Part B - Cognitive Aspects

Thinking, Memory, and Emotions

Describe the Different Features of Human Reasoning
Describe the Different Models of Human Memory
Describe Three Different Theories of Emotion

"Memory is central to all cognitive processes" Solso, Maclin, and Maclin 2008

Thinking | Early Model of Memory | Working Memory | Emotion | Exercises

Cognitive psychology distinguished itself from the dominant tradition of behaviorism by introducing the study of internal mental states.  Highlighting the study of internal mental states represented a revolution in the field of psychology.  Cognitive psychology uses the scientific method to study internal mental states. 

The Oxford English Dictionary defines the scientific method as:

a method of procedure that has characterized natural science since the 17th century, consisting in systematic observation, measurement, and experiment, and the formulation, testing, and modification of hypotheses.

In this chapter, we survey the results of research in the areas of thinking, memory, and emotion with reference to single human subjects. 


Thinking is the process of the conscious mind that encompasses reasoning, decision-making, problem solving, and recollection.  It extends to the acquisition of skill sets and the formation of mental models. 


Reasoning is the drawing of conclusions from known information.  We classify reasoning into three types:

  • abductive
  • deductive
  • inductive

Abductive Reasoning

Abductive reasoning derives a particular fact from another particular fact, action, or state.  For example,

       Observation - You look tired
       Conclusion - Therefore you did not get any sleep last night

The observation may follow from the conclusion, but the conclusion does not necessarily follow from the observation.  There may be a completely different cause to the observation.  The conclusion and the observation may be unrelated. 

Deductive Reasoning

Deductive reasoning derives a logically necessary conclusion from a general statement.  For example:

       General Statement - All humans think
       Observation - You are a human
       Conclusion - Therefore, you think

Note that a conclusion in deductive reasoning is not necessarily true.  For example:

       General Statement - All humans cannot think
       Observation - You are a human
       Conclusion - Therefore, you cannot think
This conclusion is perfectly logical and valid but not true.

Deductive logic is easily misapplied: 

       Some students study hard
       Studying hard takes time
       If you don't have time you can't be a student
This is invalid because we are not told that all students study hard.  The qualifier makes a critical difference. 

Inductive Reasoning

Inductive reasoning derives a general conclusion from a finite number of observations.  For example,

       Observation - The people that we know have two legs
       Conclusion - All people have two legs

Inductive reasoning is the common way of drawing inferences from what we know. 

We can disprove any inference by citing just one exception.  We call this falsification.  To disprove the above inference, we only need to introduce one person who does not have two legs.

Wason's Cards

Peter Wason (1924-2003) was a cognitive psychologist who studied why people make consistent mistakes in logical reasoning. 

The Wason Selection Task (Wason 1966) is an example of how we should attempt to determine the truth of an inference.  The task presents the subject with four cards.  Each card has a number on one side and a letter on the other. 

Wason's Cards
Wason's Cards

Which cards would you turn over to test the truth of the statement

If a card has a vowel on one side it has an even number on the other?

A common response is E and 4.  This response relies solely on positive evidence. 

Positive evidence is not sufficient.  To test the truth of the statement, we need to check negative evidence as well.  That is, we need to check if an odd number on one side has a vowel on the other side. 

Problem Solving

Problem solving involves finding a solution to a problem using existing knowledge where the problem is somewhat unfamiliar.  The different theories of how we solve problems include:

  • Gestalt Theory
  • Analogy
  • Problem Space Theory

Gestalt Theory

Behaviorism claims that problem solving is done either by reproducing known responses or by trial and error.  Gestalt theorists considered this approach to be too narrow

Gestalt theory asserts that problem solving has both reproductive and productive components.  Reproductive solving draws on previous experiences and may fixate the subject on known aspects of the problem.  Productive solving involves insight and restructuring of the problem to be solved.

Consider Maier's pendulum: two pieces of string are hanging from the ceiling of a room.  The room has pliers, poles and extensions.  The task is to tie the two strings together.  The strings are too far apart for anyone to catch hold of both at once.  How can a subject solve this problem.

When one of the subjects brushed against one of the strings, a solution suddenly became apparent.  This is an example of productive restructuring.  Insight overcame the fixation on the task. 

Gestalt theory has applications in user interface design to promote more intuitive use.


Analogy is a variation on Gestalt theory.  Analogy involves mapping knowledge from a similar domain to the problem domain.  We sometimes have difficulty identifying the analogy.  We often miss analogous information if it is not sufficiently close to the problem domain. 

Problem Space Theory

Gestalt theory accounts for restructuring but does not explain why or how the restructuring occurs.  Newell and Simon (1972) proposed Problem Space Theory as an alternative to Gestalt Theory. 

Problem Space Theory centers problem solving around a problem space.  The problem space consists of states.  The problem itself has an initial state and a goal state.  Problem states are generated using transition operators.  The task in solving the problem is to generate the goal state and any intermediate states using legal transition operators. 

We use heuristics to select the appropriate operators.  Heuristics include rules of thumb, educated guesses, intuitive judgements and common sense.  One example of a heuristic is straight-forward means-end analysis. 

Consider moving a desk in a cluttered room.  What needs to be done?

Searching the problem space is limited by the speed at which we can retrieve information.  Problem space theory applies in well-defined domains.  In weakly-defined, real-world domains, finding the knowledge itself may be part of the problem. 

Skill Acquisition

Skill differentiates novices from experts in any domain.  The results of experiments on chess layers suggest that novices and experts consider about the same number of moves ahead, but that experts take less time in making decisions and produce better moves.  From these results, it appears that experts remember configurations and the moves associated with them better than novices.  Experts seem to use larger chunks and therefore remember more detail. 

Computer programming is another field that requires considerable skill.  Computer programmers work with chunks or objects.  Experts tend to select better object and organize problems according to deeper conceptual understandings. 


John Anderson (1947-present) is a Canadian-born American psychologist who works in the departments of Psychology and Computer Science at Carnegie Mellon University.  In 1976, inspired by the work of Allan Newell, Anderson proposed a model of skill acquisition called the Adaptive Control of Thought (ACT) model.  ACT identified three distinct levels of skill:

  • the novice learner - uses general purpose rules to interpret facts - slow and demanding on short-term memory
  • the intermediate learner - develops rules specific to each task
  • the advanced learner - tunes the rules to accelerate performance

Proceduralization moves the learner from the first level into the second.  Generalization moves the learner from the second level into the third. 

Consider learning to bake: you start with a general rule of how long the baking pan is to be in the oven.  You subsequently identify specific rules for each type of cake.  Finally you identify general rules that exploit commonalities amongst the variety of different cakes. 

In the late 1980s, Anderson integrated his ACT model with a unifying framework and modified the name to ACT-R, where R stands for rational.  The basic assumption of his rational unification is that cognition is optimally adaptive and precise estimates of cognitive functions mirror statistical properties of the environment.

Mental Models

We build our own models to explain causality within a system.  We call these mental models.  For the most part, these mental models are incomplete, unstable and subject to change.  They can be internally inconsistent and based upon superstition or some incorrect interpretation of the evidence. 

Errors may occur whenever our mental model differs from the actual operation.  Consider a push-button light switch near an elevator push button.  Our mental model may suggest that we push one button rather than the other: turn off the light instead of calling the elevator. 

To interpret the world around us, we copy certain conventions.  Ignoring these conventions may interfere with other people's mental models.  Ideally, designs should respect these conventions. 

An Early Model of Memory

Several models of memory have been proposed since the cognitive revolution of the 1950s and 1960s.  Generally speaking, memory processing occurs in two separate directions:

  • encoding - transferring information into memory
    • shallow
    • elaborative
  • retrieval - transferring information from memory

Encoding proceeds from shallow perceptual features to deeper semantic levels.  Shallow encoding works through repetition.  Elaborative encoding involves association and stores information in deeper structures.  Retrieval transfers information from the deeper structures to shallower ones. 

Early experiments led to a model of human memory that consisted of three abstract stages: 

  • sensory memory
  • short-term memory
  • long-term memory

We call this the modal model of memory or the multi-stage model.  Atkinson and Shiffrin (1968) formalized this model.

organization of memory
(source: Wikipedia 2007 CC-BY-SA)

The modal model of memory offers a straightforward explanation of memory processing.  Stimuli received by our sense organs pass directly to sensory memory.  Some of the information in sensory memory may then pass to short-term memory.  The rest is forgotten.  Some of the information in short-term memory may then pass to long-term memory.  The rest is forgotten. 

Sensory Memory

Sensory memory receives information about 200-500 milliseconds (ms) after some external source has stimulated a sensory organ.  Sensory memory retains this information for several seconds, but no more.  Sensory memory interpret some meaning onto the information received by retaining some of the characteristics of the stimuli.  Sensory memory is outside conscious control, very short-lived, and continually being over-written by incoming signals. 

Our senses are continually being bombarded by fresh information and we have limited capacity to retain that information beyond the duration of sensory memory. 

Evidence of sensory memory dates back to the experiments of Johann Segner (1704-1777) in 1740.  Segner attached a glowing coal to a wheel and spun the wheel at different speeds.  According to him, an object on a disk needs to complete a full circular path in under 100 ms in order to appear as a continuous circle rather than as a sequence of discrete images. 

Here is a very short video on Sensory Memory by Jim Hellas (2008).

Sensory memory is classified into three distinct types:

  1. iconic
  2. echoic
  3. haptic


Iconic memory is visual senesory memory.  It a very rapidly decaying memory of all information arriving from visual stimuli.  It lasts less than 2 seconds.  An example of iconic memory (in addition to Segner's example above) is writing with sparklers. 

We use iconic memory to build up mental representations while reading.  Our eyes fixate on a word for ~300 ms.  Without sensory memory, we would find it difficult to make any sense of what we read. 

Visual memory is not always accurate.  What we expect to see may affect what we perceive. 


Echoic memory is auditory sensory memory.  It is that brief mental echo that continues to sound immediately after the cessation of an auditory stimulus.  Echoic memory can last up to 3 to 4 seconds. 

To demonstrate echoic memory, ask a friend to recite a series of numbers, suddenly stop and ask you to recall the last four.  Notice that you have to repeat the sequence to yourself to identify the last four.  If your friend pauses after stopping, then you should find your recall less accurate. 

Echoic memory lasts just long enough to permit some formation of a mental representation.  Although we only hear sounds instantaneoulsy, information about them is available over this short period of time.  We construct speech flow from a series of passing sounds. 


Haptic memory is the sensory memory of touch.  Our knowledge of this memory is still scarce and a field of ongoing research.  The duration of haptic memory is about 1.3 seconds. 

Short-Term Memory

The information that passes from sensory memory to our short-term memory (STM) is the information that has received our attention.  All other information decays rapidly and is replaced by newly arriving sensory information.  This shallow encoding is local to STM. 

Short-term memory, unlike sensory memory, has quite limited capacity.  Its limited capacity in what distinguishes short-term memory from sensory memory. 

Try this short-term memory test by Eric Chudler (2009).

The Magic Number 7

George A. Miller is an American psychologist who works in hte Department of Psychology at Princeton University and was one of the ofunders of cognitive psychology.  In 1956, he showed that the capacity of short-term memory is a function of the number of chunks processed and not the number of items.  He proposed that we can recall only about 7 chunks of information, plus or minus 2.  Our prior knowledge guides our chunking. 

Which is easier to remember?

       44 113 245 8920

Patterns can also be useful to short-term memory:


Recency Effect

Memory recall may be either serial, parallel, or both.  We remember items at the beginning of a list better.  We remember items at the end of a list better because they are still in short-term memory.  This is called the recency effect.  Placing an irrelevant item at the end of a list eliminates the recency effect.  Adding new items to the end of a list diminishes the recency effect.


Short-term memory is also limited in duration: to about 15-30 seconds.  Without active intervention, we forget information in short-term memory. 

The two possible contributors to short-term forgetfulness are decay and interference. 

STM decays rapidly in the order of 200 ms.  To overcome this time window, we need to reherse the information rehearsed continually. 

Interference occurs through distraction or diversion of attention; for example, a pop-up box requiring some action by the user. 

Long-Term Memory

Long-term memory (LTM), unlike sensory memory and short-term memory, can last for decades. 

Long-term memory is divided into two distinct types:

  • declarative - facts
  • procedural - skills


Declarative memory refers to memory that can be consciously discussed. 

Endel Tulving (1927-present) is an experimental psychologist and cognitive neuroscientist who was a professor at the University of Toronto and who studied declarative memory in depth.  In 1993, he distinguished two types of declarative memory:

  • episodic - biographical events
  • semantic - words, ideas, concepts

Episodic memory holds specific experiences, people, places and events in serial form.  We see ourselves as participants in the events that we remember.  We memorize not only the events but the entire context around those events. 

Semantic memory stores meaning, understanding, and knowledge in structured form and is unrelated to specific events.  It encompasses facts, ideas, abstractions, and principles.  The content is abstract and relational.  Arguably, semantic memories originate from episodic memories that may have become irrelevant and have decayed.  Semantic memory is independent of spatial or temporal context.  We may have a biological predisposition to seek out semantic knowledge for both practical and survival purposes. 

Here is a 5-minute video about Endel Tulving and his contribution.

Semantic networks represent associations between simple items.  For more complex structures, we use frames and scripts.  Frames are like classes in object-oriented design.  Slots in frames allow for specification of attributes.  Scripts model stereotypical knowledge.  They consist of elements that we specify later, such as:

  • entry conditions - must be satisfied to activate the script
  • props - objects involved in events described by the script
  • roles - actions performed by the participants
  • scenes - sequences of events that occur
  • tracks - variations representing alternative scenarios
  • result - is true upon termination of the script

Try this exercise that uses a semantic structure: 20 questions


Procedural memory is non-declarative.  Procedural memory is not easily verbalized.  It holds skills and is associated with different parts of the brain than episodic memory. 

A common model for procedural knowledge is the production system.  A production system consists of rules.  Each rule or production consists of two parts:

  • a sensory-precondition
  • an action

If the precondition matches the current state, we say that the production is triggered.  If the action is executed, we say that the production has fired.  That is, procedural knowledge stores condition-action (if-then) rules.  If information coming into procedural memory matches a condition, the corresponding action fires. 

Working Memory

In the modal model of memory, memory is not unitary but consists of distinct components.  There is no strong scientific evidence to support this hypothesis.  The unity of consciousness concept implies that a unitary model is more realistic.

Alan Baddeley (1934-present) is a British psychologist who works at the University of York as a professor of psychology.  Baddeley and Hitch (1974) proposed an alternative to the modal model.  Their model focuses on the processes involved in short-term memory instead of separating memory into three distinct parts.  We call this the Working Memory or the Working Attention model.  This model has become the dominant model in the field of memory studies.

working memory
Baddeley (1974) Model of Working Memory

A central executive controls and coordinates the activities of working memory.  The executive controls the flow of information to and from its slave systems: the phonological loop and the visuo-spatial sketchpad.  The phonological loop includes a store responsible for to-be-remembered items plus a control process responsible for recoding verbal items into phonological form.  Sound information is refreshed through rehearsal.  The visuo-spatial sketchpad retains visual and spatial information and is involved in mental imagery. 

Working memory is used during reading and performing mental calculations.  It can be accessed quickly, in the order of about 70 ms.

Baddeley (2000) improved this model by adding an episodic buffer.  This third slave system links information across domains to produce integrated units of visual, spatial, and verbal data with chronological order.  This buffer has links to long-term memory and semantical meaning. 

Baddeley's model has been successful in integrating a large amount of findings on memory.


When we believe that a task has been completed, we typically flush working memory to make way for the next task. 

Early ATM machines gave clients their money before returning their bank card, resulting in many clients taking the money and leaving the bank card.  Newer ATM machines return the bank card before giving the client their money. 

Working Memory and Long-Term Memory

Three processes access long-term memory:

  • storing or remembering information
  • retrieving information
  • forgetting


One method of storing information is rehearsal.  The amount of learning is directly proportional to the amount of time spent.  This is called the total-time hypothesis.  Baddeley etal found that learning is more effective if spread over time.  This is called the distribution of practice effect.  Typically, rehearsal may not be enough to store information in LTM.  If information can be associated, is meaningful and familiar, it can be related to existing structures and is more easily incorporated into memory. 


Information retrieval can be of two types: recall and recognition.  Recall is direct reproduction from LTM.  Recognition involves presentation of information as an external cue that the information sought has been encountered before. 


Forgetting can be of two types: decay and interference.  Jost's law states that if two memory traces are equally strong, the older one will be more durable.  Retroactive interference is where new information causes the loss of old information.  Proactive inhibition occurs when old information breaks through and interferes with new information. 


Emotional responses can affect our performance significantly.  A problem that is easy to solve when we are relaxed may be difficult to solve if we are either frustrated or anxious. 

James-Lange Theory

James-Lange theory asserts that emotion is the interpretation of a physiological event (not the cause of a physiological event).  Our experiences generate physiological events and emotions are the feelings that come about as a result of these events. 

Criticisms of this theory include:

  • physiological events are too slow to account for emotional reactions
  • visceral responses are the same no matter what the reported emotion
  • injuries to prefrontal cortex have shown feelings but not responses
  • injuries to somatosensory cortex have shown response without feelings

Cannon-Bard Theory

Walter Cannon (1871-1945) was an American physiologist who taught at Harvard Medical School and coined the term fight or flight response.  That is, animals react to threats physiologically. 

He developed with Philip Bard the Cannon-Bard theory of emotion, which assumes that we react to a stimulus and experience the corresponding emotion simultaneously.  The theory asserts that people feel emotions before acting on them; that is, that physiological changes are caused by emotions, not vice versa. 

For example, "I see a man outside my window.  I am afraid.  I begin to perspire." 

Criticisms include

  • emotion lacks a mechanism

Two-Factor Theory

Stanley Schachter (1922-1997) was an American psychologist who worked at Columbia University.  In 1962, he published the results of a study on 184 college students, which showed that attribute emotion to an evaluation of physical responses in light of the whole situation.  Their two factor theory of emotion states that emotions are a direct result of physiological events and an analysis of the surroundings. 

Criticisms include

  • variations in replicating their results


Generally, our bodies respond to biological stimuli and we interpret that response in some way as a particular emotion.  The biological response is known as an affect

"Negative affect can make it harder to do even easy tasks; positive affect can make it easier to do difficult tasks." (Norman 2002)

Relation to Consciousness

Here is a 6-minute video by Dr. Antonio Damasio on the role of emotions with respect to consciousness.