Descriptive
Research
Observation:
Can you see the behavior?
Is it a sensitive topic?
Do you have a lot of time?
Do you know what you are looking for?
Natural Observation
No intervention
- behavior as it normally occurs
useful for external validity or if ethical concerns about
manipulation
e.g book carrying example
Type 1 - 1 or both arms around books short edges on hip
Type 11- one arm at side long edges horizontal to the
ground
82% type 1 female 3% male
96% type 11 male 16% female
Observation with intervention
- may cause an event that is rare or normally hard to see
- investigate the change in response with stimulus
- establish link between antecedent and consequent behaviors
- manipulate independent variable.
Eg
Simon and Levin - change blindness
Crusco & Wetzel - waitress touch and tip
Participant Observation
- disguised or undisguised
- takes part in the behavior and reports on it
(journalist and anthropology)
eg Rosenhan - sane in insane places
Festinger - cognitive dissonance
Problems
-observer can lose objectivity because they are part of the event
-subject reactivity if known observation (eg stop sign observation)
- subject demand characteristics
- participating may change the event and the outcomes for others
- natural hazards for field work
Bias always a possibility
- observe what you want & fulfill expectations
eg stop observing when nothing happens so over-report
activity
- may change criteria with time
- boredom or fatigue impact
- delay in transcribing may not be exact record
Solutions
Unobtrusive or disguised observation
- or trace information
Habituation to group helps reactivity
‘Blind’ scoring helps bias
Short sessions helps fatigue
Structured Observation
Situation is set up and then participants enter more
controlled environment
-often clinical or developmental lab setting
one way mirror videos. Eg Piaget
Field Experiments
Manipulate IV in natural setting
eg leopard and chimp
Record what??
Goal is to describe behavior as fully and accurately as
possible
Need to use samples of behavior - cannot record
everything.
Typically use
1) checklist of predetermined behaviors (eg Bobo)
2) qualitative narrative
3) quantitative event counts ratings.
Time Sampling
Choose time intervals to record behavior -
systematic, random or both
eg observe children in 2 hours in morning class
may not generalize to whole day so 4 sessions of
30 minutes spread across day better
or 30 minutes randomly chosen
Event Sampling
If event infrequent or long then time sampling not
good -could miss all or part of behavior
So record how often predetermined behavior occurs
eg response to unpredictable events
cannot know when they will occur so timing will miss
eg newcomer to class
Situation Sampling
To improve external validity
observe in different locations, conditions etc
eg animals in captivity or wild
children at home or school
Subject Sampling
If there are many subjects may need to select
some for observation
Observer Reliability
use more than one observer and check inter-observer
reliability - correlation
Improves with clear definitions and training
% time agreement == 100 * number of agreement/
number of opportunities
Quantitative Measures
Typically nominal scale (categories counted)
eg book carrying example
Rating common
Interval and ratio less common
personal space
duration
Survey research
Mail survey
- quick efficient convenient
- no response bias a problem (60% + response
acceptable 30% typical)
- participants answer in any order
Written questionnaires generally need to be self-
explanatory and short
Written Questionnaire
- simple questions best
- be specific
- no double questions (eg have you experienced
headache and nausea in last week?)
- no double negatives ( if you have not had no headaches
in last week….)
- no ambiguous questions
- no leading questions
- short
- readable
- order can matter general to specific is usually best
Often rank scale
Likert scale 1-7
Specific categories best
Do you get headaches frequently rarely etc…
better choice
headaches 0-1 2-3 3-4 times a week
people may define terms like rare or often differently
Interviews
More open ended complex questions and follow up
questions possible
Order of questions controlled
phone - relatively easy but no-response bias and
selection problems (unlisted numbers etc)
in person- very time consuming and expensive
Problems recording bias and leading questions
Designs
Cross sectional design
one or more samples drawn at same time
Independent samples successive
same survey different times
eg presidential approval rating
Longitudinal
same respondents surveyed over time
huge effort to track people