Psychotherapy Volume 30/Spring 1993/Number 1

ELEMENTS OF THE SOCRATIC METHOD:II. INDUCTIVE REASONING

JAMES C. OVERHOLSERCase Western Reserve University

The Socratic method uses systematicquestioning and inductive reasoning tohelp clients derive universal definitions.The present manuscript describesinductive reasoning as used inpsychotherapy. Inductive reasoning canbe based on enumerative generalizations,analogical comparisons, and eliminativecausal reasoning. Enumerativegeneralizations use pattern identificationto support a conclusion about an entiregroup of events. Analogical comparisonshelp clients transfer knowledge fromfamiliar to novel situations. Eliminativecausal reasoning involves manipulatingenvironmental conditions to examinepossible causes of specific problematicevents. After a probable cause has beenidentified, it can be modified throughtreatment. The clinical utility andfallacies of logical reasoning arediscussed as they apply to each of thethree forms of inductive reasoning. Ingeneral, inductive reasoning plays acentral role in the Socratic method andcan be a valuable tool in psychotherapy.

The Socratic method plays an important role incognitive therapy (Beck, Rush, Shaw, & Emery,

I am indebted to Mark Fine, Abe Wolf, Sue Knell, JaneKessler, Dave Brinkman, Dalia Adams, Hilary Einhom Katz,Patti Watson, and three anonymous reviewers for valuable andsupportive comments on earlier versions of this manuscript.

Correspondence regarding this article should be addressedto James C. Overholser, Department of Psychology, CaseWestern Reserve University, Cleveland, OH 44106-7123.

1979; Beck & Emery, 1985), rational-emotivetherapy (Ellis, 1962), and psychodynamic ther-apy (Rychlak, 1968; Stein, 1991). Despite itsvalue, die Socratic method rarely is described inadequate detail. The Socratic method uses sys-tematic questioning to guide the flow of inductivereasoning in therapy sessions. The goal of theSocratic method is to cultivate abstract conceptualskills (Nelson, 1980). Often, the focus is on help-ing the client to derive a universally applicabledefinition of an abstract concept relevant to ther-apy (e.g., love, trust, success, friendship).Throughout this process, the therapist and clientcollaborate in the search for knowledge anddefinitions. Both therapist and client attempt tominimize their preconceived beliefs in order toremain cautious and skeptical about the infor-mation they possess. Although potentially capa-ble of becoming a distinct approach to psycho-therapy, the Socratic method is best viewed as astyle of clinical interviewing that is compatiblewith most forms of psychotherapy. The clinicalapplication of systematic questioning has beendescribed in a previous report (Overholser,1993). The present manuscript describes the useof inductive reasoning as part of the Socraticmethod.

Inductive reasoning is used to draw generalinferences from experience with specific eventsand therefore can be used to help clients tran-scend their personal experiences and construct abroad view of reality (Rescher, 1980). Inductivereasoning can help clients distinguish betweenfacts, beliefs, and opinions (Gambrill, 1990).The basic process in inductive reasoning involvesanalyzing similarities and differences among spe-cific experiences in order to extract a generalprinciple about a class of events (Pellegrino,1985). As used in psychotherapy, inductive rea-soning helps clients develop appropriate expecta-tions and coping strategies at a fairly abstractlevel. Three forms of inductive reasoning have

75

James C. Overholser

been identified: enumerative generalizations, an-alogical comparisons, and eliminative causal rea-soning. Each will be described as it relates topsychotherapy.

Enumerative GeneralizationsInduction by enumeration involves identifying

a general category that applies to all members ofa group based on experience with only somegroup members (Zechmeister & Johnson, 1992).Ideally, a strategy can be developed that is bothsensitive (i.e., identifies all members of thegroup) and specific (i.e., identifies only membersthat belong to that specific group) (Medin & Bar-salou, 1987). However, most categories have noclear boundaries (McCloskey & Glucksberg,1978), thus making the generalization processvery complicated.

Clinical Utility of Enumerative GeneralizationsThe goal of the enumeration process is to com-

bine information into meaningful units so as toreduce the massive amount of information clientsneed to understand (Evans, 1989). Using thisenumeration process, clients can learn to reasonfrom specific events to general principles, classi-fying specific entities into general categories(Burks, 1980). This provides much useful infor-mation about all members of that category(Medin & Barsalou, 1987) allowing clients tocomprehend situations in which limited infor-mation is available (Lingle, Altom, & Medin,1984).

Clients often perform enumerative generaliza-tions in a faulty and idiosyncratic manner. Manytypes of irrational thinking are based on inaccu-rate overgeneralizations (Ellis, 1977). Clientstend to seek information that supports their pre-established beliefs (Evans, 1989; Snyder &Swann, 1978). This bias for confirmatory evi-dence is more likely to be present when testinghypotheses about oneself (Strohmer, Moilanen,& Barry, 1988) than with hypotheses about otherpeople (Strohmer & Newman, 1983). Thus,many cognitive distortions about oneself developa self-perpetuating nature. Therapy is needed tocorrect and replace these idiosyncratic and dys-functional cognitions. Therapy can be useful inencouraging clients to seek a variety of informa-tion before inferring a generalization.

Inductive generalizations involve the gradualaccumulation of evidence. Over time, clients dis-cuss a variety of events in their lives. As the

information accumulates, it becomes easier fortherapist and client to identify patterns in the cli-ent's life. The generalization process can be ap-plied by asking the client a series of questions(e.g., "How does this problem relate to the thingswe discussed last week?" "What can we learnfrom the conflict you had with your parents thatcould help us here?"). This process helps clientssee patterns in their lives and begin to anticipaterecurrent problems. Then, useful coping strate-gies can be identified and implemented.

Subtypes of Generalizations

Inductive reasoning can be used to infer gener-alizations about people, events, goals (Medin &Smith, 1984) and abstract social concepts (Lingleet al., 1984). Generalizations about people oftenidentify categories of people (e.g., extrovert) andevaluate the extent to which a person fits the gen-eral category. Clients develop social expectationsbased on these categories. However, clients mayinappropriately generalize from a small sampleto the population as a whole. For example, asocially anxious male reported "Mary is single,unattached, and pretty, and she rejected my ad-vances. Sue is single, unattached, and pretty.Therefore she will reject my advances." The cli-ent developed the view that all attractive womenwere too good for him and would not be inter-ested in getting to know him. However, this cli-ent failed to take into account other relevant con-ditions such as their personal preferences, sexualorientation, and the timing of his interactionswith these women.

Generalizations about people also apply to theview clients hold toward themselves (i.e., self-concept). Clients with narrowly defined self-con-cepts are more likely to be influenced by lifestress and vulnerable to depression and physicalillness (Linville, 1985; 1987). Furthermore, cli-ents often focus on one type of life experience,creating a biased view of their self-concept. Forexample, a pessimistic client reported: "My mar-riage is terrible, I can't find a decent job, andI've put on 15 pounds. What a loser I've turnedout to be." This client needed to learn that hecould be a good person despite poor performancein certain areas of his life. Discussing what"loser" meant to the client helped him distancehimself from these recent events and identify abroader sample of events from which to judgehimself. Instead of focusing on his view of being

76

Inductive Reasoning

a failure, this client learned to gather evidencedocumenting his success in various activities.

Generalizations about events cluster together anumber of common actions comprising a particu-lar type of event or repetitive problem situation.For example, a client complained her life was"terrible." When asked to explain what made herlife so terrible, she complained that no one re-membered her birthday. Another example of "ter-rible" occurred when she had a flat tire in a badneighborhood on a rainy night. However, whenasked to define terrible in the abstract, she con-cluded terrible events cause permanent harm toa person. She was able to see her problems wereminor inconveniences, not terrible events. Al-though her flat tire could have resulted in terribleconsequences, she realized it was best viewed asan inconvenience.

Generalizations regarding goals focus onqualities that help the client attain a desiredoutcome. Goal-related concepts are influencedby the frequency and strength with which a par-ticular action is related to the final goal (Medin& Smith, 1984). For example, a depressed cli-ent reported: "I've applied for three jobs in ad-vertising, and didn't get any of them. I knowI'll never find a good job." This generalizationwas biased because the observed sample didnot adequately represent the characteristics ofthe population as a whole. In actuality, the cli-ent was not qualified for a job in advertising,but could easily get a job in sales. Therapeuticdiscussions helped this client see how his pessi-mism had biased his reasoning. Questionsforced him to define what a "good job" meantto him, and therefore learn to broaden his scopeof possible employment opportunities.

Generalization principles also apply to abstractconcepts such as truth, justice, friendship, cour-age, or beauty (Lingle et al., 1984). For abstractconcepts, the generalization process becomesambiguous and relies on the client's interpreta-tion more than observation (Gergen, 1988; Lin-gle et al., 1984). Clients often set themselves upfor failure by looking at their lives from extremeperspectives. Bifurcation occurs when clients ex-amine only two possible alternatives (e.g., suc-cess vs. failure), thereby ignoring the various in-termediate gradations. A false dilemma is createdwhen the client implies that only two outcomesare possible, one good and one unacceptable out-come (Zechmeister & Johnson, 1992). For exam-ple, a client emphasized the goal of perfection

throughout her everyday life. She felt that when-ever her performance was less than 100%, it wasa complete failure. Therapy and assignmentswere used to help her appreciate the vast arrayof possibilities between perfection and failure.These involved recording activities throughouther day and rating her success from 1-99%, en-couraging a range of ratings. Furthermore, shewas occasionally given paradoxical assignmentsin which she was instructed to strive for a medio-cre performance so that obtaining 100% wouldbe a failure of the assignment, whereas 50%would be optimal.

Process of Generalization as Usedin Psychotherapy

The generalization process involves threesteps: the critical evaluation of the client's gener-alization, pattern identification to revise the gen-eralization, and validation to test the new gener-alization. The use of generalizations often beginswhen the client makes a statement that includesa broad generalization. Critical evaluation is usedto identify the logical implications of the client'sgeneralization. Therapists help clients examinethe implications of their generalizations by elic-iting further statements that support or contradictthe original generalization (Nelson, 1980). Ther-apist and client use a methodical search for un-derlying contradictions that may impede the gen-erality of the statement (Navia, 1985; Seiple,1985). For example, a client complained he hadno "real friends." He was asked to define whathe meant by a "real friend." He stated a realfriend was someone he could trust with any pos-session or information he had, someone withwhom he could enjoy many social activities, andsomeone that liked him even when he was in abad mood. Therapist and client discussed differ-ent people in the client's life to evaluate thesecriteria. The client realized there were severalpeople he trusted and enjoyed, and who couldbe seen as real friends. Alternatively, the clientremembered several "real friends" from his pastwho had not met these criteria. Thus, he neededto redefine what he meant by a "real friend" soas to include the friends he already had.

After identifying weaknesses in the originalgeneralization, therapist and client collabora-tively develop a revised generalization based oncommon elements across different examples. To-gether, therapist and client construct the meaningof key terms (Keeney, 1987). The therapist re-

77

James C. Overholser

quests additional information through questionsor self-monitoring assignments conducted be-tween sessions. Pattern identification occurswhen therapist and client work together to iden-tify common elements across the diverse items(Chessick, 1982; Lingle et al., 1984) by observ-ing a sample of events and detecting underlyingregularities that allow further instances to be cat-egorized accurately (Harnad, 1987). The general-ization may need to be broken down into its com-ponent parts to identify similarities amongcategory members (Medin & Barsalou, 1987).Pattern identification is influenced by the avail-ability, relevance, and vividness of the informa-tion obtained (Evans, 1989) and the frequencyand recency with which that category has beenused (Lingle et al., 1984). Clients must learn tofocus on relevant and ignore irrelevant similari-ties (Reed, 1972).

Finally, validation involves testing the gener-alization under a variety of circumstances. Bothactual and hypothetical tests may be used. Thera-pist and client evaluate the strength and diversityof evidence supporting the generalization. As thesize and representativeness of the observed sam-ple increases, the probability of an accurate gen-eralization increases (Salmon, 1973; Zechmeister& Johnson, 1992). If a generalization is basedOB a sample that does not adequately representHie group as a whole, the generalization will notreflect die properties of the larger population.Enumerative generalizations are strengthenedby: 1) a greater number of supporting instances,2) a variety of supporting examples, and 3) abalanced distribution of types of examples (Car-nap, 1945). The number of instances is the leastimportant element, being neither necessary norsufficient for a proper inductive generalization(Manicas & Kruger, 1976). Thus, instead of sim-ply gathering a large number of supportive exam-ples, it is better to seek corroborating informationunder a wide variety of circumstances (Black,1952). If the original generalization cannot ac-count for all possible cases, a new generalizationis proposed and evaluated. A cycle of proposingand refining parameters is used to improve theidentification of features that define a categoricalgeneralization (Fried & Holyoak, 1984).

Limitations of Inductive Generalizations

Inductive generalizations are artificial group-ings imposed on natural variations and arestrongly influenced by language and culture (Ger-

gen, 1985). Furthermore, induction by enumera-tion is limited by its susceptibility to falsification(Popper, 1965). A generalization based on numer-ous supporting examples can be falsified by onecounter-instance (Carnap, 1966). In order to pro-tect against falsification, the supporting examplesshould be as diverse as possible (Carnap, 1966)and the client should check for possible counter-examples (Holland, Holyoak, Nisbett, & Tha-gard, 1986). Thus, clients need to ask them-selves: "Are the examples relevant? Typical?Adequate?."

Risk of falsification can be reduced by framinghypotheses in terms of probabilities or percent-ages instead of universal statements. Inductivereasoning estimates the probability of a conclu-sion depending on the quality and quantity ofevidence supporting it (Zechmeister & Johnson,1992). Thus, enumerative generalizations canprovide estimates about the probability of differ-ent alternatives (Burks, 1980).

Despite the importance of the falsification ap-proach, it is important to avoid overemphasiz-ing the search for conflicting data. It is often use-ful to refine or expand the generalization so itcan accommodate the conflicting data (Lakatos,1970). In essence, a generalization should not bediscarded until a better one is available (Lakatos,1970). The old generalization often retains itsusefulness, but with some limitations that havebeen identified by the counter-instances (Pop-per, 1968).

Analogical ComparisonsInductive analogies build on enumeration pro-

cesses (Hesse, 1968). However, instead of infer-ring from an observed sample to the entire popula-tion as is common in enumerative generalizations,analogies infer from one sample to another (Simco& James, 1976). Reasoning by analogy involvescomparing two objects, events or people based onrelevant but not obvious similarities. The observedsimilarities are assumed to indicate other similarit-ies exist but have not yet been identified. Analogiesare used to emphasize abstract properties that areshared across situations (Holland et al., 1986).Thus, inductive analogies lay the foundation forconclusions that extend well beyond any observedsimilarities.

Clinical Utility of Analogical ReasoningThe use of analogies can increase the client's

cognitive flexibility (Brown, 1989), helping cli-

78

Inductive Reasoning

ents understand their problems from a differentperspective. Clients are encouraged to step be-yond their conventional views of the problem anddevelop a new perspective. However, the utilityof the analogy depends on the client's ability torecognize and use similarities between the noveland familiar situations (Holland et al., 1986).Reasoning by analogy involves looking beyondsurface similarities (appearances), and examiningthe structural similarities (functions) of two pro-cesses (Holyoak & Koh, 1987), preserving therelational structure between the familiar and thenovel situation (Johnson-Laird, 1989). If this isdone properly, analogies help reduce new andcomplex problems into a simpler and more famil-iar core (Spiro et al., 1989). A sense of reliefoccurs when people realize that new problemscan be dealt with by transferring familiar infor-mation to the new problem, reducing the client'srisk of feeling frustrated and overwhelmed(Bransford, Franks, Vye, & Sherwood, 1989).

Subtypes of AnalogiesAnalogies can be intradomain and interdomain

(Holland et al., 1986). Intradomain analogies in-volve examining similar problems from the past,or similar problems in different situations (e.g.,divorce versus retirement). Clients learn to mapknowledge from one area to another, maintainingthe relationships among problem and solution(Gentner, 1989). The client can be assisted inidentifying coping strategies found effective andthose ineffective in dealing with the originalproblem. Then, parallel relationships can bedrawn between the original and current problems.

Interdomain analogies involve abstract similar-ities across blatantly different events. Interdo-main analogical arguments can be based on atleast five different content areas: natural, medi-cal, mechanical, strategical, and relational com-parisons. Analogies based on natural processesattempt to transfer information from some well-known aspect of nature and apply it to a parallelproblem situation. For example, a comparisoncan be drawn between snow on roofs during abad winter and emotional strain during a stressfulperiod. Some houses have steeply pitched roofswhere everything that hits them blows off theroof. Other homes have flat roofs requiring dif-ferent care. For these homes, it takes work to getup on the roof and shovel off the snow before itcauses unnecessary strain and excessive pressureon the foundation and structure of the house.

Analogies based on medical practices involvea description of basic medical procedures as re-lated to the problem areas. For example, the de-scription of someone getting dirt in a wound canbe used to describe how a sore will fail to healunless properly tended. Although it may hurt, thewound must be opened and cleaned (psychother-apy) in order to reduce the infection (psychopa-thology) that may develop. It is important toclean the wound in a sterile environment (psy-chotherapy) with a skilled professional.

Analogies based on mechanical processes ex-trapolate from the actions seen in mechanical op-erations. For example, describing a car as run-ning on four cylinders, where each cylinder isnecessary for smooth running, can be used topoint out the need for human functioning to bebased on more than one focus. Human function-ing can be described as based on four primaryelements: work, family, friends, and spiritual be-liefs (or other areas relevant to a particular cli-ent). Complete neglect of any element will dis-rupt overall performance, much the same way asa car attempting to run on only three cylinders.If any one of these "cylinders" is not working, itdisrupts the functioning of the entire person.

Strategical analogies are based on the tacticalstrategies involved in various activities, such asthe patience required in fishing, or the planningrequired in chess. Clients who feel comfortablewith a behavior or attitude when using it in onesituation may learn to apply the same behaviorto another. For example, Zimbardo (1977) hasdescribed the shy person as similar to a conserva-tive investor in a risky economic market, beingmore concerned about risk of loss than with pos-sibility of gain. Shy clients can learn to appreci-ate the benefits of risk-taking and begin at-tempting new behaviors in social situations.

Relational analogies are based on parallel rela-tionships between two or more parties. For exam-ple, clients can learn to extrapolate from the na-ture of the relationship of parent to child, coachto players, or general to troops. Various aspectsof the interactions can be used to exemplify newways of managing current interpersonal prob-lems. For example, one client with low self-es-teem worked as a court reporter. It was helpfulfor her to see how her difficulties identifying, herpositive features were similar to a jury that wasbiased to convict her. She tended to discount ordisqualify any positive features without objec-tively testing the evidence. In session, courtroom

79

James C. Overholser

procedures were used playfully to challenge theclient's report of recent events as if she were abiased witness. Over time, she learned to ap-praise her performance from a less criticalperspective.

Process of Analogies as used in Psychotherapy

The process of using analogical comparisonsin psychotherapy begins with systematic observa-tion. The client describes a problem situation andthe therapist uses questions or assignments toelicit background information about the problemsituation. After background information has beenobtained, pattern identification is used to identifythe main elements of the problem. Then, the ther-apist either proposes or requests the client finda parallel but familiar problem situation whichalready has been solved. Transfer of knowledgeinvolves identifying the corresponding rolesacross the two situations (Holland et al., 1986).After problem and solution are transferred fromthe analogy to the current problem situation, ther-apist and client evaluate the accuracy of thetransfer.

Analogies can guide a problem-solving pro-cess by transferring knowledge from familiar tonovel situations (Gick & Holyoak, 1980; 1983;Vosniadou & Ortony, 1989). Analogies facilitatethe generation of alternatives, identifying newcoping options stemming from the analogy. Ananalogy provides the starting point for a hypothe-sis (Werkmeister, 1957). Often, the working hy-pothesis is geared toward the future, providinginsights into new ways of coping with a problemsituation. Thus, analogies often involve a "For-ward search," asking the client "Given the re-sources available, what can be done to solve thisproblem?" (Holyoak & Thagard, 1989).

Verification involves therapist and client plan-ning to implement the new coping option. Analo-gies can paint a picture that is easy for clientsto remember (Martin, Cummings, & Hallberg,1992). Therefore, clients will be more likely tochange their behavior when confronted with theproblematic situation again. If successful, thecommon principle underlying both the problemand its analogy should be extracted (Gentner,1989).

Limitations of AnalogiesThe utility of an analogy depends on the rele-

vance of the comparison dimensions (Pellegrino,1985). It is assumed that the current situation in

the client's life is similar in some ways to theissues revealed through the analogy. To be effec-tive, the analogy must preserve the relationshipbetween problem and solution (Gordon, 1978).A faulty or misleading analogy can be producedby a comparison of irrelevant similarities (Darner,1987).

Eliminative Causal ReasoningInduction by elimination relies on principles

developed by David Hume (1748) and John Stu-art Mill (1890) to identify the necessary and suf-ficient conditions for an event to occur. The prin-ciple of elimination states that any factor that isabsent when the effect occurs can be eliminatedas a necessary condition for the effect (Skyrms,1986). Conversely, a factor that is present when-ever the effect has not occurred can be eliminatedas a sufficient condition for the effect (Skyrms,1986). Thus, causal reasoning involves examin-ing the occurrences of an effect under a varietyof circumstances (Skyrms, 1986).

Clinical Utility of Causal Reasoning

Attributions regarding causality play an im-portant role when understanding and decidingupon a course of action (Kelley, 1973). Causalreasoning most often occurs in response to unex-pected events (Hastie, 1984) and situationswhereby the client fails to obtain a desired goal(Weiner, 1985). Causal reasoning helps clientsunderstand, predict, and control their behavior.For example, problems like insomnia, impo-tence, and loneliness can be caused by numerouscommon factors. Learning which of many possi-ble causes needs to be controlled can becomeconfusing and overwhelming. Effective treatmentrelies on understanding and changing the causalsequence (Overholser, 1991). Although classify-ing one event as the cause of another can besomewhat arbitrary (Efran, Lukens, & Lukens,1990), clarifying possible causal and maintainingfactors can lay the foundation for treatment plan-ning. By changing environmental or intrapsychicfactors, the presenting problem often can be alle-viated. If a necessary causal condition can bestopped from occurring, the problem can be pre-vented (Skyrms, 1986).

Causal attributions are more susceptible to theinfluence of others if the person has low self-confidence, little social support, or has been feel-ing overwhelmed by problems (Kelley, 1967).Also, depressed clients are likely to distort the

80

Inductive Reasoning

causal attribution process so as to blame them-selves for negative events (Sweeney, Anderson,& Bailey, 1986). The attribution process maybecome biased so as to perpetuate a negativeview of oneself (Strohmer et al., 1988). Thus,many clients can benefit from therapy that fo-cuses on developing different causal attributions.

Subtypes of Causal Reasoning

Causal reasoning is based on strategies of con-firmation and disconfirmation (Downing, Stem-berg, & Ross, 1985; Mill, 1890; Schustack &Steinberg, 1981). Confirmation involves identi-fying a cause by (1) the joint presence of causeand effect and (2) the joint absence of presumedcause and effect. Thus, causal reasoning basedon confirmation examines both the occurrenceand nonoccurrence of an effect. It attempts toidentify possible causes that are associated withthe presence but not the absence of an effect. Forexample, a client with mild but chronic depres-sion was asked to record her mood and activitiesthroughout the week. She quickly noticed thatwhen she was alone and inactive, her moodworsened. When she was active and socializing,her mood improved. This hypothesis was thentested by encouraging the client to initiate a so-cial activity whenever she felt herself becomingdepressed.

Disconfirmation involves critically examininga possible cause by looking for (1) the presenceof the cause without the effect or (2) the presenceof the effect without the cause. In either case,the causal hypothesis would be rejected and anew one sought. However, most people empha-size occasions in which both presumed cause andeffect are present, and minimize the importanceof disconfirming evidence (Downing et al., 1985;Schustack & Sternberg, 1981). The emphasis onconfirmation instead of disconfirmation encour-ages evaluating whether a cause is sufficient(rather than necessary) for an effect to occur(Downing etal. , 1985).

Although the confirmation approach can beuseful, it can be misleading. For example, a cli-ent with a panic disorder observed her panicsymptoms over time. She noticed that her panicwas associated with certain events (e.g., whenshe felt tired or hungry, while bending over orclimbing stairs). Disconfirmation involved con-fronting these events to see if they did in factaggravate her symptoms. She found that fatigueand dietary factors played no role in her panic,

and the combined activities of climbing stairs andbending over produced mild dizziness but notpanic. Therefore, these could be eliminated assufficient causes of her panic.

Process of Causal Reasoning as Usedin Psychotherapy

Causal reasoning builds on the generalities es-tablished through enumerative induction (Burks,1980). Causal explanations are often influencedby the questions asked of the client (Hilton,1990). Thus, systematic questioning continues toplay an important role. The process of usingeliminative causal reasoning in psychotherapy in-volves systematic observation of the problem be-havior, hypothesis formation, and hypothesistesting.

Systematic observation occurs when the clientdescribes a problem situation and the therapistrequests additional information via questions orassignments. Gathering a wide range of informa-tion helps encourage the evaluation of many pos-sible causes (Amoult & Anderson, 1988). Forexample, one woman whose child had drownedseveral years earlier still experienced "unpredict-able" episodes of severe depression. She re-corded fluctuations in her mood over time andrealized her depression was associated withplaces, events, and images that reminded her ofher son or his death.

Hypothesis formation occurs when the thera-pist and client collaboratively identify commonantecedent events and develop a preliminarycausal hypothesis. Causal hypotheses are basedon covariation in which cause and effect system-atically covary over time (Kelley, 1973). Cuesto identifying a cause include (1) the consistentcovariation between cause and effect, (2) causepreceding effect in time (Einhorn & Hogarth,1986), (3) temporal contiguity in which there aretimes when both cause and effect are present andother times when both are absent (Kelley, 1973),(4) consistency over time so that whenever thepresumed cause is present the effect occurs (Kel-ley, 1967), and (5) distinctiveness whereby aneffect uniquely occurs while the presumed causeis present and does not occur when the cause isabsent (Kelley, 1967). Because any one cue byitself can be misleading, clients are advised touse multiple cues in combination (Einhorn & Ho-garth, 1986).

After a preliminary causal hypothesis has beenproposed, its logical implications and plausible

81

James C. Overholser

rival hypotheses need to be examined. Hypothe-sis testing involves using hypothetical or actualenvironmental changes to modify one conditionat a time. Subsequent changes in outcome areobserved. Manipulation of environmental factorscan be done both logically and empirically, in aprospective or retrospective (historical) manner.A hypothesis is tested using deductive logic(Kruglanski, 1988) to examine its ability to suc-cessfully predict future events (Giere, 1983).Thus, if circumstances can be observed or manip-ulated so the presumed cause is present, the ef-fect should occur (Skyrms, 1986). If a particulareffect occurs in two very different situations, anyelement common to the two situations is a proba-ble cause. Alternatively, if a problem occurs inonly one of two very similar situations, factorspresent in the one situation are likely to be thecause. Systematic variation of the presumedcauses can eliminate rival hypotheses (Cohen,1970). As rival hypotheses are manipulated andexamined, the most important cause often can beidentified. The importance of a cause is reducedif other plausible causes are also present (Kelley,1973). Finally, therapist and client evaluate thestrength and diversity of evidence supporting thecausal hypothesis.

Limitations of Causal Reasoning

Causal reasoning is a difficult process for manyclients. Perception of possible causes can be in-fluenced by the client's prior expectations (Alloy& Tabachnik, 1984) and can be biased by theimproper weighting of one variable in respect tothe others. Thus, some clients overemphasize therole they play in causing their problems, whileothers minimize their role. Also, clients may ne-glect a common cause, such as when two effects(e.g., poor school grades and loss of appetite)are both caused by a third variable (e.g., recenttrauma). In most situations, numerous factors in-teract to produce an effect (Mackie, 1965). How-ever, clients often look at their problems from anarrow perspective, failing to examine all rele-vant causes. For example, an adult male clientsaw his chronic depression as related to his long-standing social difficulties, but biological, devel-opmental, and cognitive variables were also rele-vant. Examining his problems from multiple per-spectives helped him gain a better understandingof himself and his depression.

Many cause-effect relationships occur on anirregular basis. For example, a client complained

of headaches several times each week, with noclear pattern readily observable. Therapist andclient examined whether the persistent headacheswere due to job stress, dietary factors, maritalconflict, or other factors. Furthermore, evenwhen a pattern was observed wherein job stresswas associated with headaches, not every episodeof the cause (job stress) resulted in the effect (aheadache). However, it was possible to manipu-late one condition and see how the problemchanged. When his diet was deliberatelychanged, it had no effect on his headaches. Later,when the client took an extended vacation, themarital problems continued but the job stresswas temporarily eliminated and his headachessubsided.

People often ignore sample size and base ratedata (Schustack & Sternberg, 1981). Adequateand representative sampling are needed to ensureaccurate causal hypotheses (Crocker, 1981). Ex-amining possible alternative explanations canavoid the premature acceptance of a faulty expla-nation (Gambrill, 1990). Confusing cause and ef-fect involves perceiving a consequence to be acausal variable (Damer, 1987). For example, anovice therapist assumed that marital discordcaused the client's excessive alcohol consump-tion. However, by observing the problems overtime, it was determined that the alcohol abuseserved to trigger most of the domestic arguments.

Cognitive biases affect causal reasoning. Theclient's beliefs about cause and effect will influ-ence what information is observed and empha-sized (Crocker, 1981; Kelley, 1973). Beliefs andexpectations can aid the inference process orcause it to go astray (Jaspars, 1988). The thera-pist can provide an independent and impartialview of the behavior when searching for possiblecauses (Groeben, 1990). It is important to ques-tion a client's premises early so clients do notcontinue with a biased line of reasoning andreach a conclusion that appears to be true but isbased on faulty assumptions (Crossley & Wilson,1979). For example, an adult female client withagoraphobia longed for a close relationship andmarriage. She reported that her problems alwayssubsided when she was involved with a man.When discussing relationships, she made vaguestatements about how her life and her agorapho-bia would be improved if she got married. Thetherapist asked a series of questions to force theclient to address her unspoken ideas: "Howwould your life be different if you were mar-

82

Inductive Reasoning

ried?", "How would your life stay the same aftermarriage?", "So what would marriage accom-plish?". Over sessions, the client was able to seethat her progress depended on her effort and atti-tude, and was largely independent of any roman-tic relationships occurring over the course oftherapy.

ConclusionsInductive reasoning is compatible with most

forms of cognitive psychotherapy. It ensures thecollaborative empiricism necessary for cognitivetherapy (Beck et al., 1979), it retains the logicalaspect of Ellis' (1962) rational-emotive therapy,and it retains the self-focused exploration of cli-ent-centered therapy (Rogers, 1961) and self-control therapies (Mahoney, 1974). Because theSocratic method helps clients examine andchange their conceptualizations of different prob-lems, the Socratic method can be incorporated asan element of many different forms of psycho-therapy. Thus, the Socratic method can providean organizational framework for some commonelements in psychotherapy. Many therapists useprinciples of inductive reasoning, but the So-cratic method makes these principles more fo-cused and explicit. The current description is acontemporary adaptation of the inductive ap-proach used by Socrates. As a rational and scien-tific approach, it emphasizes cognitive processesin psychotherapy. The Socratic method may bebest integrated with other approaches that empha-size the value of affect and the importance of thetherapeutic relationship.

The process of using inductive reasoning relieson a cooperative inquiry (Klein, 1986) betweentherapist and client. The Socratic method at-tempts to have the therapist facilitate the client'sself-discovery (Overholser, 1988), thereby pro-moting the client's autonomy (Overholser,1987). The Socratic method helps clients learnhow to seek their own solutions (Seeskin, 1987).Explicit instruction is not always an effectiveway of teaching conceptual skills (Claiborn &Dixon, 1982). Learning by discovery producesbetter understanding (Legrenzi, 1971) andstrengthens the client's ability to solve novelproblems (McDaniel & Schlager, 1990).

Fallacies of logical reasoning can be identifiedand curtailed. It is natural for clients to processinformation selectively, seeking information thatis consistent with their established beliefs (Evans,1989). However, if left unchecked, these falla-

cies allow clients to retain invalid hypotheses anddistorted views of their world. Rather than tryingto increase the objectivity of clients, it may bemore appropriate to help clients see their prob-lems from different and more adaptive perspec-tives (Efran et al., 1990). Inductive reasoningcan be used to help clients distance themselvesfrom the emotional aspects of their problems andexpand their conceptualization of their problems.

ReferencesALLOY, L. & TABACHNIK, N. (1984). Assessment of covaria-

tion by humans and animals: The joint influence of priorexpectations and current situational information. Psycho-logical Review, 91, 112-149.

ARNOULT, L. & ANDERSON, C. (1988). Identifying and reduc-ing causal reasoning biases in clinical practice. In D. Turkand P. Salovey (Eds.), Reasoning, inference, and judgmentin clinical psychology, (pp. 209-232). New York: FreePress.

BECK, A. T. & EMERY, G. (1985). Anxiety disorders andphobias: A cognitive perspective. New York: Basic.

BECK, A. T., RUSH, A. J., SHAW, B. & EMERY, G. (1979).Cognitive therapy of depression. New York: Guilford.

BLACK, M. (1952). Critical thinking. Englewood Cliffs, NJ:Prentice-Hall.

BRANSFORD, J., FRANKS, J., VIE, N. & SHERWOOD, R.(1989). New approaches to instruction: Because wisdomcan't be told. In S. Vosniadou & Ortony (Eds.), Similarityand analogical reasoning (pp. 470-497). New York: Cam-bridge University Press.

BROWN, A. (1989). Analogical learning and transfer: Whatdevelops? In S. Vosniadou and Ortony (Eds.), Similarityand analogical reasoning (pp. 367-412). New York: Cam-bridge University Press.

BURKS, A. (1980). Enumerative induction versus eliminativeinduction. In L. Cohen and M. Hesse (Eds.), Applicationsof inductive logic (pp. 172-189). New York: Oxford Uni-versity Press.

CARNAP, R. (1945). On inductive logic. Philosophy of Sci-ence, 12, 72-97.

CARNAP, R. (1966/1987). The confirmation of laws and theo-ries. Reprinted in J. Kourany (Ed.), Scientific knowledge:Basic issues in the philosophy of science (pp. 122-138).Belmont, CA: Wadsworth.

CHESSICK, R. (1982). Socrates: First psychotherapist. Ameri-can Journal of Psychoanalysis, 42, 71-83.

CLAIBORN, C. & DIXON, D. (1982). The acquisition of con-ceptual skills: An exploratory study. Counselor Educationand Supervision, 21, 274-281.

COHEN, L. (1970). The implications of induction. London:Methuen & Co.

CROCKER, J. (1981). Judgment of covariation by social per-ceivers. Psychological Bulletin, 90, 272-292.

CROSSLEY, D. & WILSON, P. (1979). How to argue: An intro-duction to logical thinking. New York: Random House.

DAMER, T. E. (1987). Attacking faulty reasoning, 2nd ed.Belmont, CA: Wadsworth.

DOWNING, C , STERNBERG, R. & Ross, B. (1985). Multi-causal inference: Evaluation of evidence in causally com-plex situations. Journal of Experimental Psychology: Gen-eral, 114, 239-263.

83

James C. Overholser

Era AN, J., LUKENS, M. & LUKENS, R. (1990). Language,structure, and change: Frameworks of meaning in psycho-therapy. New York: Norton.

EINHORN, H. & HOGARTH, R. (1986). Judging probablecause. Psychological Bulletin, 99, 3-19.

ELLIS, A. (1962). Reason and emotion in psychotherapy. NewYork: Citadel Press.

ELLIS, A. (1977). The basic clinical theory of rational-emo-tive therapy. In A. Ellis and R. Grieger (Eds.), Handbookof rational-emotive therapy (pp. 3-34). New York: Springer.

EVANS, J. (1989). Bias in human reasoning: Causes and con-sequences. London: Erlbaum.

FRIED, L. & HOLYOAK, K. (1984). Induction of category dis-tributions: A framework for classification learning. Journalof Experimental Psychology: Learning, Memory, and Cog-nition, 10, 234-257.

GAMBRILL, E. (1990). Critical thinking in clinical practice:Improving the accuracy of judgments and decisions aboutclients. San Francisco: Jossey-Bass.

GIBRE, R. (1983). Testing scientific theories. In J. Earman(Ed.), Minnesota studies in the philosophy of science,Vol. X (pp. 269-298). Minneapolis: University of Minne-sota Press.

GENTNER, D. (1989). The mechanisms of analogical learning.In S. Vosniadou and Ortony (Eds.), Similarity and analogi-cal reasoning (pp. 199-241). New York: Cambridge Uni-versity Press.

GERGEN, K. (1985). The social constructionist movement inmodem psychology. American Psychologist, 40, 266-275.

GERGEN, K. (1988). Knowledge and social process. In D.Bar-Tal and A. Kruglanski (Eds.), The social psychologyof knowledge (pp. 30-47). Cambridge: Cambridge Univer-sity Press.

GICK, M. & HOLYOAK, K. (1980). Analogical problem solv-ing. Cognitive Psychology, 12, 306-355.

GICK, M. & HOLYOAK, K. (1983). Schema induction andanalogical transfer. Cognitive Psychology, 15, 1-38.

GORDON, D. (1978). Therapeutic metaphors: Helping oth-ers through the looking glass. Cupertino, CA: METAPublications.

GROEBEN, N. (1990). Subjective theories and the explanationof human action. In G. Semin and K. Gergen (Eds.), Ev-eryday understanding: Social and scientific implications(pp. 19-44). London: Sage.

HARNAD, S. (1987). Category induction and representation.In S. Harnad (Ed.), Categorical perception: The ground-work of cognition (pp. 535-565). New York: CambridgeUniversity Press.

HASTIE, R. (1984). Causes and effects of causal attribution.Journal of Personality and Social Psychology, 46, 44-56.

HILTON, D. (1990). Conversational processes and causal ex-planation. Psychological Bulletin, 107, 65-81.

HOLLAND, J., HOLYOAK, K., NISBETT, R. & THAGARD, P.(1986). Induction: Processes of inference, learning, anddiscovery. Cambridge, MA: MIT Press.

HOLYOAK, K. & KOH, K. (1987). Surface and structural simi-larity in analogical transfer. Memory & Cognition, 15,332-340.

HOLYOAK, K. & THAGARD, P. (1989). A computationalmodel of analogical problem solving. In S. Vosniadou andOrtony (Eds.), Similarity and analogical reasoning (pp.242-266). New York: Cambridge University Press.

HUME, D. (1748/1927). An enquiry concerning human under-standing. Reprinted in C. Hendel (Ed.), Hume selections(pp. 107-193). New York: Scribner.

JASPARS, J. (1988). Mental models of causal'reasoning. In D.

Bar-Tal and A. Kruglanski (Eds.), The social psychologyof knowledge (pp. 335-358). Cambridge: Cambridge Uni-versity Press.

JOHNSON-LAIRD, P. (1989). Analogy and the exercise of cre-ativity. In S. Vosniadou and Ortony (Eds.), Similarity andanalogical reasoning (pp. 313-331). New York: Cam-bridge University Press.

KEENEY, B. (1987). The construction of therapeutic realities.Psychotherapy, 24, 469-476.

KELLEY, H. H. (1967). Attribution theory in social psychol-ogy. In D. Levine (Ed.), Nebraska symposium on motiva-tion, Vol. 15 (pp. 192-240). Lincoln: University of Ne-braska Press.

KELLEY, H. H. (1973). The processes of causal attribution.American Psychologist, 28, 107-128.

KRUGLANSKI, A. (1988). Knowledge as a social psychologicalconstruct. In D. Bar-Tal and A. Kruglanski (Eds.), Thesocial psychology ofknowledge (pp. 109-141). Cambridge:Cambridge University Press.

LAKATOS, I. (1970/1987). Falsification and the methodologyof scientific research programmes. Reprinted in J. Kourany(Ed.), Scientific knowledge: Basic issues in the philosophyof science (pp. 170-196). Belmont, CA: Wadsworth.

LEGRENZI, P. (1971). Discovery as a means to understanding.Quarterly Journal of Experimental Psychology, 23, 417-422.

LINGLE, J., ALTOM, M. & MEDIN, D. (1984). Of cabbagesand kings: Assessing the extendibility of natural object con-cept models to social things. In R. Wyer and T. Srull(Eds.), Handbook of social cognition. Vol. 1 (pp. 71-117).HiUsdale, NJ: Erlbaum.

LINVILLE, P. (1985). Self-complexity and affective extremity:Don't put all of your eggs in one cognitive basket. SocialCognition, 3, 94-120.

LINVILLE, P. (1987). Self-complexity as a cognitive bufferagainst stress-related illness and depression. Journal ofPersonality and Social Psychology, 52, 663-676.

MACKIE, J. L. (1965). Causes and conditions. American Phil-osophical Quarterly, 2, 245-264.

MAHONEY, M. J. (1974). Cognition and behavior modifica-tion. Cambridge, MA: Ballinger.

MANICAS, P. & KRUGER, A. (1976). Logic: The essentials.New York: McGraw-Hill.

MARTIN, J., CUMMINGS, A. & HALLBERG, E. (1992). Thera-pists' intentional use of metaphor: Memorability, clinicalimpact, and possible epistemic/motivational functions.Journal of Consulting and Clinical Psychology, 60, 143-145.

MCCLOSKEY, M. & GLUCKSBERG, S. (1978). Natural catego-ries: Well defined or fuzzy sets. Memory & Cognition, 6,462-472.

MCDANIEL, M. & SCHLAGER, M. (1990). Discovery learningand transfer of problem-solving skills. Cognition and In-struction, 7, 129-159.

MEDIN, D. & BARSALOU, L. (1987). Categorization processesand categorical perception. In S. Harnad (Ed.), Categoricalperception: The groundwork of cognition (pp. 455-490).New York: Cambridge University Press.

MEDIN, D. & SMITH, E. (1984). Concepts and concept forma-tion. Annual Review of Psychology, 35, 113-138.

Mm., J. S. (1890). A system of logic, ratiocinative and induc-tive: Being a connected view of the principles of evidenceand the methods of scientific investigation. Eighth Ed. NewYork: Harper & Brothers.

NAVIA, L. (1985). Socrates: The man and his philosophy.Lanham, MD: University Press of America.

84

Inductive Reasoning

NELSON, L. (1980). The Socratic method. Thinking, 2, 34-38.OVERHOLSER, J. C. (1987). Facilitating autonomy in passive-

dependent persons: An integrative model. Journal of Con-temporary Psychotherapy, 17, 250-269.

OVERHOLSER, J. C. (1988). Clinical utility of the Socraticmethod. In C. Stout (Ed.), Annals of clinical research(pp. 1-7). Des Plaines, IL: Forest Institute.

OVERHOLSER, J. C. (1991). The Socratic method as a tech-nique in psychotherapy supervision. Professional Psychol-ogy: Research and Practice, 22, 68-74.

OVERHOLSER, J. C. (1993). Elements of the Socratic method:I. Systematic questioning. Psychotherapy, 30, 67-74.

PELLEGRINO, J. (1985). Inductive reasoning ability. In R. J.Sternberg (Ed.), Human abilities: An information-pro-cessing approach (pp. 195-225). New York: Freeman.

POPPER, K. (1965). Conjectures and refutations: The growthof scientific knowledge. New York: Basic.

POPPER, K. (1968). The logic of scientific discovery, 2nd ed.New York: Harper & Row.

REED, S. (1972). Pattern recognition and categorization. Cog-nitive Psychology, 3, 382-407.

RESCHER, N. (1980). Induction: An essay on the justificationof inductive reasoning. Pittsburgh: University of Pitts-burgh Press.

ROGERS, C. (1961). On becoming a person: A therapist'sview of psychotherapy. Boston: Houghton Mifflin.

RYCHLAK, J. (1968). A philosophy of science for personalitytheory. Boston: Houghton Mifflin.

SALMON, W. (1973). Logic, 2nd Ed. Englewood Cliffs, NJ:Prentice-Hall.

SCHUSTACK, M. & STERNBERG, R. (1981). Evaluation of evi-dence in causal inference. Journal of Experimental Psychol-ogy: General, 110, 101-120.

SEESKIN, K. (1987). Dialogue and discovery: A study in So-cratic method. Albany: SUNY Press.

SEIPLE, G. (1985). The Socratic method of inquiry. Dialogue,28, 16-22.

SIMCO, N. & JAMES, G. (1976). Elementary logic. Encino,CA: Dickenson.

SKYRMS, B. (1986). Choice and chance: An introduction toinductive logic, 3rd ed. Belmont, CA: Wadsworth.

SNYDER, M. & SWANN, W. (1978). Hypothesis-testing pro-cesses in social interaction. Journal of Personality and So-cial Psychology, 36, 1202-1212.

SPIRO, R., FELTOVICH, P., COULSON, R. & ANDERSON, D.(1989). Multiple analogies for complex concepts: Antidotesfor analogy-induced misconception in advanced knowledgeacquisition. In S. Vosniadou and Ortony (Eds.), Similarityand analogical reasoning (pp. 498-531). New York: Cam-bridge University Press.

STEIN, H. (1991). Adler and Socrates: Similarities and differ-ences. Individual Psychology, 47, 241-246.

STROHMER, D., MOILANEN, D. & BARRY, L. (1988). Personalhypothesis testing: The role of consistency and self-schema.Journal of Counseling Psychology, 35, 56-65.

STROHMER, D., & NEWMAN, L. (1983). Counselor hypothesis-testing strategies. Journal of Counseling Psychology, 30,557-565.

SWEENEY, P., ANDERSON, K. & BAILEY, S. (1986). Attribu-tional style in depression: A meta-analytic review. Journalof Personality and Social Psychology, 50, 974-991.

VOSNIADOU, S. & ORTONY, A. (1989). Similarity and analogi-cal reasoning: A synthesis. In S. Vosniadou and Ortony(Eds.), Similarity and analogical reasoning (pp. 1-17).New York: Cambridge University Press.

WEMER, B. (1985). "Spontaneous" causal thinking. Psycho-logical Bulletin, 97, 74-84.

WERKMEISTER, W. (1957). An introduction to critical think-ing, 2nd ed. Lincoln, NB: Johnsen.

ZECHMEISTER, E. & JOHNSON, J. (1992). Critical thinking: Afunctional approach. Pacific Grove, CA: Brooks/Cole.

ZIMBARDO, P. (1977). Shyness. Menlo Park, CA: AddisonWesley.

85