Observation (cont)
 
 
 
 

Physical responses

• Eye movements
• Galvanic skin responses (GSR)
• Electro encephalograph (EEG)
• Evoked potentials (ERP)
• PET/CAT/fMRI
• Habituation/High amplitude sucking
• Infants and animals
• Categorical perception of phonemes (Eimas et al.)

More Measures

• Reaction time/Response time
• Speeded judgments
• Speeded responses (same/different)
• Naming
• Problem Solving Time
• How long to solve an insight problem?
• Reading times
• Ratings and scales

Compare to more open-ended measures

• Open-ended measures generate a lot of data
• Think-aloud protocols
• Perform some task and think aloud
• Towers of Hanoi task
• Much like narrative methods
• Requires data coding
• Requires measures of inter-rater reliability
• Allows access to hidden mental behavior
• Relies on introspection
• Dangerous to rely too heavily on protocols

Sampling Behavior

• You cannot observe everything
• What will you choose to observe
• Need some method for sampling behavior
• Effects of sampling
• Influences representativeness of observations
• Influences ecological validity
• Influences quality of the data

Types of Sampling

• Time sampling
• Event sampling

Time Sampling

• · Select particular times to observe
• Random times
• Studies with beepers:  How are you feeling now?
• Fixed times
• Diary studies:  entries written at a fixed time of day
• Strengths and weaknesses
• May capture a range of situations
• May miss infrequent events

Event Sampling

• When an event is rare, time sampling may not work.
• Event sampling
• Observe only occurrences of a particular event
• Observe behavior of EMTs only during accidents
• Strengths and weaknesses
• Focuses on events that are rare
• Is the event easily defined?

Ways of collecting Data.

• Observation without intervention
• Naturalistic observation
• Physical traces
• Archival data
• Observations are still theory-laden
• Observer just does not interfere
• Observation with intervention
• Experimental manipulations
• Participant observation
• Disguised observation
• Undisguised observation

Some Issues

• How "ecologically valid" are the observations?
• Jury Data
• Interval vs Nominal scale
• What influence does the observer have on the situation?

Naturalistic Observation

• · Anyone can just observe
• How is scientific observation different from nonscientific observation?
• Precision
• Definition of conditions
• · Why perform naturalistic observations?
• Observe behavior in naturalistic settings
• Ecological validity
• Avoid ethical problems of causing noxious conditions on purpose
• Feral children/Children raised in isolation
• Kitty Genovese case (stabbing)

• Disguised Observation
•  Experimenter "infiltrates" group and observes
• Behavior of a cult
•  Ethical problems with disguised observation
• Clearly there is deception here
• Group members will treat experimenter as one of their own
•  Experimenter may (unwittingly) adopt attitudes of the group
• Experimenter bias and expectancy
• May inadvertently push the group
• Suppose the cult wanted to commit a crime?
• Should experimenter talk them out of it?
• Should experimenter prod them into it?
• Can experimenter shape the behaviors of the group?

• Observer follows the group around
•  Following the inner working of a political campaign
• May be some reactivity
• Anthropology and cultural psychology use these methods
•  Studies of color names
• Studies of folk botanical categories

Problems with Naturalistic Observation

• Interesting behaviors may occur infrequently
• Some events do not occur in public
• Hard to examine processes during observation
• No control over circumstances
• Credit assignment problem
• Experimental control
• e.g., third variable problem
• A correlated with B does not mean A causes B or vice versa
• instead of A -> B, A <- C -> B
• children in day care a lot have attachment problems
• maybe the parents can't stand their kids
• Other examples? Rank and Age (page 23 of book).

How would you make this into a true experiment?
 
 
 
 

Directionality Problem

• Violent Television <-> Violent Kids

• Most psychological research involves intervention
• Can cause infrequent events
• Simulated emergencies
• Investigate the limits of an ability
• Often we try to "break" a behavior
• Where's Waldo
• Observe normally private events

• Allows comparisons with different groups (that you create)
• Can vary the settings explicitly
• Can assign conditions to people
• Eases the credit assignment problem
• Two groups differing only in a single factor
• Allows repeated observation of the same behavior

More controlled settings (no cult infiltration)
 

•  Our primary concern in this class will be experimental settings
• The experimenter will manipulate the world
• We want to set up different situations that are as similar as possible
• We want to be able to eliminate alternative explanations for behaviors
• Techniques for controlled experimental studies will unfold in the coming weeks

Example: Diffusion of Responsibility
 

• Naturalistic Observation
•  Kitty Genovese case
• Woman stabbed repeatedly
• No bystander intervention
• The event is both intriguing
• Cannot know why nobody intervened here
• Possibilities?
•  Test by using a controlled experiment

• Proposed a diffusion of responsibility explanation
• The more people present, the less likely that one will help
• How could this be operationalized?
• Perceived group size (in a conversation)
• 2:  Subject and "victim"
• 3:  2 Subjects and "victim"
• 6:  5 Subjects and "victim"

Experiment Lingo-Variable Types

• Independent Variables
• A variable that is controlled by the experiment
• e.g., drugs a rat gets
• levels
• e.g., how many c.c.s of the drug
• treatments (non-control groups)
• Extraneous Variables
• Not interested in them, but they could affect the results
• e.g., Saline injection for added control
• Dependent Variable
• The actual measure for the stats/results
• e.g., time to complete the maze

Internal Validity

• Does your experiment test what it's supposed to?
• i.e., Did changes in the independent variable cause the changes in the dependent variable?
• Extraneous variable (i.e., confounds) are a threat to internal validity
• e.g., time of day, different experimenters, etc..

• history
• multiple observations -> learning, prev items
• maturation
• e.g., fatigue
• instrumentation
• e.g., drifting scales, poor measurement
• statistical regression
• selecting based on scores (see lecture 1)
• biased selection of subjects
• always use random assignment
• mortality
• dropping out of a study

External Validity

• Why do I care about rats?
• Are UT students representative?
• Can I generalize?
• Is there a point to my research?