items more accurately, and recent items are therefore more available. For example, if I’m currently overloaded with work from this quarter’s classes, I’m more likely to report that classes in general at UW give way too much homework. Familiarity also plays a role – we estimate higher frequencies of things that are more familiar to us. If I know many divorced individuals, I’m likely to estimate higher nation divorce rates (book example).

1. MAUT is multi-attribute utility theory. It is a normative theory of decision making, meaning that it is meant to make the perfect decision under perfect circumstances. The first step in MAUT is to weight the dimensions, or the attributes of a particular choice. The next step is to rate the utility of each major on each dimension. Then, one should combine all ratings of all dimensions across alternatives (or choices). This involves multiplying the weight times the rating. Then, one should choose the alternative with the highest utility, or the highest overall rating. Utility can also be defined as taking all dimensions into account simultaneously. When using this theory, you have to be willing to make tradeoffs between dimensions – one dimension will often compensate for the other. MAUT cannot be used when trying to protect one dimension. Even though a choice may be highest in that dimension, it still may fall in overall utility to other choices.

2. One error (from the perspective of normative theory) that people make when making decisions is that they ignore the weights of the dimensions. In a study by Golotti, students selecting a major were asked to list the alternatives, dimensions, weights of the dimensions, and ratings of the alternatives on dimension. This (WS1) was then taken away from them, and they were given WS2 in which they were to rate their overall feeling of how good the given alternatives felt. Then subjects were given back WS1, and one group calculated ratings with the weights (weights and ratings summed) while another group calculated without the weights. WS1s were then compared to WS2. If people are using the weights in their head in WS2, then the calculation with the weights of WS1 will be more similar to WS2 than the calculations without the weight in WS1. However, there was no specific difference. This suggests that people are not using weights in the overall ratings (WS2) and ignore weights when they have to make decisions in their heads. The error is that normative models say that if one dimension is more important, it should have a bigger impact on the decision. In this case, it didn’t.
Another error is that people do not consider all dimensions simultaneously. One choice may have the highest utility, but it may have been cut from your list of options because you only considered one dimension at a time and that choice didn’t rate high enough on that particular dimension. You might end up missing – from a compensatory point of view – your perfect choice.

6. Damage to Broca’s area produces speech that is hesitant, effortful, and grammatically simple. The main idea is that their ability to produce language is impaired, although there may be some impairment with understanding language as well. Damage to Wernicke’s area (towards the back of the brain) typically produces serious difficulties in understanding speech. People with Wernicke’s aphasia tend to have very wordy, verbose speech but with little semantic coherence. It has been shown that people with speech disorders typically have more damage to the brain on the left side. “Lateralization of language” meaning each hemisphere of the brain has different functions, and language tends to be dominated by the left side.

Who wants to let Joel get them pregnant?

7. Chomsky argues that much of language is innate. Evidence of this includes that 1) no one is formally taught language and you can’t prevent a child from learning it; 2) children are not simply imitating their parents – they can produce new sentences; 3) very young children who have a very small spoken vocabulary still understand word order and syntax

8. According to the constructionist view of inferences, readers usually draw inferences about the causes of events and the relationships between events. It is “constructionist” because readers actively construct explanations as they integrated the current information with all the relevant information from the previous parts of the text. Evidence for this theory comes from a study where subjects are asked to read several statements. The statements relate semantically and seem to tell a story, but in the inconsistent condition, one of the statements is inconsistent with what would be expected given a previous statement. In this 2×2 design, there is also a “near” condition, where the statements are adjacent to one another, and a “far” condition, where the two statements are separated by other statements. The reaction time is significantly slower for the inconsistent condition across both the near and far conditions. This shows that readers try to connect the material within a text passage, and they consult information stored in long-term memory. During this processing, we try to construct a representation of the text that is internally consistent, even when irrelevant material intervenes.

9. Slips-of-the-tongue are errors in which sounds or entire words are rearranged between two or more different words. One such error is a sound error, which occur when sounds in nearby words are exchanged. For example, this could be snow flurries à flow snurries, doctor pepper à poctor depper, or red brick à bred rick. Another type of error is a morpheme error, which occurs when morphemes (the smallest meaningful unit of language, such as –ly or in-) are exchanged in nearby words. For example, this would be self-destruct instruction à self-instruct destruction. Word errors occur when words are exchanged. For example, “Sending a letter to my mother” would be “Sending a mother to my letter.”

10. What is the relationship between age and language acquisition? According to the critical period hypothesis, our ability, to acquire a second language is strictly limited to a specific period. In fact, by early puberty, we are likely no longer able to acquire a new language with native-like fluency. However, studies have found no abrupt drop in the ability to acquire language as age increases. In terms of phonology, people who acquire a second language during childhood are more likely to pronounce words like a native speaker of that language. In terms of vocabulary, the age of acquisition does not seem to be related to language skills. The controversy about age of acquisition is strongest when considering grammar. For subjects who are natively Korean, when we control for the number of years of education in the U.S., age of acquisition is not related to the individual’s mastery of English grammar.

3. Syntax is the idea that word order helps convey meaning. For example, we know that “red fire engine” is different from “fire engine red.” We, as humans, have a set of rules about sentence structure. The sentence (or group of words) that we read is surface structure – there is no underlying meaning; it is simply what we see and perceive. The deep structure is the deep meaning and actual semantic identity associated with the sentence. Transformational grammar explains how two different deep structures can be expressed in a surface structure or how two different surface structures can be explained by one single deep structure. “Z ate an iguana” is an example of deep structure. This is the general concept behind the sentence. “An iguana was eaten by Zelda” is surface structure; the deep structure behind it is “Zelda ate the iguana.” “The shooting of the hunters was terrible” is an example of an ambiguous sentence. It can be explained by two possible deep structures – X shot the hunters or the hunters shot X.

4. There are a few types of sentence construction that are more difficult for people to process. One is the passive voice – the active form is more basic and the transformation to the passive form requires additional words. A study supports this; subjects were asked to report whether a sentence was plausible or likely. The subjects were highly accurate responding to sentences in the active voice, but accuracy dropped to about 75% when sentences were presented in the passive voice. Another difficult construction is a nested structure. This occurs when a phrase is embedded in another phrase, such as “The plane that I want to take leaves at 12 PM.” When reading this, we experience a “memory cost” – we have to remember the first part of the sentence while trying to process the rest.

5. Case grammar is another explanation for our rules about sentences. However, it is based on semantic roles; it states that we notice meaning changes, not changes in syntax. These roles include the agent (doer), relation (action), recipient (whom receives the relation), and the instrument (what the action is done with). The key to these sentences is the instrument. Garden Path sentences show that we tend to assign case roles on the line word by word. The principles of case grammar explain garden path sentences because it shows that we make assumptions about the first part of the sentence that are incompatible with the end of the sentence. In “The tenant delivered the junk mail threw it in the trash,” the tenant is actually the recipient, not the agent. People take longer to read GPS because they have to revise their case roles.

8. A semantic network is simply another organization of semantic memory. In the network model, concepts are nodes and the relationship between nodes is represented in pathways. The relationship can be an “is a (S)” relationship or “has property of (P)” relationship. In the hierarchical network model, concepts are related hierarchically and have subset or superset properties. Another model is the spreading-activation (Collins/Loftus) model, where the length of the pathway actually indicated the degree of semantic relatedness. Spreading activation explains that the length of the pathway indicated the degree of semantic relatedness. We activate the pathways in between to become conscious of that concept. There’s no strict hierarchical categorization – some things are represented twice. This model explains semantic priming – people are faster to verify a word when it is preceded by a semantically related word. This is because our pathways have been previously activated. Other effect?

1. universals are parts that are shared by all languages

- semanticity: it conveys meaning

- arbitraryness: the connection between sound and meaning is arbitrary

- displacement: a word can mean something that is not currently present (like unicorn)

- productivity: we can make an infinite number of sentences from a finite number of components

- of these, displacement and productivity seem to be unique to humans; animals only convey what they need to because they don’t really have the ability of using higher cognitive processes; animals also have a much smaller “vocabulary” with which to work, so productivity might be only for humans, but they could always tack another another quack or bark…so who knows?

2. a phoneme is the smallest unit of sound that is perceived or the smallest unit that makes a difference to meaning; we use often use top-down processing to determine phonemes because we use what we already know to fill in the gaps. For example, in the McGurk effect, visual input influences auditory perception. In the video we saw in class, the man was saying “baba,” but his mouth was making the shape of “gaga.” When we watched the man, we thought he said “gaga.” However, when we closed our eyes, we correctly heard “baba.” This is a great example of top-down processing. The visual cues were influencing our perception of the phoneme.

The “ants and jelly” experiment studied the semantic code of LTM. Subjects heard statements that were followed by a question (to make sure that they understood the statement thoroughly). The sentences could be comprised of any number of the four idea groupings: a girl and a window, ants and jelly, a tree, or a car on a hill. Each grouping had four propositions. A proposition is the smallest unit of knowledge that can be true or false. Each sentence had a various number of propositions in it. After reading the sentences, subjects were given a five-minute break with a filler task. They were then given a recognition test where they were asked if they had seen the statement – verbatim – in the learning phase of the experiment. They were also asked to rate their confidence in their answer, with +5 being extremely confident that they did see the sentence, and -5 being extremely confident that they did not see the sentence. The results show that the subjects cannot really tell the difference between old and new sentences, but the more propositions were packed into the sentence, the more confident they were that they saw the sentence before. WHY?!

3. Bartlett argued that memory is like a photograph and is largely reconstructive. He argued that we encode certain facts of events/information to get the “gist” of what’s going on, but that we tend to fill in the details from general knowledge stored in our semantic memory. For example, take Bartlett’s own evidence for this theory. He gave subjects a story and asked the subjects to recall the story 4 months later. When recalled, subjects had changed many aspects of the story. For example, subjects noted that the characters were sailing towards an island; in the story, they really had been canoeing in a river. A lot of the supernatural information had been omitted in the retelling as well, though the subjects sometimes reported a “ghost” (or really a “manifestation of breath”) to come out of the man’s mouth instead of something black. What was added-in are known as “normalizations” – they don’t come from episodic memory (the story itself) but rather semantic memory (the general knowledge how of things occur). What is episodic are the key facts – the gist of the story. The details are filled in from semantic memory.

7. The alteration hypothesis assumes that because of the misinformation effect, the original memory has changed. Elizabeth Loftus was a pioneer in the study of the alteration effect. In experiments, she would give subjects incentive (such as $25) for the right answer, but they could still not produce more accurate recall. In a second-guessing experiment, Loftus gave subjects the opportunity to change their answer once they were told that their guess was wrong; in this paradigm, the subjects jumped from 26% being wrong to 77% being wrong. In 1978, again with a film of a car accident, Loftus divided the participants into two groups. The first saw a car overtake at a ‘STOP’ sign, and the other saw a car overtake at a ‘YIELD’ sign. Each group was subdivided into two. The first group from each film was asked “Did another car pass at the stop sign?” and the second group was asked “Did another car pass at the yield sign?” Some time later, the participants were shown pairs of slides with subtle differences and asked which was from the film they had seen. One pair showed a car overtaking at a ‘STOP’ or ‘YIELD’ sign. 75% of those given consistent questions (who were not misinformed) chose the correct slide, while only 41% of those given questions inconsistent with the film chose the right slide. Because of these results, she began to think that she was actually altering original memories. However, another hypothesis exists called the co-existence hypothesis. This states that both the original memory and misinformation coexists. Berkian and Bowers repeated the experiment with the stop and yield signs, and found that for the experimental (mislead) group, errors were significantly greater when the choice recognition was presented in random order rather than sequential order. This supports encoding specificity because the subjects were asked to recall the information in the order in which it was learned and thus had a better recall. This supports the co-existence effect.

1. Flashbulb memories are those that are encoded in a heightened emotional state. The original theory stated that subjects remember flashbulb memories better (even if they are asked to focus on an everyday event). This is in part due to the “now print” mechanism – the tendency to record as much as possible when an event is encoded under a heightened emotional state. However, studies by Talrico and Ruben give us much more information. After different retention intervals (either 1, 7, 42, or 224 days after the event), the number of consistent details recalled is slightly greater for everyday events than for flashbulb memories (though they decrease at the same rate for both conditions). Errors, or inconsistent details, are also pretty much the same for flashbulb memories and everyday events. However, subjects were more confident in their ability to remember flashbulb memories than to remember everyday events.

6. The misinformation effect is an interesting case of retroactive memory interference (where the interfering item follows the memory item before the test). The original events serves as the memory item, and the post-event information serves as the interfering item. A study describing this effect is from Loftus and Palmer (1974), where subjects were shown film of a traffic accident. After the film, they were asked how fast the cars were going when they either “smashed” “collided” “bumped” “hit” or “contacted.” The estimated speed from the subjects was overall greater when the word used in the question had a connotation that would imply faster speeds. Also, the “smashed” group was 2x likely to report broken glass (which was false). The word used in the question served as the post-event information that retroactively interfered with the original event.

8. Emotion (reaction to a specific stimulus) and mood (a more general, long-lasting feeling) both affect LTM. For example, we typically remember pleasant stimuli more accurately than non-pleasant stimuli. In a typical study, subjects will remember more items of a list of pleasant words than a list of neutral or unpleasant words. Also, there is typically more accurate recall for neutral stimuli when it is associated with pleasant stimuli. An example of this effect comes from subjects who watched commercials during a violent and equally-exciting non-violent film. The subjects recalled more information about the commercials when they were in non-violent films. Also, there, there’s typically faster recall for pleasant stimuli, and over time, unpleasant memories fade faster. Mood congruence states that memory is better when the material to be remembered is congruent with a person’s mood. Mood-dependent memory states that we are more likely to remember material if our mood at the time of retrieval matches the mood we’re in when we originally studied the material.

4. As seen by the serial position curve and Rundus and Atkinson, rehearsal is assumed to increase memory. Craik and Lockart, in their Levels of Processing Theory, explain that there are actually 2 rehearsal types. One is maintenance, where information is maintained in consciousness (and therefore does not increase coding into LTM), and the other is elaborative rehearsal, where information is encoded into LTM and analyzed on a more meaningful level. Levels of processing are best described as an increased depth of semantic processing. However, the most accurate definition is circular: we know of deeper processing because of better accuracy, but we know that better accuracy is because of deeper processing. Memory will endure longer if it’s processed at a deeper level, but increased rehearsals at a given level will not increase memory. The IP model, on the other hand, is not a continuum but rather a three step process. Depending on which step of the IP model is reached, that’s how long durability will last (from a fraction of a second to around half minute to permanently). The TV1 is completely amazing to all who have witnessed it’s marketing powers.

5. Encoding specificity argues that the greater the congruence between study and test, the greater is memory accuracy. For example, it is harder for us to name the months of the year in alphabetical order because we learned and typically know them by chronological order. A study that describes encoding specificity is by Thomson and Tulving (1970). They gave subjects a memory item in the learning phase that was either paired with a high-associative cue, a low-associative cue, or no cue at all. The subjects were then tested to recall the memory item with either a high associative cue, a low associative cue, or no cue (i.e. a 3×3 design). A surprising result was that, for the high associate testing group, the low associate learning group performed worse than the no-cue group. The low assoaciate cue messed up the subjects because it messed up encoding specificity – there was a different cue at learning than at test. This contradicts levels of processing because having a cue would seem to give the word more meaning (and a deeper level of processing) than have no cue at all.

2. There is both behavioral and neurological evidence to explain a distinction between STM and LTM. The behavioral evidence consists of the general serial position curve, which shows proportion recall over serial position in the original list of memory items. There’s a functional dissociation between primacy and recency; this means that different variables affect primacy and recency in different ways. Retention interval, for example, is one of these functional dissociations. A 30-second delay after the recall will affect recency, but not primacy. Presentation rate is also a functional dissociation. With slower presentations of the words, primacy and middle recall proportions increase, but recency remains the same. Also, errors show a difference. Typical errors in primacy are semantic (recalling “large” instead of “big”), and this suggests that the words are stored in LTM. Errors in the recency end are typically acoustical (recalling “top” instead of “pop”), which suggests that the words are stored in STM (particularly the phonological loop). Neurological evidence for a STM/LTM distinction comes in the form of amnesic patients. Anterograde amnesiacs cannot form new LT memories, but their short-term memory is fine. Amnesiac K.F. has impaired STM, but her LTM is fine. These are both functional dissociations in each patient, but they combine to form a double dissociation (where there’s one dissociation in one person and the opposite dissociation in the other person). This double dissociation is strong evidence for separate systems. Reviews of the Eastpak Backpack can be found on this site.

3. Explicit memory is, by definition, consciousness with specific reference to the episode in which it was required. It is often referred to as “deliberate memory,” and it is tested through recall and recognition. Implicit memory occurs when previous exposure to information affects behavior, but there doesn’t have to be conscious awareness of that information or the occasion on which there was exposure. An implicit memory test supports that STM and LTM systems are different. Non-impaired subjects are split into two groups that do pleasantness ratings on words. Both groups are then given a fragment completion test, and implicit memory is indicated if the experimental group fills in the blanks with phase 1 words significantly more often than does the control group. Also, dissociations lead to a difference. With varied retention intervals, implicit memory is the same while explicit memory degrades. When information is processed more deeply, explicit memory improves but implict memory is the same. Changing modality also has an effect; it will impair implicit but make no difference in explicit memory. Anterograde amnesiacs also show a difference – they are thought not able to make new LTM, and while explicit memory supports this, implicit tests often provide evidence for new LTMs.

7. Support for Baddeley’s model comes from the dual task paradigm. During a study with this paradigm, subjects were given two tasks. The first task was to remember a list of digits (from two to eight). The second task involved a reasoning skill. In this task, they would be shown something such as “AB” and be given relevant true/false statements such as “A precedes B” or “B is not preceded by A.” Then, they’d be ask to recall the memory items. The results were that the more digits they were asked to recall, the slower the time was for the reasoning task. When both tasks were very easy, there was no performance decrement. When the digit load was greater, the digit span task was therefore more difficult, they… Joel loves to interview people.

1. The serial position represents behavioral evidence for how information is stored in long-term memory. The curve is the plotted proportion of correct memory items recalled over their serial position in the original list. The curve is typically U-shaped with slightly higher recall at the recency end than the primacy end. The argument for the higher recall proportion at the recency end is STM; because of the rehearsal buffer (or what Baddeley would later call the phonological loop), as there are more and more items added to the list, more items toward the beginning of the list drop out of short-term memory. Subjects are allowed to rehearse in this paradigm, and the recency effect is due to smaller retention interval for those items in STM. The argument for the primacy effect is LTM; items at the beginning of the list are rehearsed more often – the more often subjects rehearse something, the more likely the item is to be recalled in LTM. The items in the middle have a low proportion of recall mainly because of rehearsal; they’re being kicked out of STM as new words are added and they have been rehearsed fewer times than those that have been encoded into LTM.

4. The code of information is the basic “code” of how information is represented in STM (for lack of a better description). Three types of codes have been suggested for information in STM: articulatory/acoustical (or auditory), semantic, and visual/spatial. Conrad (1964) conducted an experiment that supports the articulatory/acoustical code. He five letters to subjects and asked them to recall which letters were presented. He studied the errors; if the letter was “F,” the subjects sometimes said “X” or “S” in error. Same thing with the letter “C”- subjects sometimes said “V” or “E” in error. This shows auditory confusion because the letters recalled in error sound like those that were actually presented.

5. Two nonverbal codes that have been suggested for STM are semantic and visual/spatial. An experiment that supports the semantic code involves the remembrance of a three-word trigram, with each word relating to the other categorically. The subjects then count backwards for 15 seconds, and then are asked to recall the trigram. With a 3 trials, the percent of correct recall lowers with each trial due to proactive interference (since the trigrams are all of the same category). However, on the fourth trial, when subjects were given a trigram consisting of words that came from a completely different category, percent recalled suddenly jumped up again. Because the words don’t differ in any way besides their semantic meaning, the percent recalled increases with the level of semantic change, suggesting that the words are encoded semantically. The visual/spatial explanation suggest that for visual information, we mentally rotate information in our mind until it matches what we actually see. This feat occurs in STM, and the bigger the angle of our mental rotation, the longer the processing takes (as if we were to physically rotate an object in the real world). Hurricanes need to arrive now.

6. The Baddeley model of working memory suggests that STM is a multipart system that temporarily holds and manipulates information as we perform cognitive tasks. There are four main parts: phonological loop, visuospatial sketchpad, episodic buffer, and central executive. The phonological loop is a storage and rehearsal buffer for verbal information. It accounts for the digit span, acoustical errors, and the word length effect (remembering fewer longer words). It’s capacity is the number of items that can be repeated in about two seconds, and is therefore not a spatial limit but a temporal limit. Information decays from the phonological loop at some constant rate, and it’s gone in 2 seconds without rehearsal. The visuospatial sketchpad stores visual/spatial information; it also serves as a rehearsal mechanism for that information. The episodic buffer combines information from the phonological loop and the visuospatial sketchpad with information from LTM. The central executive is responsible for directing/allocating information from the three other parts directly to tasks. It is therefore involved in decision-making, reasoning, comprehension, and how to direct information into LTM. Trey Wattson.

7. The Feature Integration Theory of Attention states that we bind components of internal representations of objects together. The first step in this process is feature detection, where individual features of the object are detected with automatic processing. The second stop is feature integration, which requires controlled processing. Here, the features are bound together in a single objected. This integration step is often considered the “glue” of the process. However, the binding-together can be interrupted when attention is diverted. Illusory conjunctions provide support for this theory; in this example, subjects are shown a black digit in the middle of the screen and two letters off to the side. They are asked to identify the digit, then to identify the letters. People will often correctly identify the letters, but they may switch the colors. This illustrates that attention is the glue that puts the features together; it’s “illusory” because the features are correct (so step 1 occurred without problem) but the assembly is incorrect (which occurred in step 2). Zosia Stanley is the older sister of my friend Jarek.

1. The capacity of STM is 7 +/- 2 chunks of information. A chunk is several components of information grouped together on the basis of associated meaning. Chunks therefore allow us to remember more information in short-term memory. For example, I would not be able to remember 28 random numbers in order from start to finish. However, I would be able to remember 28 numbers that consisted of my home phone number, my cell phone number, my parents’ phone number, and my best friend’s phone number. Because of chunking, I’m not really holding 28 numbers in STM but rather 4 chunks of 7 numbers that have already been stored in LTM. Bobby Clarkson was a member of my fat.

2. The duration of information in STM is about 20-30 seconds. Rehearsal poses a problem for STM duration because in that duration of 20 seconds, information can be rehearsed (or mentally repeated), therefore refreshing the 20-second interval. With repeated rehearsal, information could therefore be held in STM for, theoretically, forever. A paradigm that has been used to circumvent this problem is the Brown-Peterson experiment. In this study, subjects were given a 3-letter trigram to remember. They were then given a 3-digit number and instructed to count backwards by 3s from that number to the clicks of a metronome that sounded twice every second. This was the distractor task – it prevented the subject from rehearsing. Subjects were then asked to recall the trigram. Over a number of trials, the greater the duration of the distractor task, the smaller the proportion was of correct trigrams recalled. The conclusions made from this paradigm were that STM fades slowly over time and that its duration is gone in about 20 seconds.

3. There are 3 ways that information can be lost from STM. One way is decay, or simple fading, in which memory evaporates on its own over time. In the Brown-Peterson paradigm, this was an original conclusion inferred from the results; the longer rehearsal was prevented, the shorter the STM duration (until it reached 20 seconds), and more and more information was lost. Shawn Wang was my roommate. Another way of forgetting is proactive interference, where the interfering item is presented before the memory item & affects recall of the memory item. In the Brown-Peterson paradigm, the interfering items could be thought of as the previous trigrams from previous trials. Another way is retroactive interference, where the memory item is presented before the interfering item, and subsequent recall of the memory item is affected. In the Brown-Peterson paradigm, the digits (occurring after the memory item, the trigram) are categorically dissimilar to the letters, so it’s assumed that retroactive interference is not a problem.

4. Controlled (conscious) processing is an intentional mental process. We are aware of this processing; open introspection occurs. Attentional resources are needed, and it proceeds very slowly (or serially). An example of controlled processing is seen in a young child who is just learning to walk. The child must make a conscious effort to pick up one foot and move it in front of the other, all while keeping balance and judging what movements making walking easier. Automatic processing, on the other hand, occurs without intentional control. It is not open to conscious awareness, and it needs very few attentional resources to operate. Automatic processing also occurs rapidly and in parallel. An example of this is seen when normal, adult humans walk – they don’t consciously need to decide how to do so. The difference in mental processing between walking infants and walking adults is due to practice; a task that is very well practice becomes more and more automatic, where the same stimulus will yield the same response.

5. In the Stroop effect study, subjects are asked to name the color of the word that the word is presented in. The data of the study show that subjects have trouble naming the ink colors when that color is used in printing. (For example, if the word RED is presented in the color red, subjects will respond faster than if the word RED is presented in the color BLUE.) The findings suggest that the reading part of this task is automatic. It is done without intention or control, and therefore the subjects have no choice but to have their reading of the word interfere with their processing of the color of the word. Ashley Morgan is legendary.

6. Visual search is the allocation of attention over a visual environment required to detect a target. When a target is identified according to a single feature (i.e. a blue X in a group of red X’s and O’s), reaction time is equal no matter the set size. This is called disjunction. A key characteristic of this processing is that it’s automatic – processes occur in parallel. However, when the target is identified by two features that occur together in the same object (i.e. a blue X in a group of red X’s and red and blue O’s,), this is referred to as conjunction, and reaction time is less for smaller sets than for larger sets. This is because this type of processing requires focused attention – the search is serial. Searching for an item according to a single feature is therefore like automatic processing, and searching by two features is like controlled (or conscious) processing.

9. Most research suggests that attention is managed by two regions of the cortex: the posterior attention network and the anterior attention network. The posterior attention network, located in parietal lobe, is responsible for attention given to visual searching and spatial awareness. I use this part of my brain while searching for my keys in the morning. The anterior attention network, located in the frontal lobe, is responsible for attention tasks that focus on word meaning. It also aids in top-down processing, as evident by its activity when subjects are asked to listening to a list of nouns and state the meaning of each noun. I use this part of my brain while I’m reading my cognitive textbook.

#8 from week 2: Connectionism is a very specific model of the top-down theory of pattern recognition. It includes three main units: input unties, hidden units, and output units. The input units receive and encode external stimuli, the output units yield a response, and the hidden units that lie in between receive connections from the input units and send connection to the output units. The information is stored in the connections between the units, not in the units themselves. There can be varying weights of the units: strong or weak and positive or negative. All weights start out neutral but are shaped by experience. The connectionist models are great for human cognition because they are able to make use of context. Andy Meade is a strange man. For example, in THE CAT, prior knowledge that an H will often follow a T is held in a connection. They are also great because they can learn from examples the way that humans do, and they are also parsimonious – the same group of units can be trained to recognize many different things.