Sie sind auf Seite 1von 38

Experimental Research 1

Suppose teachers wished to determine which of two methods of
reading instruction was most effectiveone that involved 20 minutes of
direct instruction in phonics each day throughout the academic year in
grade 1 or one that involved the current practice of having the teacher read
a book to the class for 20 minutes each day throughout the year in grade
one. Similarly, suppose they wished to determine whether children learn
better in a small class (i.e., with 15 students) or a large class (i.e., with 30
students). Finally, suppose they wished to determine whether requiring
students to take a short quiz during each meeting of a college lecture class
would result in better performance on the final exam than not giving
The example above is an implementation of experimental research.
Therefore, this paper will discuss more about it in order to make us
understand about this kind of research.
1. What are the definition and the purpose of experimental research?
2. How many kinds do experimental research have and how to apply
3. What are their benefits and their infirmity?
1. To know the definition of experimental research and its purpose.
2. To know the kinds of experimental research and how to apply them.
3. To know the benefits and the infirmity of experimental research.

Experimental Research 2

Experimental research is the only type of reseacrh that can test
hypotheses to estabilsh cause-effect relationships. It represents the strongest
chain of reasoning about the links between variables. You may recall from
chapter 1 that in experimental research the reseacrher manipulates at least one
independent variable, control over relevant variables and observes the effect
on one or more dependent variables is the one characteristic that differentiates
experimental research from other types of research. The independent variable,
also called the treatment, clausal, or experimental variable, is that treatment
or characteristic be lieved to make a difference. In educational research,
independent variables that are frequently manipulated include method of
instruction, type of reinforcement, arrangement of learning environment, type
of learning materials, and length of treatment. This is list by no means
exhaustive. The dependent variables, also called the criterion, effect, or
posttest variable, is the outcome of the study, the change or difference in
groups that occurs as a result of the independent variable. It gets its name
because it is dependent on the independent variable. The dependent
variable may be measured by a test or some other quantitative measure ( e.g,
attendence, number of suspension, time on task ).the only restriction on the
dependent variable is that it represents a measurable outcome.
Experimental research is the most structured of all reserach types.
When well conducted, experimental studies produce the soundest evidence
concering cause-effect relationships. The results of experimental research
permit prediction, but not the kind that is characteristic of correlational
research. A correlational study predicts a particular score for a particular
individual. Predictions based on experimental findings are more global and
often take the form if you see approach X, you will probably get better
Experimental Research 3

results than if you see approach Y of coures, it is unusualy for a singel
experimental study to produce broad generalization of results, because any
single study is limited in context and participants. However, replications of a
study using different contexts and participants often produce cause-effect
results that can be generalized widely.
The Experimental Process
The step in an experimental study are basically the same as in other types
of research: selecting and difining a problem, selecting participants and measuring
instruments, preparing a research plan, executing procedures, analyzing the data,
and formulating conclusion. An experimental study is quided by at least on
hypothesis that states an expected causal relationship between two variables. The
experimental is conducted to confirm ( support) or disconfirm ( refute ) the
experimental hypothesis. In an experimental study, the researcher will go to the
action from the very begining. He selects the groups, decides what treatment will
go to which group, controls extraneous variables, and measure the effect of the
teratment at the and of the study.
It is important to note that the experimental researcher controls both the
selection and the assignment of the research participants. That is the researcher
randomly selects participents into the different treatment conditions. It is the
ability to randomly select and randomly assign of participants to treatments,
which is also called the researchers manipulation of the treatments, is the
distriquishing aspect of experimental research and the feature that distinguishes it
from causal-comparative research has only random selection, not assignment,
because causal-comparative participants are obtained from two already-existing
populations. There can be no random assignment to a treatment from a singel
population in causal-comparative studies.
The group that receives the new treatments is often called the experimental
group, and the group that receives a different treatment or is treated as usual is
called the control group. An alternative to using these is to simply describe the
treatments as comparison groups, treatments groups, or groups A and B. The
Experimental Research 4

trems are used interchangeably. A common misconception is that a control group
always receives no treatment. This is not true and would hardly provide a fair
comparison. For example, if the independent variable was type of reading
instruction, the experimental group might be instructed with a new method, and
the control group might continue, instruction with the currently used method. The
control group would still receive reading instruction, members would not sit in a
closet while the study was being conducted. Otherwise, you would not be
evaluating the effectiveness of a new method as compared to that of a traditional
method, you would be comparing a new method to no reading instruction at all.
Any method of instruction is bound to be more effective than no instruction.
The group are not to receive the difference treatments should be aquated on
all variables that might influence performence on the dependent variable. For
example, in the previous example, initial reading readiness should be very similar
in each treatment group at the star of the study. The researcher must make every
effort to ensure that the two group start as equivalently as possible on all variables
except the independent variable. The main way that groups are equated is through
simple random or stratified random sampling. After the groups have been exposed
to the treatment for some period, the researcher collects data on the dependent
variable from the groups and determines whether there is a real or significant
difference between their perfomance. In other words, using statistical analysis of
experimental studies in detail. For now, suppose that at the and of an experimental
study one group han an average score of 29 on the dependent variable and the
other group had an average score of 27. There clearly is a different between the
groups but is a 2 point difference a meaningful or significant difference, or is it
just a chance difference produced by meansurement error. Statistical analysis
helps answer this question.
Experimental studies in education often suffer from two problems : a lack of
sufficient exposure to treatments and failure to make the treatments substantially
different from each other. On most cases, no metter how effective a treatments is,
it is not likely to be effective if students are exposed to it for only a brief period.
To adequantely test a hypothesis concerning the effectiveness of a treatment, the
Experimental Research 5

experimental group would need to be exposed to it over a period of time so that
the treatment is given a fair chance to work. Also of concren is the difference
between treatments. In a study comparing team teaching and traditional lecture
teaching, it would be vital that team teaching be operationalized in a manner that
clearly differentiated it from the traditonal method. If team teaching meant two
teachers taking turns lecturing, it would not be very different from traditional
teaching and the researcher would be very unlikely to find a meaningful
difference between the two study treaments. Also, if teachers using different
treatments converse with and borrow from each others treatments, the original
treatments become diluted and similar to each other. These problems have
detrimental effects on the outcome of the study.
Manipulation and control
Direct manipulation by the researcher of at least one independent variable
is the one single characteristic the differentiates experimental research from other
types of research. Manipulation of an independent variable is often a diffcult
concept to grasp. Quite simply, it means that the research decides what treatments
will make up the independent variable and which group will get which treatment.
For example, if the independent variable was number af annual teacher reviews,
the researcher might decide to from three groups : one group receiving no review,
a second group receiving one review, and a third group receiving two reviews. In
addition, having selected research participants from a single, well-defined
population, the researcher would randomly assign participants to treatments. Thus,
manipulation means being able to select the number and type of treatments and to
randomly assign manipulation treatments.
Independent variables in education are ethier manipulated ( active variable )
or not manipulated ( assigned variables ). You can manipulate such variables as
method of instruction, number of reviews., and size of group. You cannot
manipulate variables such as gender, age, or socioeconomic status. You can place
participants into one method of instruction or another ( active variable ), but you
cannot place participants into male or female categories because the already are
Experimental Research 6

male or famale ( assigned variable ). Althought the design of an experimental
study may or may not include assigned variables, at least one active variable must
be present.
Control refers to the researchers efforts to remove the influenceaf any
variable other than the independent variable that might affect performance on the
dependent variable. In other words, the researcher wants the groups to be as
similar as possible, so that the only major difference between them is the
treatment variables as manipulated. To illustrate, the the importance of research
control, suppose you conducted a study to compare the effectiveness of student
tutors versus parent tutors in teaching first grades to read. Students tutors might be
older children from higher grade levels, and parent tutors might be members of
the PTA. Suppose also that student tutors helped each member of their group for 2
hours per week for a month, would the comparison be fair ? certainly not.
Participants with the students tutors would have received 2 times as much help as
that provided to the parents group. thus, one variable that would need to be
controlled would be amount of tutoring . if this variable were not controlled, you
could be confronted with a dilemma. If the students tutors produced higher
reading scores that the parent tutors, you would not know whether this result
indicated that students tutors were more effective that parent tutors, that longer
periods of tutoring were more effective than shorter periods, or that type and
amount of tutoring combined were more effective. To make the comparison fair
and interpretable, both students and parents would tutor for the same amount of
time. Then time of tutoring would be controlled, and you could truly compare the
effectiveness of students and parent tutors.
A reseracher must consider many factors when attempating to identify and
control extraneous variables. Some variables that need controlling may be
relatively obvious as the reseracher in the preceding study, you would need to
examine such variables as reading readiness and prior reading instruction in
addition to time spent tutoring. Some variables that need to be controlled may not
be as obvious, for example you would also need to ensure that both groups used
similar reading texts and materials. Thus, there are really two different kinds of
Experimental Research 7

variables that need to be controlled, Participants variables and environmental
variables. A participants variable is one on which participants in different group in
a study might different enviromental variable ( such as learning materials ) is a
variable tn the setting of the study that might cause unwanted differences between
groups. The researcher strives to ensure that the characteristics and experincess of
the groups are asequal as possible on all important variables excepted the
independent variable. If relevant variables can be controlled, group differences on
the dependent variable can be attributed to the independent variable.
Control is not easy in an experiment, especially in educational studies, where
human beings are involved. It certainly ia a lot easier to control solids, liquids and
gases our task is not an impossible one, however , because we can concentrate on
identifying and controlling only those variables that might really affect or interact
with the dependent variable. An interaction occurs when different values of the
independent variable are differentianlly effective depending upon the level of the
control variable. For example, if two groups had significant differences in shoe
size or height, such differences would probably not affect the results of most
educational studies. Techniques for controlling those extraneous variables that do
matter will be presented later in this chapter.
Threats to experimental validity
As noted, any uncontrolled extraneous variables effecting performance on the
dependent variable are threast to the vadility of an experimental. An experimental
is valid if results obtained are due only to the manipulated independent variable
and if they are generalizable to individuals or contexts beyond the experimental
setting. These two criteria are referred to, respectively, as the internal validity and
external validity of an experiment.
Internal validity is the degree to which observed differences on the dependent
variable are a direct result of manipulation of the independent variable, not some
other variable. In other words, an examination of internal validity focuses on
threast or rival explenations for thr research results would have been the
differences in the amount of time the two groups tutored. The degree to which
Experimental Research 8

experimental research results are attributable to the independent variable and not
to some other rival explanation is the degree to which the study is internally valid.
External validity, also called ecological validity, is the degree to which study
results are generalizable, or applicable, to groups and environments outside the
experimental setting. In other groups, an examination of external validity focuses
on threast or rival explenations that would not permit the results of a study to be
generalized to other setting or groups. A study conducted with groups of gifted
ninth graders, for example, should produce results that are applicable to other
groups of gifted ninth grades. If research results were never generalizable outside
the experimental setting, then no one could profit from research. Each and every
study would have to be reestabished over and over. An experimental study can
contribute to educational theory or practice only if its results and effects are
replicable and generalize to other places and groups. If results cannot be replicated
in other settings by other researchers, the study has low external, or ecological,
So, all has to do in order to ocenduct a valid experiment is to maximize boyh
internal and external validity. To maximize internal validity, the researcher must
exercise very right controls over participants and conditions, producing a
labolatory like environment. However the more a research situation is narrowed
and controlled, the less realistic and generizable it become. A study can
contributed little to educational practice if its technique, which have been proven
effective in highly controlled setting, will not also be effective in a less controlled
classroom setting. On the other hand, the more natural the experimental setting
becomes, the more difficult it is to control extraneous variables. Thus, the
researcher must strive for balance between control and realism. If a choice is
involved, the researcher should err on the side of control rather than realism since
a study that is not internally valid is worthless. a useful startegy to address this
problem is to first demonstrated an effect in a highly controlled environment (
with maksimum internal validity ) and then redo the study in a more natural
setting ( to examine external validity ). In the final analysis, however the
researcher must seek a compromise between a highly controlled and highly
Experimental Research 9

natural environment. In the following pages we describe many threats to internal
and external validity. Some extraneous variables are threast to internal validity,
some area threast to external validity, and some may be threast to both. How
potential threast are classified is not of great importance what is important is that
you be aware of their existence and how to control for them. As you read, you
may begin to feel that there are just too many threast for one little researcher to
control. However, the tast is not formidable as it may at first appear, since there
are a number of experimental designs that do control many or most of threast you
are likely to encounter. Also, remember that each threat is a potential there only it
may not be problem i a particular.
Threasts to internal validity
Probably the most authoritative source on experimental design and threasts to
experimental validity is the work of donald campbell and julian stanley, and
thomas cook and donald campbell. They identifild eight main threast to internal
validity, History, maturation, testing, intrumentation, statistical regression,
diffrential selection of participants, mortality, and selection-maturation intrection.
However, before decribing these threast to internal validity, we would like to note
the role of experimental research in overcoming these threasts. you are not
rendered helpless when faced with them, quete the contary, the use of random
selection of participants, the researcher assignment of participants to treatments.
And control of other variables are powerfull approaches to overcaming the threast.
As you read about the threast, note how experimental researchsrandom selection
and assignment to treatments can control most threast.
When discussing threats to validity, history refres to any event occurring during a
study that is not part of the experimental treatment but may affect the dependent
variable. The longerv a study lasts, the more likely it is that history will be a
threat. A bomb scare, an epedemic of meales, or even general current events are
examples of events that could produce a history of fect. For example, suppose you
conducted a series of in serves workshops designed to in crease the morale of
Experimental Research 10

teacher participants. Between the time you conducted the workshops and the time
you administered a posttest measure of morale, the news media announced that,
due to state level budget problems, funding to the local school district was going
to be significantly reduced and promised pay raises for teachers would likely by
postponed. Auch as event could easily wipe out any effect the workshops might
have had, and postest morale scores might well be considerably lower than they
otherwise might have been (to say the least).
Maturation refres to physical, intellectual, and emotional changes that naturally
occur within individuals over a period of time. In a research study, these changes
may affect participants performance on a measure of the dependent variable.
Especially in studies that last a long time participants may become older, more
coordinated, unmotivated, anxious or just plain bored. Maturation is more likely
to be a problem in a study designed to test the effectivenees of a physochomotor
training program on 3 year than in a study designed to compare two method of
teaching algebra. Young participants would typically undergoing rapaid
biological changes during the training program, raising the question of whether
changes were due to the training program or to maturation.
Testing, also called pretest sensitization, refres to the threat of improved
performance on a posttest being a result of having taken a pretest. Taking a pretest
may improve participants scores on a postest, regardles of whether they received
any treatment or intructions in between. Testing is more likely to be a threat when
the time between testing is short, takin a pretest taken in september is not likely to
affect performance on a posttest taken in june. The testing threat to internal
validity is more likely to occur in studies that measure factual information that can
be recalled. For example, taking a pretest on algebraic equations is less likely to
improve posttest performance that taking a pretest on multiplication facts would.

Experimental Research 11

The instrumentation threast refres to unreliability or lack of consistenciy, in
measuring instruments that may result in an invalid assessment of peformance.
Instrumentation may threaten validity in several different ways. A problem may
occur if the researcher uses two different test, one for pretesting, and the tests are
not of equal diffiulty. For example, if the posttest is more difficult than that
pretest, it may mask improvement that is actually present. Alternatively, if the
posttest is more difficult than the pretest, it may mask indicate improvement that
is no really present. If data are collected through observation, the obseverses may
not be observing or evaluating behavior in the same way at the end of the study as
at the beginning, in fact, if they are aware of the nature of the study, they may
see and record only what they know the researcher is hypothesizing. If data are
collected through the use of a mechanical device, the device may be poorly
calibrated, resulting in inaccurate measurement. Thus, the researcher must take
care in selecting tests, observers, and mechanical devices to measure the
dependent variable.
Statisticul regression
Statistical regression usually occurs in studies where participants are selected
on the basis of their extremely high or extremely low scores. Statistical regression
is the tendency of participants who score highest on a test ( a pretest ) to score
lower on a second, similar test ( posttest ), and of subjects who score lowest on a
posttest. The tendency is for score to regress, ( move lower ) toward a mean (
average ) or expected score. Thus, extremely high scorers, or move, toward the
mean, and extremely low scorers regress ( move higher )toward the mean.
Differential Selection of Participants
Differential selection of participants is the selections of subjects who have
differences before the start of a study that may at least partially account for
differences found in a posttest. The theart that the groups were different before the
study even began is more likely when a researcher, is comparing alread-formed
Experimental Research 12

Mortality or attrition, refres to a reduction in the number of research
participants that occurs over time as individuals drop out of the study. Mortality
creates problems with validity particularly when different groupsmdrop out for
different reasons and with different frequency the change in the characteristic of
the group due to mortality can have a significant effect on the results of the study.
For example, participants who drop out of the study may be lest motivated or
uninterested in the study than those who remain. This is especially a problem
when volunteers are used or when a study compares a new treatment to an
existing treatment. Participants rarely drop out of control groups or exisiting
treatments because few or no additional demands are made on them. However,
volunteers participants using the new, experimental treatment may drop out
because too much effort is required for participation. The experimental group that
remains at the end of the study then represents a more motivated group than the
control group. As another example of mortality, suppose suzy shiningstar (a high
IQ and all the student) got the measles and dropped out of your control group.
Before suzy dropped out, she managed to infect her friends might also be the
high IQ and all that type students. The experimental group might end up looking
pretty good when compared to the control group simply because many of the food
students dropped out of the control group. The researcher cannot assume that
participants drop out of a study in a rendom fashion and should, if possible, select
a design that controls for mortality.
A researcher can assess the mortality of groups by obtaining demographic
information about the participants groups before the star of the study and the
determining if the makeup of the group has changed at the end of the study. One
wey to reduce mortality is to provide some incentive to participants to remain in
the study. Another approach is to identify the kinds of participants who drop out
of the study and remove similiar portions from the other groups.
Selection-Maturation Interaction and Other Interactive Effects
Experimental Research 13

The effects of different selection may also interact with the effects of maturation,
history, or testing to cause a threat to internal validity. What this means is that if
already-formed groups are used, one group may profit ( or less ) from a treatment
or have an intial advantage ( or disadvantages ) because of maturation, history, or
testing factors. The most common of these interactive effects is selection-
maturation interaction, which would exist. Of participants selected into the
treatment groups matured at different rates during the study. To get a better idea
of these effects, suppose, for example, that you received permission to use two of
ms. Hynees english classes and that both classes were average and apparently
equivalent on all relevant variables. Suppose, however, that for some reasons ms.
Hynee had to miss one of her classes but not the other ( maybe she had to have a
root canal ) and ms. Almamater took over ms. Hynees class. As luck would have
it, ms. Mater proceeded to cover much of the material now included in your
posttest ( remember history ). Unbeknownst to you, your experimental group
would have a definite advantage to begin with, and it might be this initial
advantage, not the independent variable, that caused posttest differences in the
dependent variable. Thus, a researcher must select a design that controls for
potential problems such as this or make every effort to determine if they are
operating in the study.
Threats to external validity
There are several major threats to external validity that can limit generalization
of experimental results to other populations. Building on the work of campbell
and stanely, bracht and glass refined and expanded discussion pf threast to
external validity. Bracht and glass clssified these therast into two categories.
Threast effecting generalizing to whom that is, threat affecting the groups to
which research results be generalizing make up to population validity. Threats
affecting generalizing to what that is threast affecting the settings, conditions,
variables, and contexts to which results can be generalized make up threast to
ecological validity. The following discussion incorporates the contributions of
bracht and glass into campbell and stanleys original ( 1971 ) conceptualizations.
Experimental Research 14

Pretest-treatment interaction
Pretest-treatment interaction occurs when participants respond or react differently
to a treatment because they have been pretested. Pretesting may sensitize or alert
subjects to the nature of the treatment, potentially making the treatment effect
different than it would have been had subjects not been pretested. Thus, the
research results would be generalizable, only be other pretest groups. The results
would not even generalizable to the unpretested population from which the
sample was selected.
The seriousness of the pretest-treatment interactions threat depends on the
research participants, the nature of the independent and dependent variables, and
the duration of the study. Studies involving sell-report meausre, such as attitude
scales and interes inventories, are especially susceptible to this threat. Campbell
and stanley illustrate this effect by pointing out the probable lack of comparability
of between two groups : one that views the antiprejudie film gentlemans
agreement right after taking a lengthy pretest dealing with anti-semitism and
another that views the move without a pretest. Individuals not pretested could
conceivably enjoy the move as a good love story and be unaware that it deals with
a social issue. Pretested individuals would be much more likely to see a
connections between the pretest and the massage of the film. In contrast, taking a
pretest on algebraic algorithms would probably have very young children, who
would probably not see or remember a connection between the pretest and the
subsequent treatment. Similarly, for studies conducted over a period of months or
longer, the effects of the pretest would probably have wron off or been greatly
diminished by the time a posttest was given. Thus, for some studies the potential
interactive effect of a pretest ia a more serious consideration than others. In such
cases, researchers should make use of unobtrusive measure, ways to collect data
that do not intrude on, or require interaction with, research participants. For
example, data can be gathered from school records, transcripts, and other written

Experimental Research 15

Multiple-treatment interference
Sometimes the same research participants receive more than one treatment in
succession. Multiple-treatment interference occurs when carryover effects from an
earlier treatment make it difficult to assess the effectiveness of a later treatment.
Suppose you were interested in comparing two different approaches to improving
classroom behavior modification and corporal punishment ( admittedly an
extreme example used to make a point ). For 2 months, behavior modification
techniques were systematiclly applied to the particapants, and at the end of this
period you found behavior to be significantly better than before the study began.
For the next 2 months, thne same participants were physically punished ( with
hand slappings, spankings and the like ) whenever they misbehaved, and at the
end of the 2 months behavior was equally as good as after the 2 months of
behavior modification. Could you then conclude that behavior modification and
corporal punishment are equally effective methods of behavior control certainly
not. In fact the goal of behavior modification is to produce self-maintaining
behavior that is, behavior that continues after direct intervention is stopped. Thus,
the good behavior exhibited by the participants at the end of the corporal
punishment period could well be due to the effectiveness of previous exposure to
behavior modification and exist in spite of rather than because of exposure to
corporal punishment. If it is not possible to select a design in which each group
receives only one treatment, the researcher should try to minimize potential
multiple-treatment interference by allowing sufficient time to elapse between
treatments and by investigating distinctly different types of independent variables.
Multiple-treatment interference may also occur when participants who have
already participated in a study are selected for inclusion in another, apparently
study. If the accessible population for a study is one whose members are likely to
have participation should be collected and evaluated before subjects are selected
for the current study. If any members of the accessible population are eliminated
from consideration because of previous research activities, a note should be made
of this limitation in the research report.
Experimental Research 16

Selection-Treatment Interaction
Selection-treatment interaction, like differential selection of participants
problem associated with internal invalidity, mainly occurs when participants are
not randomly selected for treatments. Interaction effects aside, the very fact that
participants are not randomly selected from a population severely limits the
researchers ability to generalize, because what population the sample represents
is in question. Even if intact group are randomly selected, the possiblility exists
that the experimental group is in some important way different from the control
group, the larger population or both.
When the use of nonrepresentative groups results in study findings that apply only
to the groups involved and are not representative of the treatment effect in the
extended population, this is selection-treatment interaction, another threat to
population validity. This interaction occurs when actual study participants at one
level of a variable react differently to a treatment than other potential participants
in the population, at another level, would have reacted. For example, a resarcher
might conduct a study on the effectiveness of microcomputer-assisted instruction
on the math achievement of junior high students. Classes available to the
researcher may represent an overall ability level at the lower end of the ability
spectrum for all junior high students. If a positive effect is found, it may be that it
would not have been found if the subjects were truly representative of the target
population. And similary, if an effect is not found, it might have been. Thus, extra
caution must be taken in stating conclusions and generalizations based on studies
involving existing, nonrandomized groups.
Selection-treatment interaction ia also an uncontrolled variable in designs
involving randomization. For example, the accessible population is often quite
different from the researcher attempts to generalize the result of the accessible
population to the target population. Thus, the way a given population becomes
available to a researcher may make generalizability of findings questionable, no
matter how internally valid an experiment may be.suppose that, in seeking a
sample, a researcher is turned down by 9 school system and finally accepted by a
10 th. The accepting system is very likely to be different from both the other 9
Experimental Research 17

system and the population of schools to which the researcher would like to
generalize the results. Administrators and instructional personnel in the 10 th
school likely have higher morale, less fear of being inspected, and more zoal for
imporvement than personnel in the other 9 school. In the research report,
including the number of times they were turned down, so that the reader can judge
the seriousness of a possible selection treatment interaction.
Specificity of Variables
Like selection treatment interaction, specificity of variables is a threat to
generalizability of research results regardless of the particular experimental design
used. Any given study has specificity of variables that is the study is conducted
with a specific set of circumstance. We have discussed the need to describe
research procedures in sufficient detail to permit another researcher to replicate
the study. Such detailed descriptions also permit interested readers to assess how
applicable findings are to their situation. Experimental procedures require
operational definition of the variables. When studies that supposedly manipulated
the same independent variable get quite different results, it is often difficult to
determine the reasons for the differences because researchers have not provided
clear, operational description og their independent variables. When operational
description are available, they often reveal that two independent variables with the
same name were in fact defined quite differently in the separate studies, thus
explaining why results differed. Because such terms as discovery method, whole
language, and computer-based instruction mean different things to different
people, it is impossible to know what a research means by these terms inless they
are defined without operationalized descriptions, it is not clear to what
populations a study can be generalized. Generalizability of results it also tied to
the clear definition of the dependent variable although in most cases performance
on a specific measure ia the operational definition. When there are a number of
dependent variable measures to select from, questions about the comparability of
these instruments must be raised.
Experimental Research 18

Generalizability of results may also be affected by short or long term events that
occur while the study is taking place. This threat is referred to as interaction of
history and treatment effects. It describe the situation in which events extraneous
to the study alter the research results. Short term, emotion-packed events,
extraneous to the study alter the research results. Short-term, emotion-packed
events, such as the firing of a superintendent, the release of district test scores, or
the impeachment of a president might affect the behavior of participants. Usually
however, the researcher is aware of such happenings and can assess their possible
impcat of longer term events, such as wars and economic depressions, however,
is more subtle and tougher to evaluate.
Another threat to external validity is the interaction of time of
measurement and treatment effect. This threat results from the fact that posttest
given some time after treatment might provide evidence for an effect that does not
show up on posttest give some time after treatment. Conversely, a treatment may
have a long time, but a short term effect. Thus, the only way to assess the
generalizability of findings over time is to measure the dependent variable at
various times following treatment. To deal with the threats associated with
specificity, the researcher must operationally define variables in a way that has
Treatment diffusion
Treatment diffusion occurs when different treatment groups communicate with
and learn from each other. Knowledge of each others treatment often leads to the
groups borrowing aspects from each other so that the study no longer has two
distinctly different treatmens, but two overlapping ones. The integrity of each
treatments is diffused. Often, it is the more desirable treatment the experimental
treatment or the treatment with additional resources that is diffused into the less
desirable treatment.
Experimental effects
Researcher themselves also present potential threats to the external validity of
their own studies. A researchres influences on participants or on study procedures
Experimental Research 19

are knwon as experimenter effects. Passive experimenter effects occur as a result
of characteristic or personality traits of the experimenter, such as gender, age,
race, anxiety level, and hostility level. These influences are collectively called
experimenter personal attributes effects. Active experimenter effects occur when
the researchers expectations of the study results affect her behavior and actually
contribute to production certain research outcomes. This effect is referred to as the
experimenter bias effect. Thus, an experimenter may unintentionally affect study
results, typically in the direction desired by the researcher, simply by looking,
feeling, or acting a certain way. One from of experimenter bias occurs when the
researcher affects participants behavior, or is inaccurate in evaluating their
behavior because of previous knowledge of the participants. Suppose researcher
hypothesizes that a new reading approach will improve reading skills. If the
researcher know that suzys reading skills a higher rating than they actually
warrant. This example, illustrates another way a researchers expectations may
actually contribute to producing those outcomes knowing which participants are
in the experimental and control groups mau cause the researcher to
unintentionally evaluate their performances differently.
It is difficult to identify experimenter bias in a study, which is all the more
reason for researchers to be aware of its consequences on the external validity of
their study. The moral is that the researcher should strive to avoid communicating
emotions and expectations to participants in the study. Experimenter bias effects
can be reduced by blind scoring in which the researcher doesnt know whose
performance is being evaluated.
Reactive arrangements
Reactive arrangements, also called participants effects, are threats to validity
that are associated with the wey in which a study is conducted and the feelings
and attitudes of the participants involved. As discussed previously, in order to
maintain a high degree of control and thus obtain internal validity, a researcher
may create an experimental environment that is highly artificial and hinders
generalizability to nonexeprimental settings, this is a reactive arrangement.
Experimental Research 20

Another type of reactive arrangement results from participants knowledge that
they are involved in an experiment or their feeling that they are in some way
receiving special attention. The effect that such knowledge or feellings can
have on the participants was demonstrated at the hawthorne plant of the western
electric company in chicago some years ago, studies were conducted to
investigate the relantionship between various working conductions and
productivity. As part of their study, researchers investigated the effect og light
intensity and worker output. The researchers increasted light intensity and
production went up. They increased it some more and production went up some
more. The brighter the place became, the more production rose. As a check, the
researcher decreased the light intensity, and guess what, production went up the
darker it got, the more workers producted. The researchers soon realized that it
was the attention given the workers, and not the illumination, that was affecting
production. To this day, the term hawthorne effects is used to describe any
situation in which participants behavior is affected not by treatment per se, but
their knowledge of participating in a study.
A related reactive effect, known as compensatory rivalry or the jhon henry
effect, occurs when members of a control group feel threatened or challenged by
being in competition with an experimental group and they perform way beyond
what would normally be expected. Folkhero john henry, you may recall, was a
steel driven man who worked for a railroad. When he heard that a sterm drill
was going to replace him and his fellow steel drivers, he challenged and set out to
beat, the machine. Through tremendous effort he managed to win the ensuing
contest, dropping dead at the finish line. When research participants are told that
they will form the control group for a new. Experimental methode, they act like
jhon henry they decide to challenge the new method by putting extra effort into
their work, essentially saying, well show them that our old ways are as effective
as their newfangled ways. By doing this, however, the control group performs
atypically and their performance provides a rival explanation for the experimental
group is not much better than that of the control group.
Experimental Research 21

As an antidote to the hawthorne and jhon henry effects, educational researchers
often attempt to achieve a placebo effect . The term comes from medical
researchers who discovered that any medication , even sugar and water, could
make subjects feel better , any beneficial effect caused by a persons expectation
about a treatment rather than the treatment itself became known as the placebo
effect. To counteract this effect, a placebbo approach was developed in which half
of the subjects receive the ture medication and half receive a placebo. The use of a
placebo is, of course, not known by the participants, both groups think they are
taking a real medicine. The application of the placebo effect in educational
research is that all groups in an experiment should appear to be treated the same.
Suppose, for example, you have four groups of ninth graders, two experimental
and two control, and the treatment is a film designed to promote a positive attitude
toward a vocational carrer, and the experimental participants are to be excused
and shown another film whose content is unrelated to the purpose of the
an added control, you might have all the participants told that there are two
movies and that eventually everyone will see both movies. In other words, it
should appear as if all the students are doing the same thing.
Another reactive arrangemen, or participants effect, is the novely effect, which
refers to the increased interest, motivation, or engagement participants develop
simply because they are doing something different. In other words, a treatment
may be effective because it is different, not because it is better. To counteract the
novely effect, a researcher should conduct a study over a period of time long
enough to allow the treatment newness to wear off. This is especially
advisable if the treatment involves activities very different from the subjects usual
routine. Obviously there are many internal and external threats to the validity of
an experimental ( or causal comparative ) study. You should be awaer of likely
threats and strive to nullify them. One main way to overcome many threats to
validity is to choose a research design that controls for such threats. We examine
some of these design in the following sections.
Group Experimental designs
Experimental Research 22

The validity of an experiment is a direct function of the degree to which
internal and external variables are controlled. If such variables are not controlled,
it is difficult to interpret the results of a study and the groups to which results can
be generalized. The trem confounding is sometimes used to describe an
intertwining of the effects of the independent variable with those of extraneous
variables that makes it difficult to determine the unique effects of each this is what
experimental design is all about the control of extrancous variables. Good
desaigns control many sources of invalidity poor designs control few. If you
recall, two types of extraneous variables in need of control are participants
variables and environmental variables. Participant variables include both
organismic variables and intervening variables. Organismic variables are
participant ia an example. Intervening variables intervence between the
independent and the dependent variable and cannot be directly observed but can
be controlled for anxiety and boredom are example.
Control of Extraneous Variables
Randomization is the best single way to simultaneously control for many
extraneous veriables. Thus, randomization should be used whenever possible
participants should be randomly selected from a papulation and randomly
assigned to treatment groups. To ensure random selection and assignment,
researcher use methods that rely on pure chane, usually consulting a table of
random numbers. Other randomization methods are also available. For example a
researcher could flip a coin or use odd and even numbers on a die to assign
participants to two treatments, heads or an even number would signal assignment
to treatment 1, and tailsor an odd number would signal assignment to treatment 2.
Randomization is effective in creating equivalent, representative groups that are
essentially the same on all revelant variables. As noted, the use of randomly
formed not possible with causal comparative research. The underlying rationale
for randomizition is that if subject are assigned at random to groups, there is no
reason to belive that the groups will be greatly different in any systematic way.
Thus, the groups would be expected to perform essentially the same on the
Experimental Research 23

dependent variable if the independent variable makes no differnces. Therefore, if
the groups perform differently ar the end of the study , the differnce can be
atributed to the independent variable. It is important to remember that the large
the groups, the more confidence the researcher can have in the effectiveness of
randomization. Randomly assigning 6 participants to two treatments is much less
likely to equalize extraneous variables than assigning 50 participants to two
treatments. In addition to equating groups on participants variables such as ability,
gender, or prior experience, randomization can also equalize groups on
environmental variables. Teacher, for example, can be randomly assigned to
teratment groups so that the experimental groups will not have all the carmel
kandee teachers or all the hester hartless teacher. clearly, the researcher
shoulduse as much randimization as possible. If subjects cannot be randomly
assigned to groups, then at least treatment conditions should be randomly assigned
to the existing groups.
In addition to randomization, there are other ways to control for extraneous
variables. Certain environmental variables. For example can be controlled by
holding them constant foe all groups. Recall the example of the students tutor
versus parent tutor the example, help time was an important variable that
had to be held constant, that is, made the same for both groups for them to be
fairly compared. Other such variables that might need to be held constant include
learning materials, prior exposure, meeting place and time ( students might be
more aleart in the morning than in the afternoon ), and years of research
experience. Controlling participants variables is critical. If the groups are not the
same to start with, you have not even given yourself a figthing chance to obtain
valid, intrepretable research results. Even if groups cannot be randomly formed,
there are a number of techniques that can be used to try to equate groups. These
include matching, comparing homogeneous groups or subgroups, using
participants as their own controls, and analysis of covariance, several of these
concepts were introduced the discussion of equating groups in causal
comparative studies.

Experimental Research 24

Matching is a technigue for equating groups on one or more variables, usually
ones highly related to performance on the dependent variable. The most commoly
used approach to matching involves random assignment of pairs, one participants
to each group. In other words, the researcher attempts to find pairs of participants
similar on the variable or variables to be controlled. If the researcher is
matchingon gendre, obviously the matched pairs must be of the same gendre.
However, if the researcher is matching on variables such as pretest, GRE, or
ability scores, the pairing can be based on similarity of scores. Unless the number
of participants is very large, it is unreasonable to try to make exact matches or
matches on more than one or two variables. Once a matched pair is identified,
one member of the pair is randomly assigned to one treatment group and the other
member to the other treatment group. A participants who does not have a suitable
match is excluded from the study.the resulting matched groups are idential or very
similiar with respect to the variable being controlled. Major problem with such
matching is that there are invariably participants who do not have a match and
must be eliminated from the study. This may cost the research many subject,
especially if matching is attempted on two or more variables . of course, one way
to combat loss of participants is to match less stringenrly. They will constitute an
acceptable match. This approach may increase the number of subjects, but it tends
to defeat the purpose of matching.
A related matching procedure is to rank all of the participants, from highest to
lowest, based on their score. Are the first pair. One member of the first pair is
randomly assigned to one group and the other member to the other group. The
next two highest ranked participants are the second pair, and so on. The major
advantage of this approach is that no participants are lost. the major disadvantage
is that it is a lot less precies than pair-wise matching. Advance statistical
procedures, such as analysis of convariance, have greatly reduced the research use
of matching.

Experimental Research 25

Comparing Homogeneous Groups or Subgroups
Another previosly discussed way to control an extraneous variable is to compare
groups that are homogeneous with respect to that variable. For example if IQ
were an identified extraneous variable, the researcher might select only
participants with IQ between 85 and 115 ( adverage IQ ). The resercher would
then randomly assign half the selected participants to the experimental group and
half to the control group. Of course, this proceduse also lowers the number of
participants in the population and additionally restricts the generalizability of the
findings to participants with IQs between 85 and 115 as noted in the discussion of
causal comparative research , a similar, more satisfactory approach is to form
different subgroups representing all levels of the control variable. For example,
the available participants might be divided into high ( 116 or above ), average ( 85
to 115 ), and low ( 84 and below ) IQ subgroups. Half of the participants from
each of the subgroups could then be randomly assigned to the experimental group
and half to the control group. This procedure should sound familiar, since it
describe startified sampling. If the researcher is interested not just in controlling
the variable but also in seeing if the independent variable affects the dependent
variable differently at different levels of IQ, the best appoarch is to build the
control variable right into the design. Thus, the research design would have six
cells, two treatment by three IQ levels. Draw the design for yourself, and label
each cell with its treatment and IQ level.
Using participants as their own controls
Using participants as their own controls involves exposing a single groups to
different treatments one treatment at a time. This strategy helps to control for
participants difference, because the same participants get both treatments. Of
course, this approach is not always feasible, you cannot teach the same algebraic
concepts to the same groups twice using two different method of instruction (
well, you, could, but it would not make much sense ). A problem with this
approach is a carryover effect from one treatment to the next. To use a previous
example, it would be very difficult to evaluate the effectiveness of corporal
Experimental Research 26

punishment in improving behavior if the group receive corporal punishment was
the same group that had previously been exposed to behavior modification. If one
group was available. A better approach, if feasible, would be to randoml divide
the group into two smaller groups, each of which would receive both treatments
but in a different order. Thus, the researcher could at least get some idea of the
effectiveness of corporal punishment because there would be a group that received
it before behavior modification. In situations in whichthe effect of the dependent
variable dissapers quickly after treatment, or in which a single participant is the
focus of the research, participant can be used as their own controls.
Analysis of Covariance
The analysis of covariance is a statistical method for equating randomly
formed groups of on one or more variables. Analysis of covariance adjusts scores
on a dependent variable for initial differences on some other variable, such as
pretest scores, IQ, reading readiness, or musicsal aptitude. The covariate variable
should be one related to performance on the dependent variable. Although
analysis of covariance can be used in studies when groups cannot be randomly
formed, its use is most appropriate when randomization is used. In spite of
randomization, it might be found that two groups still differ significantly in terms
of pretest scores. Analysis of covariance can be used in such cases to correct or
adjust postest scores for initial pretest differences. However, analysis of
covariance is not universally useful. For example, the relationship betweenthe
independent and covariate variables must be linear (represented by a straight line).
If the relationship is curviliniear, analysis of covariance is not useful. Also,
analysis of covariance is often used when a study deals with intact groups,
uncontrolled variables, and nonrandom assignment to treatments, all of which
weaken its results. Calculation of an analysis of covariance is a complex
The experimental design you select to a great extent dictates the specific
procedures of your study. Selection of a given design influences factors such as
Experimental Research 27

whether there will be a control group, whether participants will be randomly
selected and assigned to groups, whether the group will be pretested, and how
data will be analyzed. In selecting a design, determining which designs are
appropriate for study and for testing hypothesis is important before determining
which of these are also feasible given the constraints under which you may be
There are two major classes of experimental designs: single-variable
designs and factorial designs. A single-variable design is any design that
involves one manipulated independent variable. Meanwhile, factorial design is
any design that involves two or more independent variables with at least one
being manipulated.
Single-variable designs are classified as pre-experimental, true
experimental, or quasi-experimental, depending on the degree of control they
provide for threats to internal and external invalidity. Pre-experimental design
do not do a very good job of controlling threats to validity and should be
avoided. In fact, the results of a study based on pre-experimental designs are so
questionable they are not useful for most purposes except, perhaps, to provide a
preliminary investigation of a problem. True-experimental designs provide a
very high degree of control and are always to be preferred. Quasi-experimental
designs do not control as well as true experimental designs but do a much
better job than the pre-experimental designs. If we could give such grades
towards these three experimental designs, true experimental designs would get
an A, quasi-experimental designs would get B or C (some are better than
others), and pre-experimental designs would get a D or E. You could see on the
pictures below:

Experimental Research 28

Figure 2.1
True-experimental designs are better than
quasi-experimental designs;
Quasi-experimental designs are better than
pre-experimental designs;
Pre-experimental designs are quite better than not
doing study at all.

Factorial designs are basically elaborations of single-variable experimental
designs except that they permit investigation of two or more variables,
individually and in interaction with each other. After an independent variable has
been investigated using a single-variable design, it is often useful to then study the
variable in combination with one or more other variables. Some variables work
differently when paired with different levels of another variable. The designs
discussed next represent the basic designs in each category.

Experimental Research 29

Pre-Experimental Designs
a) The One-Shot Case Study
The one-shot case study involves a single-group treatment (X) and then
posttested (O). None of the sources of invalidity are controlled in this design.
b) The One-Group Pretest-Posttest Design
The one-group pretest-posttest design involves a single group that is pretested
(O), exposed a treatment (X), and postested (O). The success of the treatment is
determined by comparing pretest and posttest scores. This design controls
some areas of invalidity not controlled by the one-shot case study, but a
number of additional factors relevant to this design are not controlled. If
participants do significantly better on the posttest than on the pretest, it cannot
be assumed that the improvement is due to the treatment. History and
maturation are not controlled. Something may happen to the participants that
makes them perform better the second time, and the longer study takes, the
more likely it is that is something will threaten validity. Testing and
instrumentation also are not controlled; the participants may learn something
on the pretest that helps them on the posttest, or unreliability of the measures
may be responsible for the apparent improvement. Statistical regression is also
not controlled. Even if subjects are not selected on the basis of extreme scores
(high or low), it is possible that a group may do very poorly on the pretest, just
by poor luck. For example, participants may guess badly just by chance on a
multiple-choice pretest and improve on a posttest simply because, this time,
their guessing produces a score that is more in line with an expected score. The
external validity threat pretest-treatment interaction is also not controlled.
Pretest-treatment interaction may cause participants to react differently to the
treatment than they would have if they had not been pretested.
c) The Static-Group Comparison
The static-group comparison involves at least two nonrandomly formed
groups: one that receives a new or unusual treatment (the experimental
treatment) and another that receives a traditional treatment (the control
Experimental Research 30

treatment). Both group are postested. However, it is probably more appropriate
to call them both comparison groups, because each really serves as the
comparison for the other. Each group receives some form of the independent
variable (the treatments). Therefore, as an example, if the independent variable
is type of drill and practice, the experimental group (X
) may receive
computer-assisted drill and practice, and the control group may receive
worksheet drill and practice. Occasionally, but not often, the experimental
group may receive something while the control group receives nothing. The
purpose of a control group is to indicate what the performance of the
experimental group would have been if it had not received the experimental
treatment. This purpose is fulfilled only to the degree that the control group is
equivalent to the experimental group.
The static-group comparison design can be expanded to deal with any
number of groups. For three groups, the design would take the following form:
X1 O
X2 O
X3 O
Which group is the control group? Basically, each group serves as a
control or comparison group for the other two. As already emphasized, but worthy
of repeating, the degree to which the groups are equivalent is the degree to which
their comparison is reasonable. In this design, participants are not randomly
assigned to groups and there are no pretest data; thus, it is difficult to determine
just how equivalent the groups are. That is, it is possible taht posttest differences
are due to initial group differences in maturation, selection, and selection
interactions, rather than the treatment effects. Mortality is also a problem; if you
lose participants from the study, you have no information about what you have
lost because you have no pretest data. On the positive side, the presence of a
comparison group does control for history, since it is assumed that events
occuring outside the experimental setting will equally affect both groups.
Experimental Research 31

Nevertheless, the static-group comparison design is occasionally employed in a
preliminary or exploratory study. For example, one semester, early in the term, a
teacher wondered if the kind of test items given to educational research students
affects their retention of course concepts. For the rest of the term, students in one
section of the course were given multiple-choice tests, and students in another
section were given short-answer tests. At the end of the term, group performances
were compared. The group receiving short-answer test items had higher total
scores than the multiple-choice item group. On the basis of this exploratory study,
a formal investigation of this issue was undertaken.

True-Experimental Designs
a) The Pretest-Posttest Control Group Design
The pretest-posttest control group design requires at least two groups, each of
which is formed by random assignment; both group are administered a pretest,
each group receives a different treatment, and both groups are posttested at the
end of the study. Posttest scores are compared to determine the effectiveness of
the treatment. The pretest-posttest control group design may also be expanded
Experimental Research 32

to include any number of treatment groups. For three groups, for example, this
design would take the following form:
There are a number of ways in which the data from this and other
experimental designs can be analyzed to test the research hypothesis regarding
the effectiveness of the treatments. The best way to analyze these data is to
compare the posttest scores of the two treatment groups. The pretest is used to
see if the groups are essentially the same on the dependent variable at the start
of the study. If they are; posttest scores can be directly compared using a
statistic called the t test. If the groups are not essentially the same on the pretest
(random assignment does not guarantee equality), posttest scores can be
analyzed using analysis of covariance. Recall that covariance adjusts posttest
scores for initial differences on any variable, including pretest scores. This
approach is superior to using gain or difference scores (posttest minus pretest)
to determine the treatment effects.
A variation of the pretest-posttest control group design involves random
assignment of members of matched pairs to the treatment groups, in order to
more closely control for one or more extraneous variables. There is really no
advantage to this technique, however, because any variable that can be
controlled through matching can be better controlled using other procedures
such as analysis of convariance.
b) The Posttest-Only Control Group Design
The posttest-only control group design is exactly the same as the pretest-
posttest control group design except there is no pretest; participants are
randomly assigned to at least two groups, exposed to the different
treatments, and posttested. Posttest scores are compared to determine the
effectiveness of the treatment. As with the pretest-posttest control group
Experimental Research 33

design, the pretest-posttest control group design, the posttest-only control
group design can be expanded to include more than two groups.
The combination of random assignment and the presence of a
control group serves to control for all sources of internal invalidity except
mortality. Mortality is not controlled because of the absence of pretest data
on participants. However, it may or may not be a problem, depending on
the duration of the study. If it is not a problem, the researcher may report
that although mortality is a potential threat to the validity with this design,
it did not prove to be a threat because the group sizes remained constant or
nearly constant throughout the study. If the probability of differential
mortality is low, the posttest-only design can be very effective. However,
if there is any chance that the groups may be different with the respect to
pretreatment knowledge related to the dependent variable, the pretest-
posttest control group design should be used. Which design is best
depends on the study. If the study is to be short, and if it can be assumed
that neither group has any knowledge related to the dependent variable,
then the posttest-only design may be the best choice. If the study is to be
lengthy, or if there is a chance that the two groups differ on initial
knowledge related to the dependent variable, the pretest-posttest control
group design may be the best.
c) The Solomon Four-Group Design
The Solomon four-group design involves random assignment of
participants to one of four groups; two of the groups are pretested and two
are not; one of the pretested groups and one of the unpretested groups
receive the experimental treatment; and all four groups are posttested with
the dependent variable.
The correct way to analyze data resulting from application of the Solomon
four-group design is to use a 2 x 2 (two by two) factorial with treatment
and control groups crossed with pretesting and nonpretesting. There are
two independent variables in this design: treatment/control and pretest/no
Experimental Research 34

pretest. The 2 x 2 factorial analysis tells the researcher whether the
treatment is effective and whether there is an interaction between the
treatment and the pretest.

Quasi-Experimental Designs
a) The Nonequivalent Control Group Design
The nonequivalent control group design looks very much like the pretest-
posttest control group design, except that the nonequivalent control group
design, except that the nonequivalent control group design does not
involve random assignment. The lack of random assignment raises the
possibility of interactions between selection and variables such as
maturation, history, and testing. Reactive effects are minimized.
b) Time-Series Design
In the time-series design, one group is repeatedly pretested, exposed to a
treatment, and then repeatedly posttested. If s group scores essentially the
same on a number of pretests and then significantly improves following a
treatment, the researcher has more confidence in the effectiveness of the
treatment than if just one pretest and one posttest were administered.
History is a problem, as is pretest-treatment interaction.

Experimental Research 35

c) Counterbalanced Design
In a counterbalanced design, all groups receive all treatments but in a
different order, the number of groups equals the number of treatments, and
groups are postested after each treatment. This design is usually employed
when intact groups must be used and when administration of a pretest is
not possible. A weakness of this design is potential multiple-treatment

Factorial Designs
Factorial designs involve two or more independent variables, at least one
of which is manipulated by the researcher. Factorial designs are basically
elaborations of single-variable true experimental designs that permit
investigation of two or more variables individually and in interaction with each
other. The term factorial refers to a design that has more than one independent
variable, or factor. For example, method of instruction is one factor and student
aptitude is another. The factor student aptitude also has two levels, high
aptitude and low aptitude.
Experimental Research 36

The purpose of a factorial design is to determine whether the effects of an
independent variable are generalizable across all levels or whether the effects
are specific to particular levels. A factorial design also can demonstrate
relationships that a single-variable design cannot. For example, a variable
found not to be effective in a single-variable study may be found to interact
significantly with another variable.

Experimental Research 37

Experimental research is the only type of reseacrh that can test hypotheses
to estabilsh cause-effect relationships. When well conducted, experimental studies
produce the soundest evidence concering cause-effect relationships. The results of
experimental research permit prediction, but not the kind that is characteristic of
correlational research.
There are two major classes of experimental designs: single-variable
designs and factorial designs. Single-variable designs involve: pre-experimental,
true-experimental, quasi-experimental, and factorial designs. Pre-experimental
designs can be applied through the one-shot case study, the one-group pretest-
posttest design, and the static-group comparison. True-experimental designs can
be implemented in several ways: the pretest-posttest control group design, the
posttest-only control group design, and the Solomon four-group design. Quasi-
Experimental Designs are applied in the nonequivalent Control Group Design,
Time-Series Design, and Counterbalanced Design. The last, factorial designs are
basically elaborations of single-variable true experimental designs that permit
investigation of two or more variables individually and in interaction with each
An experimental research study design has several advantages and
disadvantages. The advantages of experimental research are two: the ability to
prove causal relationship and the ability to precisely manipulate one or more
variables researcher desired. Meanwhile the disadvantages are:
It is difficult to generalize in everyday life. This means that the results of an
experimental study cannot be directly used in real life or day-to- day. It is caused
by conditions that are very controlled experimental studies, so the situation is not
as in the real life. It requires considerable time. But this reason is not entirely
correct, because sometimes an experimental study conducted in a relatively short
time compared to the non-experimental research.
Experimental Research 38