Cognitive Science Introduction to the Mind Bermudez Test Bank ed 3

Why is behaviorism largely incompatible with the basic assumptions of cognitive science? A. Behaviorists study non-psychological observable phenomena and measurable behavior, while cognitive science is the study of cognition and mind. B. Cognitive scientists are not interested in behavior. C. Behaviorism is not scientific. D. Behaviorists study non-psychological observable phenomena and measurable behavior, while cognitive science is the study of association and reinforcement learning.

A. Behaviorists study non-psychological observable phenomena and measurable behavior, while cognitive science is the study of cognition and mind.

Reinforcement learning and classical conditioning are A. The same. B. Two different types of conditioning, both exploiting mechanisms of association and reinforcement. C. Different because reinforcement learning depends upon mechanisms of reinforcement while classical conditioning does not. D. Different because classical conditioning depends upon mechanisms of association while reinforcement learning does not.

B. Two different types of conditioning, both exploiting mechanisms of association and reinforcement.

What is the key feature of the maze experiments in Tolman and Honzik (1930)? A. Rats learn to run the maze by response learning. B. Reinforcement is the best way for rats to learn to run the maze. C. Rats learn the layout of the maze only when they are rewarded. D. Rats can learn the layout of the maze without any direct reinforcement.

D. Rats can learn the layout of the maze without any direct reinforcement.

In the cross-maze experiments reported in Tolman, Ritchie, and Kalish (1946), why did the rats for which the food was always in the same place learn to run the maze more quickly? A. Because they used sequences of movements to figure out the maze. B. Because they learned spatial information about the maze in terms of places. C. Because they could selectively focus their attention on the reward. D. Because it was easier to learn the location by reinforcement.

B. Because they learned spatial information about the maze in terms of places.

The cross-maze experiments reported in Tolman, Ritchie, and Kalish (1946) led him to propose which concept? A. Cognitive map B. Reinforcement learning C. Latent learning D. Hierarchical behavior

A. Cognitive map.

In Lashley’s work on explaining complex behaviors, which concept is essential? A. Hierarchical organization B. Cognitive map C. Deep structure and surface structure D. Representation

A. Hierarchical organization.

Which of the following statement is not compatible with the hypothesis of subconscious information processing? A. Complex behaviors are often hierarchically organized. B. We are often not aware of our high-level plans. C. We are always aware of the information processing involved in carrying out a plan. D. We often behave by breaking the task down into more basic ones.

C. We are always aware of the information processing involved in carrying out a plan.

Which is a good example of illustrating the idea of task analysis? A. Recalling one’s password B. A rat learning to run a maze C. Playing tennis D. A, B, and C

D. A, B, and C.

Which key idea in the development of cognitive science is attributable to Alan Turing? A. Information processing is fundamentally algorithmic. B. Not all behavior can be explained through conditioning. C. The idea of hierarchically organized behavior provides a clue for how mind might process information. D. The idea of syntax provides a clue for how mind might process information.

A. Information processing is fundamentally algorithmic.

In symbolizing the sentence “John has hit the ball” as the string of symbols “NP1 + Aux + V + NP2”, which aspect of the sentence has been revealed? A. Its deep structure B. Its meaning C. Its surface structure D. Its pragmatic structure

A. Its deep structure.

According to the Church-Turing thesis, every computable function can be computed by ––– A. a Turing machine

A. a Turing machine.

In his theory of information, Claude Shannon uses which concept to measure information? A. A bit B. A chunk C. An informatron

A. A bit.

An information channel with a channel capacity of n bits can discriminate how many pieces of information? A. 2ⁿ pieces of information (so that a channel with a capacity of 3 bits can discriminate 6 items) B. n² pieces of information (so that a channel with a capacity of 3 bits can discriminate 9 items) C. 2ⁿ pieces of information (so that a channel with a capacity of 3 bits can discriminate 8 items)

C. 2ⁿ pieces of information (so that a channel with a capacity of 3 bits can discriminate 8 items).

Why (according to George Miller) are people better at making relative judgments than absolute judgments? A. Because making relative judgments requires less channel capacity. B. Because the information channels for making absolute judgments have lower capacity. C. Because making absolute judgments requires more attention. D. Because absolute judgments require hierarchical information-processing.

A. Because making relative judgments requires less channel capacity.

Decimal and binary number systems are different forms of A. Chunking numerical information B. Algorithmic computing C. Subconscious information processing D. Deep numerical structure

A. Chunking numerical information.

When you read in a noisy coffee shop, you focus on your reading and are not distracted by the conversations around you. This is an example of? A. Selective attention B. Chunking information C. Hierarchical organization D. Spatial learning

A. Selective attention.

In Broadbent’s dichotic listening experiments, what best explains the fact that people performed best when they reported the stimuli ear by ear? A. The selective filter makes people focus on one string of information at a time. B. Their auditory systems break the task into sub-tasks. C. People use different transformational rules for information they get from different ears.

A. The selective filter makes people focus on one string of information at a time.

According to Broadbent’s theory of attention, perceptual systems are information channels. A. True B. False

A. True.

On Broadbent’s theory of attention, information is semantically processed before it enters the selective filter. A. False B. True

A. False.

Broadbent illustrated the stages of information processing through –– A. A flowchart B. A Venn diagram C. A cognitive map D. An algorithm

A. A flowchart.

Terry Winograd’s SHRDLU is a general-purpose language program, as flexible as a human language. A. True B. False

B. False

In Terry Winograd’s SHRDLU, linguistic understanding is achieved through which of the following? A. Algorithmic procedures B. Conversational chatterbots C. Phrase structure trees

A. Algorithmic procedures

Which of the following does Terry Winograd’s SHRDLU not contain? A. A syntactic analysis system B. A pragmatic analysis system C. A semantic analysis system D. A general cognitive-deductive system

B. A pragmatic analysis system

Terry Winograd’s SHRDLU operates in a micro-world and so does not have deal with the semantic aspects of language A. True B. False

B. False

Terry Winograd’s SHRDLU operates in a micro-world and so does not have deal with the pragmatic aspects of language A. True B. False

A. True

Shepherd and Metzler’s rotation experiments found that there was no relation between (a) the length of time subjects took to determine whether one shape was a rotation of another, and (b) the degree of rotation between the shapes. A. True B. False

B. False

Which of the following is an example of digital information processing? A. A map B. A sequence of 0’s and 1’s C. A photograph

B. A sequence of 0’s and 1’s

Which of the following is an example of analog or imagistic information processing? A. A map B. A sequence of 0’s and 1’s C. A recipe in a cookbook

A. A map

According to Stephen Kosslyn, the information processing in mental imagery involves operations carried out on imagistically represented information. A. True B. False

A. True

According to Shepard and Kosslyn, if information is imagistically encoded, then the length of processing time will be a function only of the complexity of the information, measured in bits A. True B. False

B. False

The imagery debate is a debate about which of the following? A. How we consciously experience mental images. B. The subconscious information-processing that underlies conscious mental images. C. The informational differences between pictures and photographs

B. The subconscious information-processing that underlies conscious mental images.

Marr’s analysis of the visual system is a classic example of a ––– approach in cognitive science. A. Bottom-up B. Top-down C. Holistic

B. Top-down

In Marr’s analysis of the visual system, the primal sketch, the 2.5D sketch, and the 3D sketch are constructed ­­–– A. In sequence B. In parallel C. Each independently of what happens at other levels

A. In sequence

The representational primitives in Marr’s primal sketch include A. Rods and cones B. Zero crossings and edges C. Representations of orientation and depth

B. Zero crossings and edges

According to Marr, the function of the visual system is to identify objects in the distal environment A. True B. False

B. False

In Marr’s theory, the primal sketch, 2.5D, and 3D sketch are all elements at which level of analysis? A. The computational level B. The algorithmic level C. The implementational level

B. The algorithmic level

According to Marr, Elizabeth Warrington’s work with neuropsychological patients showed that the visual system processes information about objects’ shape independently of information about their function. A. True B. False

A. True

Depth and orientation information appears in which stage of visual processing, on Marr’s model? A. The primal sketch B. The 2.5D sketch C. The 3D sketch

B. The 2.5D sketch

In Marr’s theory, the 3D sketch A. Contains information about what objects are for B. Is viewer-centered C. Represents objects independently of the viewer’s perspective D. Is computed in the primary visual cortex

C. Represents objects independently of the viewer’s perspective

Detailed studies of neurobiology and neuroanatomy are most relevant at –– A. The computational level B. The algorithmic level C. The implementational level

C. The implementational level

Cognitive scientists who hold that mental capacities are multiply realizable tend to think that we can understand those capacities without looking at the details of how the brain works. A. True B. False

A. True

If cognitive processes are multiply realizable then that suggests that they are ––––– processes A. Implementational B. Functional C. Digital D. Analog

B. Functional

The cerebral cortex is part of the –– A. Forebrain B. Midbrain C. Hindbrain D. Limbic system

A. Forebrain

The term “gray matter” is used to describe the tissue in which part of the human brain A. Limbic system B. Cerebral cortex C. Subcortical areas

B. Cerebral cortex

The term “white matter” is used to describe the tissue in which part of the human brain A. The temporal lobe B. Cerebral cortex C. Subcortical areas

C. Subcortical areas

Visual information about the distal environment is processed contra-laterally, so that information from the left side is processed in the right hemisphere. A. True B. False

A. True

The ventral pathway goes from the primary visual cortex to –– A. The occipital lobe B. The temporal lobe C. The frontal lobe D. The parietal lobe

B. The temporal lobe

The dorsal pathway goes from the primary visual cortex to –– A. The occipital lobe B. The temporal lobe C. The frontal lobe D. The parietal lobe

D. The parietal lobe

According to Mishkin and Ungerleider, the ventral pathway primarily carries information relevant to –– A. Locating objects B. Identifying objects C. Reacting emotionally to objects

B. Identifying objects

According to Mishkin and Ungerleider, the dorsal pathway primarily carries information relevant to –– A. Locating objects B. Identifying objects C. Reacting emotionally to objects

B. Identifying objects

Artificial neural networks (connectionist networks) are like digital computers in that they use highly specific programs that vary according to the particular task they are trying to solve. A. True B. False

B. False

In artificial neural networks (connectionist networks), information is carried by A. The pattern of weights across the network B. Individual units that form language-like structures C. The structure of the hidden layers D. The error signal

A. The pattern of weights across the network

Neural networks learn by adjusting the weights of individual units to reduce overall error. A. True B. False

A. True

Neural networks are well suited to tasks of pattern recognition. A. True B. False

A. True

Which technology measures the BOLD signal? A. EEG B. fMRI C. PET D. ECG

B. fMRI

A blocked design involves subjects performing a single task for extended periods of time, so that the BOLD signal can be compared and contrasted in different conditions. A. True B. False

A. True

Which method allows the BOLD signal to be measured in short duration tasks? A. Blocked design B. Subtraction methods C. Event-related design

C. Event-related design

Functional neuroimaging directly measures cognitive activity. A. True B. False

B. False

The local field potential measures –– A. How connected a population of neurons is B. The average firing rates of neurons C. The “input” to neurons, irrespective of whether they fire

C. The “input” to neurons, irrespective of whether they fire

Using a combination of fMRI and microelectrode recordings Logothetis and his collaborators found evidence that the BOLD response is correlated with ––– A. Average neuronal firing rates B. The local field potential C. Multi-unit spiking maxima

B. The local field potential

The aim of neural network modeling is –– A. To match accurately the behavior of individual biological neurons B. To approximate the behavior of populations of neurons C. To study the activity of distinct regions of the brain

B. To approximate the behavior of populations of neurons

A biological neuron has A. One axon and one dendrite B. Many axons and a single dendrite C. A single axon and many dendrites D. Many axons and many dendrites

C. A single axon and many dendrites

A sending neuron’s axon comes close to a receiving neuron’s dendrite at A. A neurite B. A synapse C. An activation threshold

B. A synapse

All activation functions in artificial neural networks are linear A. True B. False

B. False

Artificial neural network units with the same inputs but different activation functions can have different outputs A. True B. False

A. True

Which of the following mathematical operations is not a function? A. Squaring B. Taking square roots C. Addition D. Subtraction

B. Taking square roots

Boolean functions are defined over –– A. Truth values B. Numerals C. Numbers D. Network units

A. Truth values

“Neurons that wire together, fire together” is a description of which type of learning? A. Hebbian learning B. Perceptron convergence C. Backpropagation of error

A. Hebbian learning

Perceptron convergence learning does not require feedback about error A. True B. False

B. False

Any Boolean function can be represented by a single-layer network? A. True B. False

B. False

The Boolean function OR (inclusive OR) assigns which truth value to the input pair (TRUE, TRUE)? A. TRUE B. FALSE

A. TRUE

The Boolean function XOR (exclusive OR) assigns which truth value to the input pair (TRUE, TRUE)? A. TRUE B. FALSE

B. FALSE

The Boolean function XOR cannot be represented by a single-layer network because it is not linearly separable. A. True B. False

A. True

The perceptron convergence learning rule can be used for both single-layer and multi-layer networks. A. True B. False

B. False

Which of the following is true of feedforward networks? A. Activation spreads between units in a given layer B. They contain feedback loops C. Activation spreads through the network in a series of discrete time steps corresponding to the number of layers

C. Activation spreads through the network in a series of discrete time steps corresponding to the number of layers

Which type of learning rule can work for networks with hidden units? A. Hebbian learning B. Perceptron convergence C. Backpropagation of error

C. Backpropagation of error

All learning algorithms for multilayer networks require supervision and feedback. A. True B. False

B. False

Which of the following types of learning is supervised? A. Hebbian learning B. Competitive learning C. The backpropagation algorithm

C. The backpropagation algorithm

How does an artificial neural network store information? A. In language-like physical symbol structures B. In its learning algorithm C. In the pattern of weights holding across individual units D. In the output units

C. In the pattern of weights holding across individual units

There is no clear distinction between information storage and information processing in artificial neural network models A. True B. False

A. True

Which of the following is an example of a dynamical system? A. A dripping tap B. A Turing machine C. The solar system D. All of the above

D. All of the above

Dynamical systems theorists accept that the basic currency of cognitive science is information-carrying representations. A. True B. False

B. False

Dynamical systems theorists think that traditional cognitive scientists are mistaken in trying to reverse-engineer the mind as an information-processing system. A. True B. False

A. True

Dynamical systems theorists emphasize which of the following basic ideas? A. The human mind is an autonomous problem-solver B. The multiple realizability of cognition C. Human beings are embodied and embedded in their environments

C. Human beings are embodied and embedded in their environments

Which type of equations allow us to model quantities that change continuously over time? A. Difference equations B. Differential equations

B. Differential equations

Which type of equations allow us to model quantities that change in discrete time steps? A. Difference equations B. Differential equations

A. Difference equations

The state space of a dynamical system always has –– A. Four dimensions (the three spatial dimensions, plus time) B. As many dimensions as there are dimensions of variation in the system C. Infinitely many dimensions

B. As many dimensions as there are dimensions of variation in the system

The (hypothetical) computational governor is a good illustration of dynamical systems theory because it is (i) representational, (ii) computational, (iii) sequential, and (iv) decomposable. A. True B. False

B. False

The Watt governor is a good example of a dynamical system because A. The flywheel and the throttle valve are directly coupled, so that speed can be almost instantaneously regulated. B. It uses the difference between actual speed and desired speed to calculate the required throttle adjustment. C. It is made up of distinct and semi-autonomous sub-systems

A. The flywheel and the throttle valve are directly coupled, so that speed can be almost instantaneously regulated

The state spaces of dynamical systems often have attractors, which are regions on which multiple trajectories converge. A. True B. False

A. True

Which types of dynamical system have multiple basins of attraction? A. Linear dynamical systems B. Nonlinear dynamical systems C. All dynamical systems

B. Nonlinear dynamical systems

Dynamical systems theorists hold that the mind–environment complex is a system whose behavior evolves as a function of a small number of variables. A. True B. False

A. True

Dynamical systems models of infant walking behavior explain the U-shaped trajectory of infant walking in terms of –– A. Cortical maturation B. The role of parameters such as gravity, leg strength, inertia, and leg fat C. Comparator mechanisms

B. The role of parameters such as gravity, leg strength, inertia, and leg fat

It is evidence against computational models of motor control that infant stepping behavior can be produced by manipulating variables that are not part of the central nervous system. A. True B. False

A. True

According to dynamical systems theorists, the A-not-B error occurs because infants’ representational abilities are insufficiently developed. A. True B. False

B. False

Dynamical systems models of motor control explicitly calculate trajectories from starting positions to end-point positions. A. True B. False

B. False

The A-not-B error can be made to disappear by changing the infant’s posture. A. True B. False

A. True

Infant performance on the A-not-B task can be improved by tapping the right side of the visual field. A. True B. False

A. True

Then and Smith’s dynamical field model includes parameters for A. Memory input B. Cortical maturation C. Level of executive functioning

A. Memory input

The fact that some aspects of cognition can be as dynamical systems proves that there is no need to appeal to representations and information in cognitive science. A. True B. False

B. False