Science in Context

1
Science in Context
Experimental Ecology
Subject Effects
The Hawthorne Effect
- the performance of subjects may be altered by
the simple fact that they are aware that they are
taking part in some kind of research
2
Science in Context
Experimental Ecology
Subject Effects
economic considerations
- what do the subjects get in return for
participating in the research?
- money
- grades
- access to certain treatments
- will subjects present inaccurate pictures of
themselves just to get into the study?
3
Science in Context
Experimental Ecology
Subject Effects
In general, it is important to consider why
subjects are doing what they do...
- in response to the experimental conditions
- in response to the experimenter
- in response to some idea in their own heads
4
Science in Context
Experimental Ecology
Subject Effects
placebo effects
- an effect produced by the subjects belief that
the manipulation will have an effect
- most obvious in cases where no actual
manipulation was performed (e.g., subjects in no
treatment control groups)
5
Science in Context
Experimental Ecology
Subject Effects
placebo effects
- the coronary bypass placebo
- subjects reported a decrease in angina (chest
pain) following sham surgery
6
Science in Context
Experimental Ecology
Subject Effects
placebo effects
- the best defense against placebo effects is to
perform double-blind studies
1. person getting the treatment doesnt know if
it is a placebo or the real thing
2. person giving the treatment and/or collecting
the data doesnt know if the subject gets a
placebo or the real thing
7
Science in Context
Experimental Ecology
Subject Effects
demand characteristics
- are subjects responding to the type of
experiment or setting rather than the
manipulation of the IV?
- e.g., studies of the effects of recreational
drugs
- quasi-control groups
8
Quasi-Experimental Designs
Closed and Open Systems
Cause and effect interpretations require a high
level of experimental control
- a situation where we control all the relevant
variables is a closed system
- e.g., the operant chamber
9
Quasi-Experimental Designs
Closed and Open Systems
Closed systems are suitable for the various kinds
of experimental designs previously discussed
- possible to determine cause and effect
- possible to isolate the effects of individual
variables
- external validity is always an issue
- a truly closed system may be difficult or
impossible for us to create
10
Quasi-Experimental Designs
Closed and Open Systems
We may be interested in the effects of variables
that we cannot control for practical or ethical
reasons
- e.g., studies of the effects of the Loma Prieta
earthquake discussed in text
- such studies are taking place in open systems,
where many of the relevant variables are not
controllable by researchers
11
Quasi-Experimental Designs
Closed and Open Systems
Open systems will never allow us to make cause
and effect statements, but they can still provide
useful information
- reliance on logical rather than statistical
interpretation of the data
- e.g., Robbers Cave study
- may lead to more controlled experimental designs
12
Quasi-Experimental Designs
Time Series Designs
Single-Group Pretest-Posttest designs
- essentially a within-subjects design with no
control group
- e.g., a study on the effects of changing the
diet on the general health of nursing home
residents
- may be no practical/ethical way to create a no
treatment control group
13
Quasi-Experimental Designs
Time Series Designs
Single-Group Pretest-Posttest designs
- a change in health status may reflect the
effects of a change in diet
- or it may reflect a change that would have
happened anyway for some other reason unrelated
to diet
- it would be useful to know how the health of
these people tended to change over a longer
period of time
14
Quasi-Experimental Designs
Time Series Designs
Interrupted Time Series design
Some of these issues can be addressed by taking
multiple measurements both before and after the
manipulation
- may give a much more accurate picture of the
normal range of variation in the variable of
interest
- allows discrimination of the effects of our
manipulation from normal variation
15
Quasi-Experimental Designs
Time Series Designs
Multiple Time Series design
- inclusion of control group
- multiple pretest and posttest measures for both
groups
- experimental group get the manipulation,
control group doesnt
- how is this critically different from the
between-subjects experimental designs already
discussed?
16
Quasi-Experimental Designs
Time Series Designs
Non-Equivalent Before-After design
- inclusion of control group
- single pretest and posttest measure for both
groups
- DV is the difference score when before and
after conditions are compared
- may be useful if you have strong reason to
believe the 2 groups are different to start with
17
Quasi-Experimental Designs
Time Series Designs
Retrospective / Ex Post Facto designs
- useful when one or more events of interest are
in the past
- ex post facto designs specifically look at the
relationship between a past event and a current
situation
- other relationships among past events may be
examined through archival materials, etc...
18
Correlational Designs
- determine the nature of the relationship (if
any) between 2 variables of interest
- may or may not actually involve the calculation
of a correlation coefficient
- data may come from present, past, or both
- no experimental control, so third variable
problem is always an issue
- no cause and effect explanations are possible
19
Naturalistic Observation
- wont tell us about cause and effect or even
relationships
- can tell us if our ideas (e.g., about human
behavior) match up with the real world
- can tell us about things we were previously
unaware of
- reactive behavior can be an issue (similarity
to Hawthorne Effect)
20
Naturalistic Observation
Reactive Behavior observed behavior is altered
by the knowledge that observation is taking place
- if behavior is highly reactive, it may be
better to appear as a participant than as a
detached observer
- e.g., Rosenhans study of the treatment of the
insane
21
Naturalistic Observation
Reactive Behavior observed behavior is altered
by the knowledge that observation is taking place
- will your participation alter the behavior of
others?
- will your participation alter your ability to
make unbiased observations?
- will your participation allow you to better
understand the situation by experiencing it
yourself?
22
Naturalistic Observation
The naturalistic approach has been extremely
useful in many fields
- e.g., animal behavior (Tinbergen, Lorenz, et
al.)
- e.g., treatment of psychopathology (Rosenhan)
23
Naturalistic Observation
Data analysis, however, can become complicated
- unobtrusive observation will not interfere with
the observed behaviors...
...but it also means very little control can be
exercised
- much of the data may be either redundant or
incomplete
24
Single Subject Designs
All the data come from a single individual, so
issues relating to sample selection, calculating
group means, etc... do not apply
- many instances of single subject designs in
early psychological research (Ebbinghaus, Weber,
Pavlov, etc...)
- 2 basic types of design descriptive and
quasi-experimental
25
Single Subject Designs
Naturalistic Case Studies (descriptive)
- the extensive examination and description of an
individual and their circumstances
- widely used as a starting point for the
investigation of a phenomenon
- many famous case studies in psychology
- Anna O.
- H.M.
- Phineas Gage
26
Single Subject Designs
Naturalistic Case Studies (descriptive)
- relevant information in a case study may come
from several sources
- observation of subject
- interviews with subject
- interviews with friends relatives
- institutional records
27
Single Subject Designs
Naturalistic Case Studies (descriptive)
The One-Shot Case Study
- this approach generally involves following an
individual through the course of some treatment
and reporting on the effects (if any)
- may be followed up using more controlled
experimental methodology
28
Single Subject Designs
Quasi-Experimental Single Subject Designs
We may want to observe the effects of some
manipulation on an individual subject
- time series approach
- replication intrasubject vs intersubject
29
Single Subject Designs
Quasi-Experimental Single Subject Designs
the reversal design
- baseline and treatment conditions are
alternated (usually at least 2 cycles)
- multiple cycles can help determine if behavior
really shifts in response to the manipulation,
not just due to some uncontrolled variable
- may be referred to as ABAB design
30
Single Subject Designs
Quasi-Experimental Single Subject Designs
the reversal design
31
Single Subject Designs
Quasi-Experimental Single Subject Designs
the multiple baseline design
- in some cases we may be interested in a number
of different behaviors in a single subject
- e.g., multiple disruptive behaviors in an
institutionalized person
32
Single Subject Designs
Quasi-Experimental Single Subject Designs
the multiple baseline design
- first, a baseline measure of all the behaviors
is taken
- then treatment for a specific behavior is
given, and additional baseline measures of the
other behaviors are taken
33
Single Subject Designs
Quasi-Experimental Single Subject Designs
the multielement design
- in some cases we may be interested in a number
of different treatments in a single subject
- similar to the reversal design, but generally
involves multiple reversals
- may involve gt2 treatments (some
counterbalancing may be required)