the experimental research approach - higher · pdf filetypes ways to field laboratory ......

20
64 The Experimental Research Approach 3 Experimental Research Causation Definition Advantages and Settings Disadvantages Types Ways to Field Laboratory Identify CH03.QXD 5/22/2003 9:41 AM Page 64

Upload: hadung

Post on 11-Mar-2018

213 views

Category:

Documents


0 download

TRANSCRIPT

64

The Experimental Research Approach

3Experimental Research

Causation Definition Advantages and SettingsDisadvantages

Types Ways to Field LaboratoryIdentify

CH03.QXD 5/22/2003 9:41 AM Page 64

In June 1999, the American Psychological Association published a special issue of TheMonitor on psychology and cancer.The thrust of this special issue was the contributionmade by psychology in cancer research and cancer caregiving. At first glance, it mayseem strange to focus on the role of psychology in cancer because cancer is typicallythought of as a physical disorder with little or no psychological component. However,we are increasingly finding that the mind and the body are not separate entities but areinterconnected. One has only to be associated with a cancer patient to realize the psy-chological impact of this illness.

Consider, for example, the case of Mrs. A, a forty-nine-year-old businesswoman.Mrs. A’s mother was diagnosed with cancer when she was twelve. For the next five years, Mrs. A’s mother underwent a series of treatments consisting of a radicalmastectomy, radiation that burned and scarred her skin, and chemotherapy so toxicthat it left her crouched on her hands and knees, vomiting onto newspapers by herbed. Mrs A’s mother was totally devastated by her illness and forced her daughter to be her only caregiver while at the same time forcing her husband out of the bed-room. By the time her mother died, Mrs. A had become her virtual slave.Today, Mrs. Ais not only filled with rage at the disease that robbed her of her childhood, but she isalso a patient at a high-risk clinic designed to educate women about breast cancerprevention and help them manage their emotions (Clay, 1999).This is just one exam-ple of the emotional turmoil that may accompany cancer.

So what kind of contribution can psychology make to cancer research and treat-ment? Remember that cancer can cause emotional turmoil not only in the cancervictim but also in the caregivers of the victim and other family members. Is it possi-ble that a psychological intervention, or psychotherapy, can assist in preventing can-cer or even prolong life in those that have cancer? We know that prolonged stressaffects just about every system in the body.This leads to the question of whetherindividuals living in very stressful conditions are more susceptible to cancer. Ifresearch demonstrates that they are, can a stress management program reduce therisk of developing cancer? Many other legitimate questions could also be asked.These are questions that need to be asked and that are currently being researched.

How do we seek the answers to these questions? The primary way in whichresearchers have begun to provide answers to these questions is by conducting apsychological experiment or using the experimental research approach, theapproach discussed in this chapter.

65

CH03.QXD 5/22/2003 9:41 AM Page 65

Chapter PreviewIn the present chapter I discuss the experimental research approach. This is theresearch approach that is typically used to investigate causality. Although otherapproaches are used, the experimental approach is the primary one because of itscharacteristics. Because causality is the type of relationship investigated whenusing the experimental research approach, one might assume that we have agood understanding of the meaning of causality and how to identify causal rela-tionships. This is an inaccurate assumption. To drive this point home, I first dis-cuss the meaning of causation and, hopefully, convince you that it is difficult toidentify causality. Then I discuss the characteristics of the experimental researchapproach, its advantages and disadvantages, and the settings in which psycholog-ical experiments are conducted.

IntroductionThe experimental research approach is a quantitative approach designed todiscover the effects of presumed causes. The key feature of this approach is thatone thing is deliberately varied to see what happens to something else, or to dis-cover the effects of presumed causes. This is something that people do all the time.For example, individuals try different diets or exercise to see if they will loseweight. Others may get an education to see if that will lead to a better job. Stillother individuals may change their diet to see if it will lower their blood cholesterollevel. As you can see, both scientists and nonscientists use experimentation to tryto identify causal relationships. However, scientific experimentation differs fromnonscientific experimentation in that the scientist makes a deliberate attempt tomake observations that are free of bias. Both approaches attempt to identify causalrelationships. I therefore begin by exploring the concept of causation and then dis-cuss scientific experimentation and the nature of causation that experiments test.

CausationCausation is one of those terms that people frequently use but often don’t reallyunderstand. People ask questions like “What causes cancer?” “What causes a per-son to murder someone else?” “What causes a man to beat his wife?” What dothey really mean? Common sense suggests that causality refers to a condition inwhich one event—the cause—generates another event—the effect. However,causality is much more complex.

When individuals discuss the effects of events, they tend to use the words causeand effect rather informally. People are likely to assume that manipulation isimplicit in the concept of causation. If we manipulate or do something, we expectsomething else to happen. If something does happen, the thing or event wemanipulate is called the cause and what happens is called the effect. For example, ifwe spank a child for coloring on a wall and then observe that he no longer colorson the wall, we assume that the spanking caused the child to stop the coloring.This temporal relationship between events such as spanking and ceasing a behav-ior such as coloring on a wall gives people an intuitive sense that they understand

66 | The Experimental Research Approach

CausationA term whose meaning isdebated by philosophers,but that in everydaylanguage seems to implythat manipulation of oneevent produces anotherevent

Experimental researchapproachThe research approach inwhich one attempts todiscover the effects ofpresumed causes

CH03.QXD 5/22/2003 9:41 AM Page 66

the meaning of cause and effect. However, the definition of these terms has eludedphilosophers for centuries.

CauseThe intuitive definition most of us have is that a cause is something that makessomething else exist. However, this intuitive definition of cause is far too simplisticbecause all causal relationships are contextually dependent. For example, depres-sion can occur in many different ways. Eating a diet that does not contain the pre-cursor of the central neurotransmitter serotonin, having a baby, being fired from ajob, getting a divorce, and numerous other events can cause an onset of depres-sion. However, none of these events by itself is sufficient to cause depression.Many women have babies without experiencing postpartum depression. For someindividuals losing a job causes depression, whereas others view it as an opportu-nity to develop another stimulating career. The point is that many factors are usu-ally required for an effect to occur, and we rarely know all of them and how theyrelate to each other. This means that any causal relationship occurs within the con-text of many factors and, if any of these other factors change, the causal relation-ship previously identified may or may not be replicated. This is why causalrelationships are not deterministic but probabilistic (Shadish, Cook, & Campbell,2002). In spite of the difficulty with identifying the cause of some event, it is stilluseful to think of cause as something that produces something else. However, wemust never lose sight of the fact that any causal relationship occurs contextuallyand that the generalization of any effect is always an issue.

EffectAn effect is the difference between what would have happened and what didhappen. In an experiment, the effect is the difference between what did happenwhen a treatment was administered and what would have happened to this samegroup of individuals if the treatment had not been administered. Note the emphasison the same group of individuals. However, it is impossible for the same group ofpeople to both have and not have a treatment, so identifying a true effect is notpossible. What we attempt to do within the context of an experiment is to obtainan imperfect measure of this difference by doing such things as having two differ-ent groups of individuals and administering the treatment to one group and not tothe other group. The point is that it is never possible to obtain a true measure of aneffect because this requires participants to both be exposed to something and notbe exposed to something and that is not possible.

Causal RelationshipWhat kind of evidence is needed to know if a cause and effect are related? JohnStuart Mill (1874), a British philosopher, set forth canons that could be used toexperimentally identify causality. These canons form the basis of many of theapproaches currently used.

Causation | 67

CauseA contextually dependentevent that makessomething else exist

EffectThe difference betweenwhat would havehappened and what didhappen when a treatmentis administered

CH03.QXD 5/22/2003 9:41 AM Page 67

1. The first canon is the method of agreement, by which one identifies causalityby observing the element common to several instances of an event. This canon canbe illustrated by the frequently cited case of the man who wanted to find out sci-entifically why he got drunk. He drank rye and water on the first night and becamedrunk. On the second night, he drank scotch and water and became drunk again.On the third night, he got drunk on bourbon and water. He therefore decided thatthe water was the cause of his getting drunk because it was the common elementeach time. This method, as you can see, is inadequate for unequivocally identifyingcausation because many significant variables—such as the alcohol in the rye,scotch, and bourbon—may be overlooked.

2. The second canon is the method of difference, by which one attempts to iden-tify causality by observing the different effects produced in two situations that arealike in all respects except one. The method of difference is the approach taken inmany psychological experiments. In an experiment designed to test the effect of adrug on reaction time, the drug is given to one group of participants while a placebois given to another group of matched participants. If the reaction time of the drug-taking group differs significantly from that of the control group, the difference isusually attributed to the drug (the causal agent). This method provides the basis fora great deal of work in psychology aimed at identifying causality.

3. The third canon set forth by Mill is the joint methods of agreement and dif-ference. This method is exactly what its name implies. The method of agreementis first used to observe common elements, which are then formulated as hypothe-ses to be tested by the method of difference. In the case of the man who wanted tofind out why he got drunk, the common element, water, should have been formu-lated as a hypothesis to be tested by the method of difference. Using the method ofdifference, researchers would give one group of participants water and a matchedgroup another liquid (such as straight bourbon). Naturally, the group drinking onlywater would not get drunk, indicating that the wrong variable had been identifiedeven though it was a common element.

4. The fourth canon is the method of concomitant variation. This method statesthat a variable is either a cause or an effect, or else is connected through some fac-tor of causation if variation in the variable results in a parallel variation in anothervariable. Plutchik (1974) interprets this canon to be an extension of the method ofdifference in that, rather than just using two equated groups in an experiment, theresearchers use three or more, with each group receiving a different amount of thevariable under study. In the previously cited drug example, rather than just aplacebo group and a drug group, one placebo and several drug groups could beused, with each drug group receiving a different amount of the drug. Reactiontimes could then be observed to determine if variation in the quantity of the drugresults in a parallel variation in reaction time. If this parallel variation is found,then the drug is interpreted as being the cause of the variation in reaction time.

Some writers interpret this canon as including correlation studies. One is onextremely shaky ground in attempting to infer causative relationships from corre-lational studies because many correlational studies simply describe the degree ofrelationship. However, recent work is making strides in enabling causation to beinferred from correlational studies.

68 | The Experimental Research Approach

Method of agreementThe identification of thecommon element inseveral instances of anevent

Method of differenceThe identification of thedifferent effects producedby variation in only oneevent

Joint methods ofagreement anddifferenceThe combination of themethods of agreement anddifference to identifycausation

Method of concomitantvariationThe identification ofparallel changes in twovariables

CH03.QXD 5/22/2003 9:41 AM Page 68

From looking at the works of such people as Mill, one gets the idea that wehave a fairly adequate grasp of what causation is and how to obtain evidence of it.This belief is reinforced when we see that many studies of causation are based onMill’s canons. Such philosophizing and experimentation have not completely clar-ified the meaning of the word cause, however.

Exhibit 3.1, which presents Morison’s (1960, pp. 193–194) review of the his-tory of attempts to find the cause of malaria, illustrates the ambiguity of this word.Essentially, the method of agreement was used first to hypothesize that the bad air

Causation | 69

Whatever the reason, medical men have found itcongenial to assume that they could find some-thing called The Cause of a particular disease. If onelooks at the history of any particular disease, onefinds that the notion of its cause has varied with thestate of the art. In general, the procedure has beento select as The Cause that element in the situationwhich one could do the most about. In many casesit turned out that, if one could take away this ele-ment or reduce its influence, the disease simply dis-appeared or was reduced in severity.This wascertainly desirable, and it seemed sensible enoughto say that one had got at the cause of the condi-tion.Thus in ancient and medieval times malaria, asits name implied, was thought to be due to the badair of the lowlands. As a result, towns were built onthe tops of hills, as one notices in much of Italytoday.The disease did not disappear, but its inci-dence and severity were reduced to a level consis-tent with productive community life.

At this stage it seemed reasonable enough toregard bad air as the cause of malaria, but soon theintroduction of quinine to Europe from SouthAmerica suggested another approach. Apparentlyquinine acted on some situation within the patientto relieve and often to cure him completely.Toward the end of the last century the malaria par-asite was discovered in the blood of patients suf-fering from the disease.The effectiveness ofquinine was explained by its ability to eliminatethis parasite from the blood.The parasite nowbecame The Cause, and those who could afford the

cost of quinine and were reasonably regular intheir habits were enabled to escape the most seri-ous ravages of the disease. It did not disappear as apublic health problem, however, and further studywas given to the chain of causality.These studieswere shortly rewarded by the discovery that theparasite was transmitted by certain species ofmosquitoes. For practical purposes The Cause ofepidemic malaria became the Mosquito, and atten-tion was directed to control of its activities.

Entertainingly enough, however, malaria hasdisappeared from large parts of the world withoutanyone doing much about it at all.The fens ofBoston and other northern cities still produce mos-quitoes capable of transmitting the parasite, andpeople carrying the organism still come to theseareas from time to time, but it has been manydecades since the last case of the disease occurredlocally. Observations such as this point to the prob-ability that epidemic malaria is the result of anicely balanced set of social and economic, as wellas biological factors, each one of which has to bepresent at the appropriate level.We are still com-pletely unable to describe these sufficient condi-tions with any degree of accuracy, but we knowwhat to do in an epidemic area because we havefocused attention on three or four of the most nec-essary ones.

From “ ‘Gradualness, Gradualness, Gradualness’ (I. P. Pavlov)”by R. S. Morison, 1960, American Psychologist, 15,pp. 187–198. Copyright 1960 by the American PsychologicalAssociation. Reprinted by permission of the author.

E X H I B I T 3 . 1

Morison’s Discussion of the History of Attempts to Find the Cause of Malaria

CH03.QXD 5/22/2003 9:41 AM Page 69

in the lowlands caused malaria because a common element among people livingon top of the hills was better air. Subsequent investigation using the method of dif-ference revealed that only individuals with the malaria parasite in their blood suf-fered from the disease. The problem with this second explanation is that it did notexplain how the parasite came to exist in the bloodstream, until it was found thatthe mosquito transmitted it. As you can see, the various canons set forth by Millenable us to identify the relationships that exist among a set of variables. However,they do not help us to name the single factor that causes an effect, just as they didnot enable scientists to identify the single factor that causes malaria. This is becausethe identification of causation can occur only when no alternative interpretationsfor an effect exist other than the one specified. When we have reached this stage,we have essentially identified both the necessary and the sufficient conditions forthe occurrence of an event. A necessary condition refers to a condition that mustbe present in order for the effect to occur. (To become an alcoholic, you must con-sume alcohol.) A sufficient condition refers to a condition that will always pro-duce the effect. (Destroying the auditory nerve always results in a loss of hearing.)

A condition must be both necessary and sufficient to qualify as a cause becausein such a situation the effect would never occur unless the condition were presentand, whenever the condition was present, the effect would occur. If a conditionwere only sufficient, then the effect could occur in other ways. (There are severalways one can lose one’s hearing in addition to destruction of the auditory nerve.)In like manner, a necessary condition does not mean that the effect will necessarilyoccur. (All people who consume alcohol do not become alcoholics, but one mustconsume alcohol to become an alcoholic.)

To state that we have found the cause for an event means that both the neces-sary and the sufficient conditions have been found. It means that a completeexplanation of the occurrence of the event has been isolated and that, unlike thetheory of malaria, the explanation will never change.

It is presumptuous to assume that we will ever find the conditions necessary andsufficient for the occurrence of an event, however, because the behavior of organ-isms is extremely complex. Seldom do we encounter situations or behaviors thatcannot be explained in several different ways. Popper (1968) has perhaps beenmost explicit in his insistence on the necessity of ruling out alternative explana-tions. According to inductive logic, science must be capable of deciding between thetruth and falsity of hypotheses and theories. In other words, if we conduct a scien-tific experiment testing the hypothesis that depression can be treated with psy-chotherapy, we should be able to decide if this hypothesis is true or false. Popperrejects the notion of such inductive logic. He maintains that we cannot use theresults of one or even several scientific experiments to infer that a given hypothesisor theory is true, or proven. Even if five experiments show the success of treatingdepressives with psychotherapy, this is not conclusive proof that psychotherapy cansuccessfully treat depressives. The attained relationship could be due to flaws in theexperiments or to unknown variables operating simultaneously with the psy-chotherapy. To Popper, a confirmation of an experiment states only that thehypothesis tested has survived the test. On the other hand, if the experiment fails toconfirm a prediction or a theory, the prediction or theory being tested is falsified.Therefore, Popper focuses attention on a position of falsification rather than a

70 | The Experimental Research Approach

Position of falsificationThe belief that the bestthat can be said about atheory or prediction is thatit is “not yet falsified”

Necessary conditionA condition that must existfor an effect to occur

Sufficient conditionA condition that willalways produce the effectunder study

CH03.QXD 5/22/2003 9:41 AM Page 70

position of confirmation. For him, a theory or prediction can only achieve the sta-tus of “not yet disconfirmed”; it can never be proved. In other words, if an experi-ment supports the prediction that psychotherapy is beneficial in treatingdepressives, Popper would not state that the prediction has been confirmed; hewould merely state that this prediction has maintained the status of not yet discon-firmed. This status is very precious in science, however, because it means that thetheory or prediction has passed the test of rigorous experimentation and thus onlystates one of the possible true explanations.

Deese (1972) provides yet another view of causation. He sees causation as alarge network of cause-and-effect relations. Any given cause-and-effect relationthat is isolated in a study is only one such relation embedded in a matrix of oth-ers. Consider the case of Morison’s discussion of malaria, in which he illustratesthe covariation between a number of events and malaria, each of which was onceconsidered a specific cause-and-effect relationship. The bad air of the lowlandswas found to covary with the incidence of malaria. Later it was found that thepresence of a parasite covaried with the appearance of malaria, and even later itwas found that a certain species of mosquito caused malaria because it carried themalaria parasite. Note that a number of specific cause-and-effect relationships, interms of covariation of events, were involved in the history of trying to identifythe cause of malaria. It is apparent that none of these specific relationships couldbe labeled as the cause of malaria, because many of the so-called causative events(such as the mosquitoes) still exist and yet the presumed effect of malaria nolonger occurs. For a given effect to occur, as Morison pointed out, a nicely bal-anced system of interrelated conditions must exist. For malaria to occur, the mos-quitoes and parasites must exist in a system of other specific social and economicconditions. Any one condition by itself is not sufficient to produce the effect. Pro-ponents of this view of causation advocate study of the relationship among thelevels or amounts of the variables operating within a system rather than study ofthe covariation between one variable, which can be labeled the cause, andanother, which can be labeled the effect. Such a viewpoint sees any given studyas representing only a small part of the overall system, and the relationship foundin a given study exists only if certain relationships exist among the remainder ofthe elements of the system.

It is clear that causation is subject to quite a bit of debate. [Brand (1976) pro-vides a detailed discussion of this debate.] Where does this leave the psychologistwho is attempting to identify causal relations? The behavior of organisms isextremely complex and multidetermined. It is rarely—if ever—caused by oneevent. Therefore, not only is it impossible in most instances to name the cause of anevent, but in reality a single cause for a behavior seldom exists. Given this state ofaffairs, we must conduct our scientific investigations in a manner that will enableus to identify most of the interacting causes rather than attempting to find the sin-gle cause of an event or given behavior.

S T U D Y Q U E S T I O N 3 . 1 • Why can we never identify the cause of an effect?• Identify and discuss the different ways of viewing causation.• What view or views of causation seems to be represented by

the psychological experiment?

Causation | 71

CH03.QXD 5/22/2003 9:41 AM Page 71

The Psychological ExperimentJohn Stuart Mill, in his classic 1874 analysis, identified three criteria that must bemet to identify a cause and effect relationship.

• The cause must precede the effect

• The cause must be related to the effect

• No plausible alternative explanations must exist for the effect other than thecause

These three characteristics mirror what happens in the psychological experiment.Zimney (1961, p. 18) defines a psychological experiment as “objective

observation of phenomena which are made to occur in a strictly controlled situa-tion in which one or more factors are varied and the others are kept constant.” Thisdefinition seems to be a good one because of the components that it includes,each of which are examined separately next. Analysis of this definition, with oneminor alteration, should provide a definition of an experiment, an appreciation ofthe many facets of experimentation, and a general understanding of how exper-imentation meets the three criteria set forth by Mill to identify a cause and effectenables causative relationships.

1. Objective Observation Impartiality and freedom from bias on the part ofthe investigator, or objectivity, was previously discussed as a characteristic that thescientist must exhibit. In order to be able to identify causation from the results ofthe experiment, the experimenter must avoid doing anything that might influencethe outcome. Rosenthal (1966) has demonstrated that the experimenter is proba-bly capable of greater biasing effects than one would expect. In spite of this, andrecognizing that complete objectivity is probably unattainable, the investigatormust strive for freedom from bias.

Science requires that we make empirical observations in order to arrive atanswers to the questions that are posed. Observations are necessary because theyprovide the database used to attain the answers. To provide correct answers, exper-imenters must make a concerted effort to avoid mistakes, even though they areonly human and therefore are subject to errors in recording and observation. Forexample, work in impression formation has revealed the biased nature of ourimpressions of others. Gage and Cronback (1955, p. 420) have stated that socialimpressions are “dominated far more by what the Judge brings to it than by whathe takes in during it.” Once scientists realize that they are capable of making mis-takes, they can guard against them. Zimney (1961) presents three rules that inves-tigators should follow to minimize recording and observation errors. The first ruleis to accept the possibility that mistakes can occur—that we are not perfect, that ourperceptions and therefore our responses are influenced by our motives, desires, andother biasing factors. Once we accept this fact, we can then attempt to identifywhere the mistakes are likely to occur—the second rule. To identify potential mis-takes, we must carefully analyze and test each segment of the entire experiment inorder to anticipate the potential sources and causes of the errors. Once the situationhas been analyzed, then the third rule can be implemented—to take the necessarysteps to avoid the errors. Often this involves constructing a more elaborate scenario

72 | The Experimental Research Approach

PsychologicalexperimentObjective observation ofphenomena that are madeto occur in a strictlycontrolled situation inwhich one or more factorsare varied and the othersare kept constant

CH03.QXD 5/22/2003 9:41 AM Page 72

or just designing equipment and procedures more appropriately. In any event,every effort should be expended to construct the experiment so that accurateobservations are recorded.

2. Of Phenomena That Are Made to Occur Webster’s dictionary defines phe-nomenon as “an observable fact or event.” In psychological experimentation, phe-nomenon refers to any publicly observable behavior, such as actions, appearances,verbal statements, responses to questionnaires, and physiological recordings.Focusing on such observable behaviors is a must if psychology is to meet the previ-ously discussed characteristics of science. Only by focusing on these phenomenacan we satisfy the demands of operational definition and replication of experiments.

Defining a phenomenon as publicly observable behavior would seem toexclude the internal or private processes and states of the individual. In the intro-ductory psychology course, such processes as memory, perception, personality,emotion, and intelligence are discussed. Is it possible to retain these processes if westudy only publicly observable behavior? Certainly these processes must beretained because they also play a part in determining an individual’s responses.Such processes are studied diligently by many psychologists. In studying theseprocesses, researchers investigate publicly observable behavior and infer fromtheir observations the existence of internal processes. It is the behavioral manifes-tation of the inferred processes that is observed. For example, intelligence isinferred from responses to an intelligence test, aggression from verbal or physicalattacks on another person.

In the discussion of control as a goal of science, we saw that the psychologistdoes not have a direct controlling influence on behavior. The psychologist arrangesthe consequences and the antecedent conditions that result in the behavior ofinterest. In an experiment, the experimenter precisely manipulates one or morevariables and objectively observes the phenomena that are made to occur by thismanipulation. This part of the definition of experimentation refers to the fact thatthe experimenter is manipulating the conditions that cause a certain effect. In thisway, experimenters identify the cause-and-effect relationships from experimenta-tion by noting the effect or lack of effect produced by their manipulations.

3. In a Strictly Controlled Situation This part of the definition refers to theneed to eliminate the influence of variables other than those manipulated by theexperimenter. As you have seen, control is one of the most pressing problems fac-ing the experimenter and one to which considerable attention is devoted. Withoutcontrol, causation could not be identified. Because of the magnitude of this issue, itis given extended coverage in later chapters.

4. In Which One or More Factors Are Varied and the Others Are Kept Con-stant The ideas expressed in this phase of the definition are epitomized by the ruleof one variable, which states that all conditions in an experiment must be kept con-stant except one, which is to be varied along a defined range, and the result of thisvariation is to be measured on the response variable. The two major ideas expressedin the rule of one variable are constancy and variation. Constancy refers to control-ling or eliminating the influence of all variables except the one (or ones) of interest.This requirement is necessary to determine the cause of the variation on the

The Psychological Experiment | 73

PhenomenonA publicly observablebehavior

CH03.QXD 5/22/2003 9:41 AM Page 73

response variable. If the constancy requirement is violated, the cause for the varia-tion cannot be determined and the experiment is ruined. A learning experimentcan easily illustrate the constancy component of the rule. Assume that you areinterested in the effect of the length of a list of words on speed of learning. How doesincreasing the length influence the speed with which one learns that list of words?The length of the list of words could be systematically varied and related to thenumber of trials needed to learn the list. In such an experiment, some factors thatcould influence learning speed must be controlled, including the difficulty of thewords, the participants’ ability level, the participants’ familiarity with the words,and the participants’ motivation level. Only if these factors are held constant (andtherefore do not exert an influence) can you say that the difference in speed oflearning is a function of the change in the length of the list of words. The idea of vari-ation means that one or more variables must be deliberately and precisely varied bysome given amount to determine their effect on behavior. In the learning experi-ment, the length of the list of words must be changed by an exact, predeterminedamount. The questions that frequently arise are how and how much is the variable tobe varied? The answers to these questions will be discussed later in the book.

Advantages of the Experimental ApproachCausal Inference The psychological experiment has been presented as amethod for identifying causal relationships. Indeed, its primary advantage is thestrength with which a causal relationship can be inferred. However, in looking atthis advantage, it is important to distinguish between causal description and causalexplanation (Shadish et al., 2002). Causal description refers to describing theconsequences attributable to deliberately varying a treatment. For example, manystudies have demonstrated that drugs such as Prozac help ameliorate depression.Such a study is causal description because it describes the causal connectionbetween administering the drug and the consequence of amelioration of depres-sion. However, this study does not provide an explanation of why the drugworked. This is the purview of causal explanation.

Causal explanation refers to clarifying the mechanisms by which a causalrelationship holds. In other words, causal explanation involves taking a causalrelationship and identifying the features that produce the causal relationship. Forexample, identifying a causal descriptive relationship between Prozac and amelio-ration of depression is not sufficient. After identifying this causal descriptive rela-tionship we want to know why the relationship holds. We want to know howProzac works to affect depression. Currently we know that it has an influence onthe central neurotransmitter serotonin and that serotonin is involved in depres-sion. But how is serotonin involved and why does it take some time for Prozac towork when its effect on increasing serotonin is rather immediate? There are toomany questions whose answers still remain unknown for us to have a full expla-nation of how the treatment (Prozac) mediates its influence on the outcome, theamelioration of depression.

The practical important of causal explanation is emphasized when a personwith depression obtains no relief from taking Prozac. Such instances not onlyemphasize the importance of causal explanation, but also help explain why the

74 | The Experimental Research Approach

Causal explanationClarifying the mechanismsby which a causalrelationship holds

Causal descriptionDescribing theconsequences ofdeliberately varying atreatment

CH03.QXD 5/22/2003 9:41 AM Page 74

bulk of science is directed toward explaining why and how something happens. Italso illustrates why it is easier to identify causal description and more difficult toachieve causal explanation.

Control The inferential strength that the experimental approach has in identify-ing a causal relationship is, to a large degree, obtained from the degree of controlthat can be exercised. Control, as stated in Chapter 1, is the most important char-acteristic of the scientific method, and the experimental approach enables theresearcher to effect the greatest degree of control. In an experiment, one is seekingan answer to a specific question. In order to obtain an unambiguous answer, it isnecessary to institute control over irrelevant variables by either eliminating theirinfluence or holding their influence constant. Such control can be achieved bybringing the experiment into the laboratory, thereby eliminating noise and otherpotentially distracting stimuli. Control is also achieved by using such techniques asrandom assignment and matching, which are discussed in Chapter 7.

Ability to Manipulate Variables An advantage of the experimental approachis the ability to manipulate precisely one or more variables of the experimenter’schoosing. If a researcher is interested in studying the effects of crowding on a par-ticular behavior, crowding can be manipulated in a very precise and systematicmanner by varying the number of people in a constant amount of space. If theresearcher is also interested in the effects of the gender of the participant and degreeof crowding on some subsequent behavior, male and female participants can beincluded in both the crowded and noncrowded conditions. In this way, the experi-menter can precisely manipulate two variables: gender of participant and degree ofcrowding. The experimental approach enables one to control precisely the manip-ulation of variables by specifying the exact conditions of the experiment. The resultscan then be interpreted unambiguously, because the research participants shouldbe responding primarily to the variables introduced by the experimenter.

Disadvantages of the Experimental ApproachDoes Not Test Effects of Nonmanipulated Variables Although the experi-mental research approach is the best method we have for identifying causal rela-tionships, it is limited to testing the effect of things that can be manipulated, suchas the amount of attention parents give children, the severity of punishmentadministered, or the type of therapy used to treat people with depression. Theworld in which we live includes many events that are not capable of being con-trolled by an experimenter and, therefore, not capable of being deliberatelymanipulated. For example, we cannot deliberately manipulate people’s ages, theirraw genetic material, the weather, the appearance of sunspots, or terrorists’ activ-ities. We live in a world with many nonmanipulable events that produce effects.These nonmanipulable causal events are not capable of being investigated withinthe context of an experiment.

This does not mean that we cannot or should not investigate the effects of non-manipulable events. We not only can but should by whatever means are availableand useful because investigating these events helps in the search for manipulable

The Psychological Experiment | 75

CH03.QXD 5/22/2003 9:41 AM Page 75

events that lead to solutions to problems. As we progress through this book, I dis-cuss a number of ways in which these nonmanipulable events are investigated.You will see that it is much more difficult to identify a causal relationship wheninvestigating nonmanipulable events because of the difficulty in making an unam-biguous connection between the event and the presumed cause.

Artificiality The most frequently cited and probably the most severe criticismleveled against the experimental approach is that laboratory findings are obtainedin an artificial and sterile atmosphere that precludes any generalization to a real-life situation. The following statement by Bannister (1966, p. 24) epitomizes thispoint of view:

In order to behave like scientists we must construct situations in which subjectsare totally controlled, manipulated and measured. We must cut our subjectsdown to size. We must construct situations in which they can behave as littlelike human beings as possible and we do this in order to allow ourselves to makestatements about the nature of their humanity.

Is such a severe criticism of experimentation justified? It seems to me that thecase is overstated. Underwood (1959), taking a totally different point of view, doesnot see artificiality as a problem at all. He states:

One may view the laboratory as a fast, efficient, convenient way of identifyingvariables or factors which are likely to be important in real-life situations. Thus,if four or five factors are discovered to influence human learning markedly, andto influence it under a wide range of conditions, it would be reasonable to sus-pect that these factors would also be important in the classroom. But, one wouldnot automatically conclude such; rather, one would make field tests in the class-room situation to deny or confirm the inference concerning the general impor-tance of these variables.1

The artificiality issue is a problem only when an individual makes a generalizationfrom an experimental finding without first determining whether the generaliza-tion can be made. Ideally, competent psychologists rarely blunder in this fashionbecause they realize that laboratory experiments are contrived situations. Realisti-cally, psychologists seem to frequently make risky generalizations from their work,although there are times when such generalizations are warranted. In mostinstances we cannot generalize from animals to humans. Instead, we must iden-tify the effects on organisms such as rats and then verify the existence of sucheffects using humans as the research participants. Sulik, Johnston, and Webb(1981) present one of the more dramatic instances in which a direct generalizationcan be made from animals to humans. Figure 3.1 illustrates the similarity in thepattern of facial malformations that occur in the mouse and in humans as theresult of maternal consumption of alcohol, a syndrome called fetal alcohol syndrome.

76 | The Experimental Research Approach

1From “Verbal Learning in the Educative Process” by Benton J. Underwood, Spring 1959,Harvard Educational Review, 29, pp. 107–117. Copyright © by President and Fellows of HarvardCollege.

CH03.QXD 5/22/2003 9:41 AM Page 76

This figure reveals that the pattern of facial malformations is strikingly similar,indicating a phenomenon that is common to humans and infrahumans.

Additional difficulties of the experimental approach include problems in design-ing the experiment and the fact that the experiment may be extremely time con-suming. It is not unusual for an experimenter to have to go to extreme lengths toset the stage for, motivate, and occasionally deceive the research participant. Then,when the experiment is actually conducted, the experimenter and perhaps one ortwo assistants are often required to spend quite some time with each participant.

Inadequate Method of Scientific Inquiry A final criticism that has beenaimed at the experimental approach is that it is inadequate as a method of scien-tific inquiry into the study of human behavior. Gadlin and Ingle (1975) believethat a number of anomalies inherent in the experimental approach make it aninappropriate paradigm for studying human behavior. They state that the experi-mental approach promotes the view that humans are manipulable mechanisticobjects because twentieth-century psychology mirrors the mechanistic methodand assumptions of nineteenth-century physics. Gadlin and Ingle recommend thesearch for an alternative methodology that is not fraught with such inadequacies.It appears as though this criticism has been satisfied through the use of qualitativemethodologies, although Kruglanski has pointed out that the mechanistic manip-ulable assumption exists only to the extent that the experimenter arranges a set ofconditions that may direct the individual’s behavior in a given manner. This in noway “suggests that the subject is an empty machine devoid of feelings, thoughts, ora will of his own” (Kruglanski, 1976, p. 656).

The Psychological Experiment | 77

Narrowforehead

Smallmidface

Long upper lipwith deficient

philtrum

Short palpebralfissures

Small nose

(c) (a) (b)

F I G U R E 3 . 1Malformation of facial features in the mouse and human caused by maternal alcohol consumption: (a) mouse embryo with fetal alcohol syndrome; (b) normal mouse embryo; (c) child with fetal alcohol syndrome.

(Reprinted from “Fetal Alcohol Syndrome:Embryogenesis in a Mouse Model”by K.K.Sulik, M.C.Johnston, and M.A.Webb, 1981, Science, 214, no.4523,pp.936–938 (figure on p.937), November 20, 1981.© Copyright 1981 by the American Association for the Advancement of Science.Reprinted by permission.)

CH03.QXD 5/22/2003 9:41 AM Page 77

S T U D Y Q U E S T I O N 3 . 2 • What is a psychological experiment? Explain the meaning of thecomponents of this definition.

• What are the advantages and disadvantages of the psychologicalexperiment?

Experimental Research SettingsThe experimental approach is used in both laboratory settings and field settings.Field and laboratory experimentation both use the experimental approach, butthey have slightly different attributes that deserve mention.

Field ExperimentationA field experiment is an experimental research study that is conducted in a real-life setting. The experimenter actively manipulates variables and carefully controlsthe influence of as many extraneous variables as the situation will permit. Reganand Llamas (2002), for example, wanted to find out if a female shopper’s appear-ance influenced the amount of time it took for an employee of a store to approachand acknowledge her. The basic procedure they used was to have a female con-federate dress either in formal work clothes and grooming (skirt, blouse, nylons,and dress shoes, with makeup and her hair down) or informal sports clothes andgrooming (tights, T-shirt, and tennis shoes, with no makeup, and her hair in aponytail) and then enter a randomly selected group of women’s stores betweenthe hours of 3:00 and 4:00 P.M. on two consecutive Thursdays. Upon entering thestore, the confederate activated a stopwatch and proceeded down the first openaisle, giving the appearance of shopping for clothing. As soon as an employeeapproached and spoke to her she stopped the timer. As Figure 3.2 indicates,females dressed in formal work clothes were approached more quickly by storeemployees than were females dressed in informal sports clothing.

This is an example of a field study because it was conducted in the natural set-ting of a mall while engaging in daily activities. It also represents an experimentalstudy because variable manipulation was present (type of dress) and control waspresent (the stores entered were randomly selected from those in the mall). Fieldexperiments like this one are not subject to the artificiality problem that exists withlaboratory experiments, so field experiments are excellent for studying many prob-lems. Their primary disadvantage is that control of extraneous variables cannot beaccomplished as well as it can be in laboratory experiments. In the Regan and Lla-mas study, even though the clothing stores were randomly selected from thoseavailable in the mall, the study was limited to only the types of store that rent froma mall. In addition, the store employees were not a random selection of possibleemployees. Consequently, a selection bias may have existed. Even though it is moredifficult to exercise control in field experiments, such experiments are necessary.

Tunnell (1977) states that field experimentation should be conducted in amanner that makes all variables operational in real-world terms. The Regan andLlamas (2002) study included the three dimensions of naturalness identified byTunnell: natural behavior, natural setting, and natural treatment. The natural

78 | The Experimental Research Approach

Field experimentAn experimental researchstudy that is conducted in areal-life setting

CH03.QXD 5/22/2003 9:41 AM Page 78

behavior investigated was a store employee approaching a shopper. The settingwas natural because the study took place in a mall; the natural treatment was typeof dress. In reality, the treatment was imposed by a confederate, but it mirrored abehavior that could have occurred naturally. These are the types of behaviors Tun-nell says we must strive for when we conduct field experimentation, as opposed toasking participants to provide self-reports or to recall their own behavior in someprior situation. Asking for such retrospective data only serves to introduce possiblebias into the study.

Laboratory ExperimentationThe laboratory experiment is the same type of study as the field experiment, butwhere the field experiment is strong the laboratory experiment is weak, and wherethe laboratory experiment is strong the field experiment is weak. The laboratoryexperiment epitomizes the ability to control or eliminate the influence of extrane-ous variables. This is accomplished by bringing the problem into an environmentapart from the participants’ normal routines. In this environment, outside influ-ences (such as the presence of others and of noise) can be eliminated. However, theprice of this increase in control is the artificiality of the situation created. This issuewas covered in detail in the discussion of the disadvantages of the experimentalapproach. Even though precise results can be obtained from the laboratory, theapplicability of these results to the real world must always be verified.

The laboratory experiment is a study that is conducted in the laboratory andin which the investigator precisely manipulates one or more variables and controlsthe influence of all or nearly all of the extraneous variables. For example, Kassinand Kiechel (1996) realized that there were police reports of individuals who con-fessed to crimes that they had not committed. They realized that there was noempirical proof of this phenomenon and were interested in determining if theycould experimentally demonstrate that vulnerable individuals, under the right

Experimental Research Settings | 79

80

70

60

50

40

30

20

10

0

Formalclothing

Informalclothing

Am

ount

of t

ime

in s

econ

ds

Employee approachesa shopper.

F I G U R E 3 . 2Amount of time for a storeemployee to approach andacknowledge theconfederate.(Adapted from “Customer Service as a Function of Shopper’s Attire”by P.C.Regan and V.Llamas, 2002, PsychologicalReports, 90, pp.203–204.)

Laboratory experimentAn experimental researchstudy that is conducted inthe controlledenvironment of alaboratory

CH03.QXD 5/22/2003 9:41 AM Page 79

circumstances, would confess to an act that they did not commit and internalizethis confession to the point that they would confabulate details in memory consis-tent with the confession. To investigate this phenomenon, Kassin and Kiechel hadto construct a situation in which they manipulated the vulnerability of theresearch participants as well as the presence of a person falsely incriminating them.In addition, they had to control other variables such as the presence of witnesses,other individuals refuting or confirming the false accusation.

To precisely manipulate vulnerability and the presence of a witness and to con-trol for the impact of extraneous variables, Kassin and Kiechel created a situationwithin the context of a laboratory setting in which the research participants had toperform a task at either moderate or rapid speed. A rapid-speed completion of thetask created a vulnerable condition because the more rapidly the participants hadto respond, the greater the likelihood of making a mistake. The results of this studyrevealed that individuals were more likely to confess to making a mistake they hadnot made in the vulnerable condition and when a confederate, or witness, saidthat the research participant had made the error. More important, these vulnera-ble individuals were more likely to internalize the false confession and tell othersthat they had committed the error.

S T U D Y Q U E S T I O N 3 . 3 What are the different research settings in which experimental research isconducted and what are the advantages and disadvantages of each setting?

Summary The experimental approach is a quantitative research method in which oneattempts to identify cause-and-effect relationships by conducting an experiment.However, the cause of an effect is difficult to identify because all causal relation-ships are contextually dependent. In spite of this, it is useful to think of a cause assomething that produces something else. An effect is the difference between whatwould have happened without some treatment and what did happen when thetreatment was administered. Although an exact measure of this difference is notpossible to obtain, we approximate it within the context of the experiment.

To be able to state that a causal relationship exists, even though this relation-ship is contextually dependent, we need evidence. John Stuart Mill set forth fourcanons—the methods of agreement, difference, concomitant variation, and thejoint methods of agreement and difference—that he said could be used in identi-fying causation. However, in order for one to be able to state that the cause of agiven effect has been found, this condition must qualify as being both necessaryand sufficient. Because behavior is multidetermined, it is highly unlikely that onecan overrule all possible alternative explanations for behaviors. This is why Popperrejects the confirmationists’ position and takes the position of falsification, main-taining that the best status a theory can attain is one of “not yet disconfirmed.”Deese believes that causal relations are embedded in a matrix of other causal rela-tions. A given relationship between a cause and an effect will continue to existonly if all the other variables within the matrix or system remain constant.

The psychological experiment achieves the goal of the experimental approachby allowing the researcher to observe, under controlled conditions, the effects of

80 | The Experimental Research Approach

CH03.QXD 5/22/2003 9:41 AM Page 80

systematically varying one or more variables. The experimental approach has theprimary advantage of being able to infer causal relationships. However, it is easierto identify causal description, which describes the consequences of deliberatelyvarying a treatment, than it is to achieve a causal explanation, which clarifies themechanisms by which a causal relationship holds. A second advantage of theexperiment is that it controls for the influence of extraneous variables. Otheradvantages are that it permits the precise manipulation of one or more variables,produces lasting results, suggests new studies, and suggests solutions to practicalproblems. The experimental approach has the disadvantages of not being able totest for the effects of nonmanipulated variables, creating an artificial environment,and frequently being time consuming and difficult to design.

The experimental approach is used in both field and laboratory settings. In afield setting, the researcher makes use of a real-life situation and thereby avoidscriticism for having created an artificial environment. Typically, however, there isnot as much control over extraneous variables. In a laboratory setting, the experi-menter brings the participants into the laboratory, where there is maximum con-trol over extraneous variables; however, this usually means creating an artificialenvironment.

Key Terms and Concepts

Related http://gateway1.gmcc.ab.ca/~digdonn/psych104/think.htmInternet Sites This site includes an exercise that focuses on distinguishing between correlational stud-

ies and experiments. Use of this site should further solidify the distinction betweenthese two types of studies and the conclusions that can be drawn from them.

http://www.cia.gov/csi/books/19104/art14.htmlThis Internet site discusses many of the perceptual biases we have in arriving at causalconclusions.

http://www.nizkor.org/features/fallacies/confusing-cause-and-effect.htmlThis site provides a brief discussion of some of the things that cause us to falsely draw acausal connection between two variables.

http://www.isd77.k12.mn.us/resources/cf/SciProjInter.htmlThis site gives a step-by-step approach to conducting an experimental study and givesa slightly different description of the steps than is presented in this book. Reading thissite will show that the basic steps are the same, but that different people focus onslightly different components of the scientific method.

Related Internet Sites | 81

Experimental researchCausationCauseEffectMethod of agreementMethod of differenceJoint methods of agreement

and differenceMethod of concomitant variation

Necessary conditionSufficient conditionPosition of falsificationPsychological experimentPhenomenonCausal descriptionCausal explanationField experimentLaboratory experiment

CH03.QXD 5/22/2003 9:41 AM Page 81

http://www.fordham.edu/halsall/source/beacon2.htmlThis site gives Roger Bacon’s 1268 discussion of what he called “Experimental Science.”It is an interesting historical presentation of some of the initial approaches to scientificthinking.

http://plato.stanford.edu/entries/causation-probabilistic/ andhttp://plato.stanford.edu/entries/causation-backwards/These two sites discuss probabilistic causation and backward causation.

Practice Test Answers to these questions can be found in Appendix A.

1. To think of a cause as an event that produces some other eventa. Is too simplistic a representation of cause.b. Refers to the contextual relationship between a cause and effect.c. Reveals that a cause is a probabilistic event.d. Refers to the difference between what would have happened and what did

happen to something.

2. Assume that one hundred studies have indicated that alcoholics, to remainsober, must totally avoid alcohol. However, one study refutes this finding. If asa result of this one study you take the position that alcoholics do not necessar-ily have to avoid alcohol, you have based this position on which approach tocausation?a. Method of differenceb. Joint method of agreement and differencec. That alcohol is a necessary but not sufficient condition to become an

alcoholicd. Position of falsificatione. That alcoholism exists within a system of other factors

3. Dr. Zilstein knew that many children with Attention Deficit Hyperactivity Dis-order take the drug Ritalin. He conducted another study in which he was try-ing to find out how the drug Ritalin worked in the brain to produce itsbeneficial effect. What was he trying to achieve with this study?a. Causal descriptionb. Causal explanationc. Control of the effect of the drugd. Better description of the effect of the drug on children’s behaviore. Causal prediction

4. The September 11, 2001, terrorist attack on the World Trade Center causedemotional turmoil for many individuals. Assume that you were interested inconducting an experimental study to assess the effects of this terrorist attackand I stated that you could not do so. I would have based my statement onwhat factor?

82 | The Experimental Research Approach

CH03.QXD 5/22/2003 9:41 AM Page 82

a. The fact that the study would require you to identify individuals who lostloved ones in that attack

b. The fact that the study would have to be conducted outside the context ofa university-based laboratory

c. The fact that you would not have control over the manipulation of thecausal variable of interest

d. The fact that individuals would probably not want to participatee. The fact that it is not possible to study such events

Challenge The challenge exercises included in this chapter represent topics for discussion. For Exercises most of them I have included a Web site (see the Related Internet Sites) that

addresses each of these topics. You may want to login to these Web sites to getmore information on each of these topics.

1. When we conduct our experiments we attempt to identify a cause-and-effectrelationship. Is it more accurate to say that any relationship we find is deter-ministic or probabilistic? In other words, would it be more accurate to state thatthe presumed cause determined the effect or that the presumed causeincreased the probability of the effect occurring? Make sure that you explainand defend your answer.

2. Prior to discussing the psychological experiment I stated that one of the criteriafor identifying cause-and-effect relationships is that the cause must precede theeffect. Is it possible for the effect to precede the cause and what would be nec-essary for this to happen?

3. Consider the two variables of inflation and unemployment. Should a causalrelationship exist between these two variables, and which one should be thecause and which the effect?

4. Consider each of the following situations and identify the presumed cause andthe presumed effect. Then discuss the likelihood that the presumed cause actu-ally did produce the observed effect. Explain why someone might think thesetwo variables were causally related and then consider the fact that other vari-ables could also have produced the effect.a. The Republicans passed a law giving a tax break that benefits wealthy

Americans. Shortly after the tax break went into effect, the stock marketwent down and the economy went into a recession. The Democratsclaimed that the tax break caused the economic decline and attempted torepeal the tax break.

b. Bill purchased a new piece of software for his computer and installed itimmediately. The next time he started his computer, it froze up on him, soBill concluded that it was the software that caused the computer to freeze.

Challenge Exercises | 83

CH03.QXD 5/22/2003 9:41 AM Page 83