Cognitive Processes Exam Ii

Front 1

1. Define each system within the dual systems theory (as described by Sloman 1996 and Evans 2008) and give one example of a phenomenon that is characteristic of each system. Describe one criticism of dual systems theory outlined by Keren and Schul (2009).

Back 1

Associative system: based on similarity and temporal structure, vs. causal or mechanical structure. Based on correlations between stimuli informed by experience. More automatic and rapid. E.g., phenomenon of associative memory, in which recalling one memory activates automatic recall of another linked memory.

Rule-based system: system of processing based on logical structure and set of variables. Rules can be instructions, laws, etc. More deliberate and slow. E.g., phenomenon of deliberation, in which a problem is worked through consciously and deliberately.

Criticism of dual systems theory: the systems are not in fact isolable. I.e., system 1 can’t carry out its operations without function of system 2, and vice-versa. Both do higher-order cognitive processing that is interdependent. Therefore, they are not discrete systems.

Front 2

2. In class and in the articles we read about dual systems theory the belief bias effect was often mentioned. Describe the belief bias effect. What does this demonstrate about human reasoning? Is belief bias a System 1 or System 2 phenomenon? Explain? How would a mental models account explain this effect?

Back 2

Belief bias effect: we are more likely to think that believable conclusions are valid. For example, when a syllogism is invalid, we are more likely to think the conclusions are valid if they are believable.

This shows that we use believability as a heuristic to establish validity. This is probably a System 1 phenomenon, because, for example, it has been shown that belief bias increases in use as time pressure to make a decision increases.

A mental model in which people search harder for counterexamples to conclusions that violate their knowledge might explain this effect.

Front 3

3. In the study of human judgment what is a heuristic and what is a bias? Describe two heuristics and their associated biases using examples given in class?

Back 3

Heuristic: strategy for solving a problem by using a probabilistic rule, rather than examining all available evidence (algorithm). Bias: error when heuristic is applied but wrong solution is produced.

Availability heuristic: probabilities are evaluated by ease with which instances or occurrences come to mind. Can lead to encoding bias, e.g., people assume incorrectly more people die of homicide than stomach cancer, because they are more exposed to information about homicide.

Representitativeness heuristic: probabilities are evaluated by degree to which A is representative of B. E.g., nerdiness is representative of librarian. Can lead to bias of conjunction fallacy, in which a more specific statement with conjunction of two events is thought to be more probable than single event, even though this can never be true.

Front 4

4. When comparing insight and non-¬‐insight problem solving from a phenomenological point of view how would Metcalfe and Wiebe (1987) distinguish the problem types?

Back 4

Insight: problems should be solved in a flash. Warmth and knowing low until end.

Non-insight: opposite. Problems solved deliberately and sequentially. Warmth and knowing gradually increase.

M&W would say that problems should be tested on these dimensions of warmth and knowing, and categorized post hoc into insight or non-insight problems.

Front 5

5. Knoblich et al. (1999) describe two processes critical to solving insight problems. What were they? Provide an example of each using the matchstick arithmetic task. Would Knoblich et al. (1996) disagree with Metcalfe and Wiebe’s (1987) claim about the phenomenology of insight?

Back 5

Relaxation of constraints: how constrained you are will determine how easy impasse is to overcome. E.g., will changing a fundamental rule help you solve the problem? Is a screwdriver only useful for screwing? Matchstick example: need to be able to relax a constraint to solve the problem. E.g., operator constraint – need to be able to change plus sign to equals sign.

Decomposition of chunks: if units of the problem are easy to break down into smaller meaningful units, insight problems are more solvable. Matchstick example: can break down plus sign into number 1 and a minus sign to try to solve problem.

Seems that Knoblich is incompatible with M&W. Knoblich sets out to define aspects of insight problems ahead of time, whereas M&W want to do it post hoc.

Front 6

6. Use Thorndike’s (1898) “cat in box” experiment to describe initial states, goal states, and trial and error learning. What is a problem space and how might errors in representing the problem space inhibit problem solving? Provide two examples of difficult problems that likely derive their difficulty from a problem representing the problem space. Explain each example.

Back 6

Problem space: internal representation of the initial state, the goal state, subgoal states, and operations that can be applied within to move from state to state. Errors in representing problem space might not allow you to apply the correct operations to move from state to state, or to not identify the subgoal states; i.e., seeing problems as source of error.

Example 1: Number sequence problem. What rule generated the sequence 8,5,4,1,7,6,10,0? Hard to figure out that it’s alphabetical, as most will try to do mathematical operations.

Example 2: Candle on wall problem. Get fixated on use of box as container of tacks. Real solution is to use as shelf for candle.

Front 7

7. Categories have both a vertical and a horizontal dimension. Describe what these concepts mean (use a diagram if need be). Describe one important attribute (discussed in class) about each dimension. Lastly, describe the prototype and essentialist approaches to determining category membership.

Back 7

Vertical dimension: describes level of inclusiveness of category. E.g., Superordinate – furniture; basic – chair; subordinate – kitchen chair. Must strike ideal balance between not enough and too much information about object! Basic probably most ideal for most situations.

Horizontal dimension: distinguishes between different concepts at same level of inclusiveness. E.g., robin, canary, sparrow are types of bird. Has a graded structure – some category members are better than others. Robin is better representative of bird than ostrich!

Prototype approach to category membership: Based on similarity – enough attributes in common with other objects in the category determine membership. There’s a prototype “bird” feathers, beak, etc.

Essentialist: based on individual’s theory about essence of object. E.g., bird or butterfly? Which does it look more similar to? Is this a cow or a pig – does it moo or say oink?

Front 8

8. What is a category specific visual agnosia (CSVA)? Provide a description of the two general competing accounts of CSVA. Which account does Dixon et al.’s (2002) work with ELM support and why?

Back 8

In category-specific visual agnosia (CSVA), visual identification deficits following temporal-lobe damage selectively affect some categories of objects
but spare others.


Sensory/Functional Hypothesis vs. dominant knowledge hypothesis


SFH- categories are made up of a feature-based organizing system that is made up of visual and semantic attributes of objects. CSVA stems from damage
to any visual or semantic system that may hinder one from accessing the visual or semantic “demon”, therefore interrupting the process of accessing all the information to “categorize” an object. In short, we organize things
based on non-categorical features, we do not organize things within categories.


DVH- we organized information into “categories” using our domain- specific knowledge. Therefore CSVA stems from the inability to access that “category” of knowledge due to damage to specific neural pathways that are responsible
for retrieving information from a specific organizing system for a specific “category”. In short, we actually categorize things initially based on our domain- specific knowledge.


Dixon’s work describes a feature-based theory of CSVA, whereby identification problems stem from difficulties in “unpacking” information about semantic and visual features of objects that are highly related???.This is the sensory/functional hypothesis.

Front 9

9. What is a phoneme? How does coarticulation pose a problem for speech recognition? Describe the phonemic restoration phenomenon and how it might partially solve this problem?

Back 9

Phoneme: smallest unit of sound that can be used to differentiate words (e.g., b in “bat”).

Coarticulation means that phonemes do not appear as discrete units in a speech spectrogram; instead, sound of phoneme influenced by surrounding phonemes. Words in sentences also run together. Speech recognition would probably work better with discrete units without overlap.

Phonemic restoration is when phonemes that are not heard are “filled in” based on context. State legislators cough example. Might solve this problem by helping to fill in the gaps.

Front 10

10. The dominant theory of reading is the dual route theory. Explain this account of reading. Which route or routes can successfully read regular words? Which route or routes can successfully read irregular words? Which route or routes can successfully read pseudowords? How would this account explain surface dyslexia and phonological dyslexia?

Back 10

Dual route theory says that there are two routes by which we read words. Goes print → features → letter identification → lexical OR nonlexical route → speech

Either route can read regular words.

The lexical route works by activating lexicon, or mental dictionary of known items. This is the route that can read irregular words, or words that don’t follow rules of phonemes, like “pint.” You have to know it to read it.

The nonlexical route works by turning graphemes to phonemes. This is the route that can read pseudowords, like “yint.” Makes sense, because there is no lexical object, so you have to process word piece by piece.

Phonological dyslexia: can’t read pseudowords; regular and irregular words ok. Probably deficit in nonlexical route, problems with processing phonemes.

Surface dyslexia: Can’t read irregular words. Regular and nonwords ok. Probably deficit with lexical system.

Front 11

11. According to embodied account of language, language involves the motor system in tasks like comprehension. Describe three pieces of experimental evidence for this claim discussed in class. How does this evidence support the embodied account? Describe one problem Mahon and Caramazza have with the embodied account of language.

Back 11

Piece 1: Glenberg et al. 2008 – fatigue motor neural system by moving beans for 20 min. Showed interaction between direction of bean movement and toward/away language – slower to make judgment about sentences when sentence matched direction of bean movement. Supports embodied account showing that understanding a sentence is affected by a motor task.

Piece 2: Glenberg & Kaschak 2002 – showed “action compatibility effect,” in which similar interaction was shown. Sentences with action in one direction made sensibility judgments in the other direction more difficult. Same interaction as above. Same as above in that understanding sentence is affected by motor task.

Piece 3: Havas et al 2007 – judgment times for whether sentence makes sense are faster when emotion from facial expression (pen method) is congruent with valence of sentence. Shows that similar to motor function, emotion simulation has an effect on language processing.

Mahon & Carmazza – one criticism states that findings from apraxia patients, i.e., people who have impairment for using objects not explained by sensory or motor impairments, show that these patients can be impaired for using an object but can still name the object and recognize others pantomiming its use. This by default falsifies a “strong” embodied cognition.

Front 12

12. What is artificial intelligence (AI)? Are all AI systems modeled on human cognition? Explain? Is Deep Blue modeled on human cognition? Systems like Deep Blue naturally raise questions about whether the system is “intelligent” or “able to think” Describe the Turing test as a means to answer such questions.

Back 12

AI is the development of machines or computers to engage in intelligent behavior. Some AI systems are modeled on human cognition, i.e., try to solve problems the way humans do. Others focus on just solving problems that humans can solve, but do so by other methods (called ideal). Deep Blue is not - "brute force" or algorithmic process. Very effective - beat world champion!

The Turing test is test of machine’s ability to exhibit intelligent behavior equivalent or indistinguishable from an actual human. In the original form, a person converses with an entity not visible to them. It’s either a human or a software program. If the person can’t tell the software program isn’t human, it has passed the Turing test. This is based on the premise that you can tell if a machine is thinking by comparing its behavior to that of a human – which is a controversial premise. E.g., different people might make different judgments about the machine, or comparing behavior is not enough to show that “thinking” is going on.

Front 13

13. Breazeal (2003) describes Kismet, a “sociable” robot. What is a sociable robot? Name three behaviors that Kismet displays that are meant to engage humans socially. Were humans able to interact “normally” with Kismet? Describe both its successes and failures in this respect.

Back 13

Sociable robot is a robot that people use a social model to interact with them, vs. how we might typically interact with a computer program.


To engage humans, Kismet might acquire speaking floor by breaking eye contact and leaning back, end a speaking turn by stopping babbling, and hold the floor by looking to the side rather than making eye contact.


Kismet was generally able to engage people in “normal” protodialogue, about 82% of the time based on the study we read. Ps weren’t told how to interact with Kismet, but Kismet still able to regulate interactions, e.g., with behaviors described above. However, there were frequently “hiccups,” in which Kismet might interrupt, might lag with a response, or otherwise break the flow of smooth communication.