Metatheory and Metamethodology in Marketing: A Lakatosian Reconstruction
Author(s): Siew Meng Leong
Reviewed work(s): Source: Journal of Marketing, Vol. 49, No. 4 (Autumn, 1985), pp. 23-40 Published by: American Marketing Association Stable URL: http://www.jstor.org/stable/1251429 . Accessed: 26/07/2012 19:21 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. . American Marketing Association is collaborating with JSTOR to digitize, preserve and extend access to Journal of Marketing. http://www.jstor.org Siew Meng Leong Metatheory and Metamethod ology in Marketing: A Lakatosian Reconstruction The Lakatosian sophisticated method ological falsification frameworkis introd uced , compared with other philosophies of science, and applied to reconstruct marketing science. Fromthisthe location and intensity of extanttheoreticalconflictsare pinpointed and explained , marketing's scientificstatusevaluated , and implications forresearch method ology in the d iscipline d iscussed . W HEN Converse (1945), A ld erson and Cox (1948), and Bartels (1951) fired the opening salvos concerning the scientificstatusof marketing, little d id they realize thatthe resulting d ebate would rage foranotherfourd ecad es. Muchof the contro- versy hasfocused on the appropriate subject matter and bread thforthe marketing d iscipline (c.f. A rd t 1981). However, this d ialog has recently been over- shad owed by an even more fund amental d ispute over the metatheoretical und erpinnings of marketing sci- ence. Resolution of thisd ebate isessentialforfuture progress in marketing because the ad option of a par- ticular philosophicalperspective affects what facts, theories, and method ologies will subsequently be ac- ceptable to the d iscipline. Some scholars, notably Hunt (1983a, 1984), have urged the perpetuation of logicalempiricism, while otherssuchas Zaltman, LeMasters, and Heffring Siew Meng Leong isad octoral cand id ate, University of Wisconsin-Mad - ison, on leave fromthe Schoolof Management, National University of Singapore. The author wishesto thankPaul Busch, Gilbert A . Churchill, Jr., Deborah Roed d er John, George John, and the JM reviewers and ed itorfor helpful comments. Special thanks are extend ed to J. PaulPe- terand Ronald Serlin fortheircontributions ateach stage of the writing of thisarticle. Journalof Marketing Vol. 49 (Fall1985), 23-40. (1982), Peterand Olson (1983), and A nd erson (1983) have argued persuasively forthe ad option of arela- tivistic perspective. Logicalempiricism assumesthat science is objective and emphasizesrigorous mea- surementand hypothesistesting. In contrast, relativ- ismtend sto emphasize the subjective and creative as- pects of science. The time seems ripe forsome form of a Hegelian d ialectical synthesis between the ap- parentlypolar extremes represented by the logical em- piricist thesisand the relativistantithesis respectively. Id eally, thisnew perspective should incorporate the bestthatcan be offered by boththese philosophies while minimizing theird rawbacks. One potential cand id ate for accomplishing thistask is sophisticated method ological falsification (SMF) proposed by the late Imre Lakatos (1978) in his at- tempt to resolve asimilar controversy between phi- losophers ThomasKuhn and Karl Popper. Kuhn (1970) had proposed amod elof scientific progress found ed on the notion of a parad igm. Science is viewed by Kuhn to progress in a cycle commencing fromone parad igmaccepted throughout a particular scientific community and researched extensivelyd uring a pe- riod of "normalscience." A sanomalies accumulate, acrisis stage mayresult, followed by the fallof the old parad igm and the emergence of anotherin a"sci- entificrevolution." Metatheory and Metamethod ology in Marketing / 23 While parad igms d ealwiththe proper d omain of a science, what research questions are to be ad - d ressed , and the rulesto follow in the interpretation of results, severalcriticismshave been d irected at Kuhn'sthesis. Masterman (1970), among others, has noted thatKuhn usesthe term parad igm in a variety of d ifferent ways, whichlead sto varied interpreta- tions. For example, A rd t (1983) classified the polit- ical economy mod elasa parad igm, butin 1985 con- sid ersita metaphor und erthe sociopoliticalparad igm. A nother point of d ifference iswhether marketing sci- ence hasbeen guid ed by a single parad igm. Most scholars (A rd t 1985, Carman 1980, O'Shaughnessy and Ryan 1979, Roberts 1984) have suggested that marketing hasno central exemplarparad igm. In the Kuhnian sense, the d iscipline isin a preparad igmatic stage. If thisis the case, then the sociopolitical(or anyother) "parad igm" cannotbe labeled assuch. Most importantly, Popper(1970) has questioned the Kuhnian representation of scientific enterprise as asuccession of bold parad igms and theird ramatic overthrows. Popper hold sthatscientific progressmay be viewed asthe proliferation of rival theories, that is, of revolution in permanence. In resolving thiscon- troversy, Lakatos (1978) introd uced the notion of a research program. Instead of the multiple d efinitions associated withthe term parad igm, the Lakatosian no- tion is unambiguous with respect to its constituent characteristics. The Lakatosian (SMF) perspective urges the existence and d esirability of multiple theoretical found ationsin a d iscipline, aformulation consistent with Popperian and Feyerabend ian (1980) tenetsand the contemporary structure of marketing science. Within eachresearch program, the Lakatosian perspective re- tainsthe Kuhnian notion of solving specificproblems (puzzles) withassociated (mid d le-range) theories. Moreover, as willbe later d etailed , the Lakatosian perspective blend s together bothtrad itionaltenetsof empiricism(i.e., falsification) aswellasmore con- temporary relativistnotions (i.e., contextin research in the formof research programs). Italso recognizes the myopic nature of d ogmatic falsification that ig- noresthe tenacity withwhichscientifictheoriesare held in the face of seeminglyd isconfirming evid ence and rejects Kuhn'snotion of a single d ominant par- ad igm in a d iscipline. Thus, the firstand primarypurpose of thisarticle is to reconstruct marketing science fromthe sophis- ticated method ological falsification perspective. Through this process itis hoped thatsome of the ap- parentd isarray within the d iscipline can be system- atically accounted forand marketing's scientificstatus evaluated . A second aryobjective isto extend the Lak- atosian frameworkto the analysis of researchmeth- od ology in marketing. Extant philosophicalperspec- tives suggest a range of method ologicald irection, from the use of a specificapproach suchasstructural equa- tions mod eling, to method ologicalanarchy. The Lak- atosian perspective ad vocates method ologicalplural- ism, the need to more d eeply examine researchmethod s in marketing, and the ad option of researchmethod s, bothstatisticaland qualitative, that (1) expose theo- riesto severe threatsof refutation, and (2) aid in the d iscovery of new find ings in the d iscipline. The remaind erof thisarticle is organized asfol- lows:The nextsection will provid e ahistorical per- spective on the d evelopment of falsification. In par- ticular, the three major versionsof falsification will be d etailed . Withthatas background , a comparative analysis of various approaches in the philosophy of science willbe presented , withaview toward sad - vocating the Lakatosian perspective asa possible means of reconciling d ifferencesbetween empiricism and relativism. Marketing science willthen be recon- structed fromaLakatosian perspective to id entify and explain the location and intensity of extanttheoretical conflictsin the d iscipline and to evaluate itsscientific status. The second part of this article willad d ress method ological issuesfromaLakatosian perspective. The need for method ologicalpluralism is d iscussed , followed by an examination of researchmethod s- bothstatisticaland qualitative-from the Lakatosian perspective. Limitationsof the Lakatosian perspective are then presented , followed by some conclud ing comments. Falsification Itwas Popper(1959, 1962) who first d eveloped fal- sification asan alternative method of theoryjustifi- cation aimed at overcoming severald ifficultiesasso- ciated with logicalempiricism. Falsificationists argue thatscientific practice cannotbe rationally d efend ed if it proceed sind uctively, and thatallscientificthe- oriesare hence equallyunprovable and improbable (Serlin and Lapsley1985). The three major versions of falsification are d ogmaticfalsification, naive meth- od ologicalfalsification, and sophisticated method o- logical falsification. Dogmatic Falsification Dogmatic falsification assertsthatonce a theory is d isproved by the d iscovery of a single refuting in- stance, itshould be eliminated fromthe bod y of sci- entifictheories (Serlin and Lapsley1985). While all theoriesare d eemed fallible, d ogmatic falsificationists assume the existence of an infallible empirical basis. Thiscleard emarcation between factsand theories per- mitsthe unequivocalappeal to "hard facts" in eval- uating fallible theories. There are several shortcomings of the d ogmatic falsificationist perspective. First, ithasbeen posited 24 / Journal of Marketing, Fall1985 thatthere isno strict psychologicalbound aryd iffer- entiating factfrom theory(Kuhn 1970, Polanyi 1958). A s Lakatos (1978, his emphasis) asserts, ". .. there can be no sensations unimpregnated byexpectations, and therefore there isno naturald emarcation between observationaland theoretical propositions" (p. 15). Second , itis impossible to conclusively refute athe- ory because realistictestsituations d epend on much more than the theory und er investigation (Duhem1953). Lakatos (1978) argues that anyempirical test involves assumptions aboutinitial cond itions, measuring in- struments, and auxiliaryhypothesesconstituting the ceteris paribus clause. A n alleged refutation of the theory can be easily d eflected bysuggesting that something else in the ceteris paribus clause caused the result (Laud an 1977). In other word s, since alltheo- riescontain aceteris paribusclause, itisa theoryplus the ceteris paribus clause thatis subjected to empirical testing. Since itis alwayspossible to replace the cet- eris paribusclause, anysingle testof a theory is of little consequence, thus lead ing to the uncomfortable conclusion thatalltheoriesare not onlyequally un- provable and improbable butalso equallyund isprov- able (Serlin and Lapsley1985). Naive Method ological Falsification Popper's(1962) naive method ological falsification perspective attempts to rescue science from skepti- cism byd emonstrating thatscience isnot only acor- pus of assertionsbutalso a system of conventions. The empiricaltesting process is recognized to be im- possible without making aseriesof method ological d ecisions. Since no pure observations exist, whatare to be regard ed asfactsmustbe conventionallyagreed upon in light of a". . . 'relevant technique' suchthat 'anyone who haslearned it' willbe able to d ecid e that the statementis 'acceptable'" (Lakatos1978, p. 22). Hence, potential theoreticalfalsifiersare granted ob- servationalstatus by d ecision. The truth-value of such observationsisarrived at by arelevantresearchmethod . Popperian falsificationists recognize that research method sand scientifictheoriesare fallible, but, by d ecision, assume that they constitute unproblematic background knowled ge subsumed by the ceteris par- ibusclause. While thiscircumventsthe problem of how factis d emarcated from theory, there isstillthe problem of how, given the ceteris paribusclause, a specific the- ory is subjected to refutation. Popper maintainsthat thisis achieved bymaking another method ological d ecision. The researcherd ecid es before an experi- mentiscond ucted whatstate of affairsisd eemed ac- ceptable asafalsification of the theory und er test, ir- respective of ceteris paribus. Thus, the Popperian perspective recognizesthattheoriescannotbe sub- jected clearly to empiricaltesting because of ceteris paribus and auxiliary theories. However, to permit "objective" assessment, aresearchermust stipulate in ad vance whatevents would constitute falsification withoutad hoc appeals to ceteris paribus. Thus, the Popperian perspective hand lesthe auxiliary-theory is- sue in theorytesting bymaking a public, conventional d ecision to consid era theory falsified given the ob- servation of specified outcomes. Even so, the Popperian reconstruction of science has rarely been in accord withthe actual history of scientific progress. Lakatos (1978) points outthat many scientifictheorieshave ad vanced d espite apparent ref- utations byempiricald ata, thus testifying to the te- nacity withwhichtheoriesare held in the face of seeminglyd isconfirming evid ence. Sophisticated Method ological Falsification The sophisticated method ological version of falsifi- cation appears to offeramore powerful reconstruction of science. Sophisticated method ological falsification (SMF) hold sthatwhile science cannot prove theories orestablishtheir probabilityusing probability calcu- lus, it can, withthe mod us tollens,1 d isprove them (Serlin and Lapsley1985). However, in the SMF framework, mere criticismvia refuting evid ence is neversufficientfor theoryfalsification, since one never evaluatesa single theory butaseriesof theorieswithin research programs. Thisview of science argues thataresearch pro- gram containsa"hard core" of fund amental assump- tionsand theoretical propositionsaccepted asincon- trovertible by scientistswithin the research program. Scientistsare posited to insulate thishard core from refutation witha "protective belt" of auxiliary theo- ries. This "negative heuristic" stemsfromameth- od ological d ecision to cord on off the core of theirre- search program fromthe threatof refutation. That is, the mod ustollensis forbid d en to be d irected atthe hard core but, instead , itisthe auxiliary theoriesthat are subjected to rigoroustesting. In contrastto the negative heuristicwhichtellssci- entistswhat path not to follow, research programs contain a "positive heuristic" which guid es the d irec- tion of research. The positive heuristicis composed of asetof guid ing research questions and specifies the formtheorieswillassume and what they willlook like. The positive heuristic proceed s in the face of counterevid ence and refutation, and there isno need to consid erthe presence of empirical anomaliesas being d ecisive. Thus, refutationsare not ignored butare consid ered inconclusive untilthe positive heuristicis able to confrontthe d isconfirming evid ence. The SMF 'The mod us tollens is the logical formof Popper's (1959) falsifi- cation criterion which permits the d ed uctive testing of theories. It may be represented asfollows: A implies B; not B; therefore, notA . Metatheory and Metamethod ology in Marketing / 25 perspective thusaccountsforthe relative autonomy and tenacity of scientificresearch programs. A band onmentof aresearch program is onlypos- sible when (1) there existsarivalresearch program powerfulenough to accountforallthe "facts" of the former program, (2) thatrivalresearch program offers novel, excess empirical contentoverthe former pro- gram, some of whichshould be corroborated , and (3) the former program isno longerprogressive (i.e., is d egenerative) in thatits positive heuristicisno longer able to generate novel find ings. Letusillustrate the d istinction between the criti- cismof aresearch program and itsaband onmentwith an example. Recently, concern hasbeen raised on the neglect of mood and emotion in cognition (Taylor1980) and the almost ubiquitousregard of affectas being postcognitive by information processing theorists (Za- jonc1980, 1984; Zajonc and Markus 1982). Similar sentimentshave been echoed by consumerresearchers who question the assumption that purchase behavior is necessarilypreced ed by choice and d ecision pro- cesses (Kassarjian 1978, Olshavsky and Granbois 1979, Sheth 1982). More recently, however, there have been growing ind icationsthatthe positive heuristicof the informa- tion processing research programappearsread y to confrontthese issuesand problems. Lazarus (1984) has proposed that "cognitive formulationsshould specify how various personalagend as . . . shape cognitive appraisal, and , in so d oing, affectthe pro- pensity to experience certain emotionsin particular environmentalcontexts" (p. 129). In this connection, Bower (1981) has proposed an associative network theory in whichan emotion servesasa memory unit thatcan enterinto associationswithcoincid entevents. The activation of thisemotion unitis posited to aid retrievalof eventsassociated withitand also primes emotionalthemataforuse in free association, fanta- sies, and perceptualcategorization. Similarly, Fiske (1982) argues forthe implication of schematic pro- cesses in interpersonal affective responses. In her mod el, affectisassumed to be stored withthe generic knowled ge structure and isthusavailable immed iately upon categorization. Hence, evaluationsand affectare cued byfitting an instance to aschema. Otherresearchershave investigated how affect may influence d ecision-making strategies and risk-taking tend encies (Isen etal. 1982). In consumer research, Holbrookand Hirschman (1982) ad vocate thatthe in- formation processing approach be enriched by an ex- perientialperspective whichconsid ersconsumerfan- tasies, feelings, and fun. Moreover, Petty and Cacioppo (1981) have proposed an Elaboration Likelihood Mod el of attitud e change incorporating the trad itionalinfor- mation processing assumption of d iligent consid era- tion of information (centralroute) as wellas less thoughtful meansof persuasion via simple cue asso- ciation (peripheralroute). A sa consequence of these d evelopments, few re- searcherswould ad vocate the complete aband onment of the information processing research program. For, even if anew research programemerges thatcan ac- countforallits find ings, the positive heuristicof the information processing research programappears at present to be able to generate novel find ings. Comparative A nalysis of the Various A pproaches Itisnow appropriate to compare the SMF framework withalternative approaches in the philosophy of sci- ence. Suchan analysis would provid e a properper- spective forlaterd iscussion of the implications of the SMF frameworkfor marketing and its application to the reconstruction of marketing science. To facilitate this comparison, Table 1 summarizessome of the conflicting views of positivists/empiricists(column 1) and relativists/constructionists(column 4) as per- ceived by Peterand Olson (1983). Column 2 d epicts aview of science called mod er empiricismby Hunt (1984), while the SMF frameworkisfeatured in Col- umn 3.2 The layout of Table 1 is notaccid ental. A t the extremesare the trad itional positivistic view and the more contemporaryrelativistic/constructionist per- spective. In between are the more mid d le-of-the-road positions ad vocated by mod er empiricism and so- phisticated method ological falsification. The location of the SMF perspective along thiscontinuumis par- ticularlypertinent to the extentthatit is a hybrid blend ing together severald ifferent trad itions, viz. em- piricism, conventionalism, and the Kantian activitist approach to the theory of knowled ge. Ind eed , the SMF perspective may offera possible reconciliation between mod ern empiricism and rela- tivismin thatit assigns the objectivity and rationality of scientificresearchto alevelof abstraction d ifferent fromthatof theoreticalcommitment. A s points 1 and 4 of Table 1 show, the SMF framework d istinguishes the world of propositions and id easfromthe world of mental states, beliefs, and consequences. Logicalpos- itivismand mod er empiricismgenerally ad d ressthe former, while relativismisconcerned withthe latter levelof abstraction. 2Except forthe Lakatosian perspective, the otherclassifications in Table 1 are notconsensus ones. Nonetheless, since these classifica- tionshave been previously used bymarketing researchers, they are probably the ones most familiarto us. Moreover, they seemto be ad equate forthe presentexposition. Interested read ers may referto Suppe (1977), Brown (1977), Chalmers (1976), and Lakatos and Musgrave (1970) ford etailed comparative analyses of specificphi- losophers of science. 26 / Journal of Marketing, Fall1985 TA BLE 1 Major Differences between LogicalPositivism, Mod ern Empiricism, Sophisticated Method ological Falsification, and Relativism LogicalPositivism/ Empiricism Mod ern Empiricism (Peterand Olson 1983) (Hunt1984) 1. Science d iscovers the true nature of reality. 2. Only the logic of justification is need ed to und erstand science. 3. Science can be und erstood without consid ering cultural, social, political, and economicfactors. 4. Science is objective. 5. Scientific knowled ge isabsolute and cumulative. 6. Science is capable of d iscovering universallawsthat govern the external world . There is arealworld and , although science attempts to d iscoverthe nature of reality, the "true" nature of reality can neverbe known with certainty. Itis usefulto d istinguish between the proced ures thatscience uses to d iscoverits knowled ge- claims fromthose that science uses to accept or reject(justify) its knowled ge-claims. The acad emic d iscipline of philosophy of science historically focused on issues in justification. The proced ures that science uses to justify its knowled ge-claims should be ind epend ent of cultural, social, political, and economicfactors. A lthoughcomplete objectivity is impossible, science is more objective in justifying its knowled ge- claimsthan nonsciences, e.g. med icalscience is more objective than palmistry. Scientific knowled ge is neverabsolute. Muchof scientific knowled ge is cumulative, i.e., we really d o know more aboutthe causes of infectious d iseases tod ay than we d id 100 years ago. Science attempts to d iscover regularities among the phenomena in the realworld . Some of these regularities are stated in universalform and others are stated in probabilisticform. Sophisticated Method ological Falsification (Lakatos1978) The d irection of science is d etermined mainlyby human creative imagination and notthe universe of facts. Thus, the rationalreconstruction of scientific progress occurs in the world of id eas. The process by which research programs are established , justified , and accepted throughout a scientific community are need ed to appreciate science. Criticismof scientific theories is based on empiricalevid ence, but aband onment of a research program involves extra-experimental consid erations. The d emarcation of science from pseud oscience is objective in the world of id eas and propositions. Theoretical commitment is an issue of the world of mental states, beliefs, and consequences. Scientific knowled ge is neverabsolute. A ppraisal must involve aseries of theories; anypart of the bod y of science is replaceable but only in a "progressive" way, i.e., substitutes must successfullyanticipate novelfacts. Whethera proposition is a factora proposition in a test situation d epend s on a scientist's method ological d ecision within the context of aresearch program. Relativism/ Constructionism (Peterand Olson 1983) Science creates many realities. The process by which theories are created , justified , and d iffused through aresearch community are need ed to und erstand science. Science is asocial process and cannot be und erstood without consid ering cultural, social, political, and economicfactors. Science is subjective. Scientific knowled ge is relative to a particular context and period of time in history. Science creates id eas that are context-d epend ent, i.e., relative to aframe of reference. Metatheory and Metamethod ology in Marketing / 27 TA BLE 1 (continued ) Sophisticated LogicalPositivism/ Method ological Relativism/ Empiricism Mod ern Empiricism Falsification Constructionism (Peterand Olson 1983) (Hunt1984) (Lakatos1978) (Peterand Olson 1983) 7. Science prod uces Muchof scientific theoriesthatcome closerand closerto absolute truth. 8. Science is rational since itfollows formalrulesof logic. 9. There are scientific rulesfor d oing science valid ly(e.g., falsification). 10. Scientists subject theirtheories to potential falsification throughrigorous empiricaltesting. 11. Measurement proced ures d o not influence whatis measured . 12. Data provid e objective, ind epend ent benchmarksfor testing theories. knowled ge iscumulative. A bsolute truthis not knowable byscience. Science is rationalsince its purpose isto increase our und erstand ing of the world . Itd oes so through d eveloping theories, mod els, lawlike generalizations, and hypotheses which purport to d escribe, explain, and pred ictphenomena. There are normsfor d oing good science. For example, theories should be testable, measures should exhibit reliability and valid ity, and d atashould notbe fabricated or otherwise fraud ulently collected . Theoriesare subjected to the empiricaltesting process. A bsolute perfection in measurement proced ures is impossible. The empiricaltesting process provid esgood ground s for accepting some knowled ge-claims and rejecting others. A bsolute truth may not be achievable byscience, but scientificresearch programsmayyet in the long run lead to evermore true and fewerfalse consequences and , thus, have increasing verisimilitud e. If science aims at truth, it mustaimat maintaining consistency as an importantregulative principle. Inconsistency must be seen as a problem to be resolved . There are normsfor d oing good science. In ad d ition to Hunt'scriteria, research programs must be evaluated by theirheuristic power-how many new facts d o theyprod uce and how capable are they in explaining refutations d uring their d evelopment. A theory is "scientific" only if ithas corroborated excess empirical content overits rival, i.e., only if it lead s to the d iscovery of novel"facts." While experience still remainsan impartial arbiterin scientific (measurement) controversy, the importance of d ecisions in method ology should not be neglected . The empiricaltesting process provid es the ultimatelynecessary but neversufficient ground s forrefutation. No experiment, observation state, orwell-corroborated low-level falsifying hypothesis can lead to falsification withoutthe emergence of better theories. A historical emphasis is required . Truthis a subjective evaluation thatcannot be properly inferred outsid e of the context provid ed by the theory. Science is rationalto the d egree thatitseeks to improve ind ivid ualand societal well-being by following whatevermeans are usefulfor d oing so. There are manyways of d oing science valid ly that are appropriate in d ifferent situations. Scientists seek supportive, confirmatory evid ence in ord erto markettheir theories. Nothing can be measured without changing it. Dataare created and interpreted by scientists in terms of a variety of theories, and thus are theory-lad en. 28 / Journalof Marketing, Fall1985 In regard to the nature of scientific knowled ge (points 5 through 8 in Table 1), allbutthe logical positivistperspective ad vocate thatscientific knowl- ed ge isneverabsolute. While the sophisticated meth- od ological falsificationist agrees withthe mod emem- piricist thatscientific knowled ge is mainlycumulative (points 5 and 7), he/she is cognizant that scientific researchinvolves method ological d ecisionsset within a particularcontext, i.e., thatof aresearch program. Moreover, while the empiricistposition in column 2 simply statesthatscientific knowled ge is cumulative, the Lakatosian perspective provid es a d escription of the und erlying d ynamics of why science is progres- sive in stating thatthe appraisal of research programs involvesaseriesof theories, and thatold theoriescan only be aband oned when new ones emerge thatsuc- cessfullyanticipate new facts. The mod em empiricist and the sophisticated meth- od ological falsificationistare mostin agreement in re- gard to the role of empiricaltesting of theories (points 9 through12). While the mod em empiricist makesa rather vague claimthatthe empiricaltesting process provid es"good ground s" for acceptance of knowl- ed ge claims, the sophisticated method ological falsi- ficationistclarifiesthis position byasserting thatthe empiricaltesting proced ure is ultimatelynecessary but neversufficient ground s forrefutation (point12). Fal- sification and subsequent aband onmentis onlypos- sible withthe emergence of bettertheorieswithexcess empirical contentovertheir rivals, a point not ad - d ressed by the mod em empiricist(point10). In sum, the Lakatosian perspective mayprovid e the vehicle of sophisticated method ological falsifica- tion to the mod em empiricist and brid ge the gap be- tween relativistsand empiricists in emphasizing the rolesof contextin researchin the formof aresearch program and the importance of method ological d eci- sionsin research. Marketing Science Reconstructed Thissection is concerned withaLakatosian recon- struction of marketing science (see Figure 1). Itis proposed that marketing science be consid ered amas- terresearch programcomprising a collectivity of d is- tinctbutrelated research programsforming its pro- tective belt. Further, the Lakatosian reconstruction is extend ed here to incorporate the notion of amid d le- range theory(Merton 1957) atthe nextlevel. A mid - d le-range theory isone whichis "intermed iate to the minor working hypotheses evolved in abund ance d ur- ing the d ay-to-d ay routinesof research, and the all- inclusive speculations comprising amaster conceptual scheme" (pp. 5-6). Finally, atthe outer periphery of the proposed scheme are the working hypotheses of ind ivid ualresearcheffortsof marketers. FIGURE 1 Mod ified Lakatosian Reconstruction of Marketing Science A = hard core (guid ing research questions, generalproposi- tions/assumptions, integrated mod els, and classification schemas) B = protective belt(research programs) C = mid d le-range theories D = working hypotheses Hard Core and Protective Belt Perhaps the mostd ifficult aspect of the reconstruction isthe d elineation of the hard core of marketing sci- ence. However, we may be fortunate in thatmuchof the infrastructure necessary in d etermining the basic tenetsand propositions of marketing science hasre- cently been presented . Hence, Hunt (1983b) views marketing asthe behavioralscience seeking to explain exchange relationships between buyers and sellersand provid es fourfund amental explanand a thatcan serve asthe centraltenetsof marketing science in the re- construction: * Buyersengage in behaviord irected atconsum- mating exchanges. * Sellers engage in behaviord irected atconsum- mating exchanges. * Institutionalframeworksexist whichare d i- rected at consummating and /orfacilitating ex- changes. * The consummation and facilitation of exchange between buyers, sellers, and institutionalmech- anismsaffect society. Metatheory and Metamethod ology in Marketing / 29 Beyond a listing of the basictenetsof marketing science, the SMF framework requires thatthe prop- ositions/assumptionsgenerallyaccepted within ad is- cipline be d efined . Fern and Brown (1984) provid e initial insight forthis purpose. They reviewed 20 gen- eral marketing textsand found some 193 marketing "principles." Fromthese theyd eveloped three prop- ositions generalizable acrossind ustrialand consumer contexts. These generalpropositions centeron how buyer behavior (e.g., frequency of purchase, levelof knowled ge, and the numberof buyers in the market) affects sellerbehaviorand the institutionalframe- worksthatexistto consummate and /or facilitate ex- changes. A lthough itis beyond the scope of this ar- ticle to d evelop an exhaustive set of such general propositions, the Fern and Brown approach offers in- teresting possibilities forthe construction of market- ing theory, forits premise is thatonce afacthas achieved "textbookstatus" ittend sto become part of a d iscipline'sbod y of knowled ge. Itisnow possible to explicate the positive heuris- ticof the marketing d iscipline. Hunt (1983b, p. 13) provid es asetof guid ing research questions associated witheachof the centraltenets: * Why d o which buyerspurchase what theyd o, where, when, and how? * Why d o whichsellers prod uce, price, promote, and d istribute what theyd o, where, when, and how? * Why d o whichkind sof institutions d evelop to engage in whatkind sof functionsoractivities to consummate and /or facilitate exchanges, and when willthese institutions d evelop, where, and how? * Why d o whichkind sof behaviorsof buyers, sellers, and institutionshave whatkind sof con- sequences on society, when theyd o, where, and how? The positive heuristicalso containsa prescription forwhatformthe theories d eveloped fromthe guid ing research questions willassume and what they willlook like. Based on current theorizing, these willinclud e integrated mod elsand classification schemasthat por- tray the exchange behaviorof and between humans, institutions, and society. Eachmustbe well-ground ed in one ormore of the marketing d iscipline's richthe- oreticalbases that includ e psychology, sociology, economics, organizationaltheory, and political sci- ence. In ad d ition, classification schemasshould meet Hunt's (1983a) five criteriaof specification ad equacy, ad equacy of characteristicsused in classification, mu- tualexclusivenessand collective exhaustivenessof categories, and utility to the marketing community. The protective beltof research programs in mar- keting are those in buyerbehavior, sellerand com- petitive behavior, institutional behavior, and environ- mentalbehavior. These includ e the politicaleconomy, microeconomic, conflict resolution, generalsystems, functionalist, social exchange, behavior mod ification, and information processing research programs. A san illustration, Figure 2 outlinesthe structure of the in- formation processing research program. The d ashed linesind icate thatthere are no rigid bound ariesbe- tween the various mid d le-range theoriesand working hypotheses. Figure 2 isnotintend ed to be exhaustive or d efinitive butis presented to id entifyexamples from varioussubfield sof marketing affected by this partic- ularresearch program. Similarreconstructionscan be mad e forotherresearch programs in marketing. A tleasttwo points come to mind in analyzing Fig- ure 2. First, while the information processing research programprimarily ad d resses problems in the areasof buyer behaviorand promotionalstrategy, italso af- fectsothersubfield sin marketing. For example, the transactionscosts approach(Williamson 1975) used in the marketing channelsand strategy areahasbound ed rationality asone of itsbehavioral postulates. Thisas- sumption isfound ed on the notion thathumanshave limited information processing capacity. Second , in ad d ition to the competition between research pro- grams d iscussed below, there isalso competition be- tween mid d le-range theorieswithin aresearch pro- gram. For example, the issue concerning knowled ge representation in memory can be viewed fromacat- egorical versusaschematic perspective (Mand ler1979). Itisin the protective beltthatsome of the conflict between marketershasarisen. Yet, based on the Lak- atosian reconstruction, thisis precisely where points of d isagreement should emerge. Thus, und erthe SMF framework, the rival explanations of consumerbe- havioroffered by behaviorism (McSweeney and Bier- ley1984, Nord and Peter 1980, Peterand Nord 1982) and information processing (Bettman 1979; Lachman, Lachman, and Butterfield 1979; Sterthaland Craig 1982) is not unexpected . Ind eed , the generation of counter-instances by one conceptualization against the otherd oesnotthreaten the refutation of one orthe othernorof the hard core of marketing itself. More- over, the rivalry is not necessarilyd ysfunctional as long asboth conceptual accountsare progressive in the sense thatboth generate new find ings and no other account emerges thatis sufficientlypowerful to ex- plain allthe factsthat they d o. Mid d le-Range Theories and Working Hypotheses A tthe nextlevelof the proposed scheme are the mid - d le-range theorieswhichconstitute another major source of d ebate and controversy. The d egree of fragmen- 30 / Journalof Marketing, Fall1985 FIGURE 2 Illustrative Structure of the Information Processing Research Program Central Buyers and sellers Buyers and sellers Buyer and seller Buyersand sellers Tenets process and acton store and represent behaviorresultsfromhave limited information received information they the interaction of information processing and attend ed to. process in memoryfor innate information capabilities whichmay laterretrievaland use. processing capacities resultin the use of and learning heuristics in problem experiences. solving and d ecision making. Guid ing How d o buyers and How d o buyers and What cognitive skills Whatare the Research sellers acquire, sellers represent d o buyers and sellers information processing Questions integrate, and acton information in possess, and how are capabilities of buyers information presented memory, and how d o these d eveloped and and sellers? Whatare to them? they retrieve itfrom refined ? the heuristics used and memory? in what situations are theyapplied ? How d o these affect behavior? T Mid d le-range Signal d etection. Single vs. multistore Message learning Jud gmentalheuristics. Theories Information 'mod els of memory. theories. Information overload . integration. Categorical vs. Stage theories of skill A ttribution. Reasoned action. schematicmod els of learning. Low involvement. Cognitive response. memory Developmental representation. theories of cognitive A nalogical vs. 'processes. propositional representationsof knowled ge. 1 4, 1 1 i1 Working Buyerperception of ,Recognition vs. recall Effectsof familiarity Illusorycorrelation and Hypotheses and attention toward of prod uct information 'and expertise on the price-prod uct ad vertisements. 'and ad vertisement 'choice strategies. qualityrelationship. A d d ing vs. averaging content. Effectsof learning Vivid ness effect and of information. Schema-triggered goals on memory ,the effects of case vs. Compensatory vs. :affect. retrievaland choice statisticalinformation noncompensatory 'Information processing behaviorin purchase in ad vertisements. approaches to buyer 'of picturesand text in :situations. :A vailability-valence and sellerd ecision ad vertisements. Proced uraland 'hypothesis and the making. d eclarative knowled ge foot-in-the-d ooreffect. Effectsof context on and theireffects on Effectsof bound ed information acquisition ad aptive selling rationalityand in consumerand 'behavior. information ind ustrial buying ' impacted ness on behavior. vertical integration issues. tation and inconsistency is more intense atthis level. One reason forthis is thatthere are simply a greater numberof mid d le-range theories (both within and across research programs) compared to higher levelresearch programs. Second , much empirical researchis nec- essarily restricted in scope, examining only selected aspects of a theory(Bagozzi 1984), thus giving rise to the d iversity of topics stud ied . For example, consumer behavior may be inter- preted in terms of a cognitive response, message learning, oracombinational approach within the in- formation processing research program(Petty and Ca- cioppo 1981). In ad d ition, it may be consid ered from the behaviormod ification research program in terms of aclassical cond itioning, operant cond itioning, or vicarious learning approach(Nord and Peter 1980). In the marketing channels and strategyarea, the mid d le- range theories subsumed by the politicaleconomy re- search programinclud e the transactions costs ap- proach(Williamson 1975), the agency cost mod el (Jensen and Meckling 1976), Pfeffer and Salancik's (1978) resource d epend encyperspective, and A nd er- son's (1982) constituency-based theory of the firm. A tthe lowest levelin the proposed scheme are the Metatheory and Metamethod ology in Marketing / 31 working hypotheses. In the SMF framework, few ex- periments concerned with testing ind ivid ual working hypotheses are consid ered cruciallyimportant. In- d eed , the extentof fragmentation isatits heightfor, in ad d ition to the vastnumberof hypothesesgener- ated , there isthe problem thatfew of themare so well- articulated that they issue froma single mid d le-range theory. The latterthus gives rise to possible multiple interpretations of the same d ata. For example, both self-perception and cognitive d issonance theory have been proposed to explain whypeople change theirat- titud es (Petty and Cacioppo 1981, pp. 170-171). On the issue of pricing through ad istribution channel, al- ternative accountshave been provid ed concerning the function of quantity d iscounts. These includ e the transferof inventory/prod uction costs, channelco- ord ination, and price d iscrimination (Jeuland and Shugan 1983). The problem of multiple interpreta- tionsof the same d atacan be alleviated if potential falsifiersare specified in ad vance, thus preventing re- course to alternative explanations subsumed und erthe ceteris paribus clause. However, und erthe SMF framework, itis inap- propriate to suggest thata theoryhaving "seven facts forit against three against itis notin good shape" (Meehl1978, p. 823). Und erthe SMF framework, theoriesare pitted against each other, withconfir- mationsalso provid ing outcomesof interest (Serlin and Lapsley1985). Ind eed , the history of physics is re- plete with subsequently successfulresearch programs that proceed ed in "oceansof anomalies" and d iscon- firming evid ence (Lakatos1978). It is too rashto overthrow a theory because of arecalcitrant fact, and few researcherswould d o so in any of the mature sci- encesorin marketing. Und erthe SMF framework, the rational proced ure is to examine empiricald iscrepanciesbythorough testing of the ceteris paribus clause and applying the positive heuristicto uncovernew facts. Given this, the variety of mid d le-range theoriesd erived from higher levelresearch programs in the "protective belt" of the marketing d iscipline should be in aconstantstate of contemporary and historicalevaluation (Savitt1980). None can be refuted withoutthe emergence of a supersed ing theorysatisfying the criteriaof offering novel, excess information, some of whichshould be corroborated . Hence, theoryproliferation is encouraged und er the Lakatosian framework, as competition between research programs and mid d le-range theoriesis es- sentialforscientific progress. On this point, Lakatos isin agreement with Popper(1970) and Feyerabend (1980) in arguing thatKuhn (1970) wasincorrectto suggest thatthe benefitsof theoretical monopoly ex- ceed ed thatof theoretical pluralism. However, the SMF frameworkd oesnotand isnotmeantto provid e sanc- tuary forthose theoriesthatd o notd erive fromwithin the d iscipline'spositive heuristic. Further, it grants no philosophical license forthe retention of those theo- riesthat prod uce d egenerative program shiftsin the long run, particularly withthe emergence of superior explanations. Two implications for marketing science emerge. First, itcan be observed that many of the mid d le-range theorieshave been borrowed (e.g., in Figure 2, from psychology). This points to the need for build ing more ind igenoustheoryguid ed by the d iscipline'spositive heuristicand issuing fromatleastone of itsresearch programs. This may take the formof synthesizing or extend ing extant mid d le-range theoriesthat may be borrowed fromrelated d isciplines. Second , care should be exercised in borrowing theoriesfromotherd isci- plines. The six principles of borrowing provid ed by Robertson and Ward (1973) may be usefulin thisre- gard . These principles callforthe need to (1) justify whatis borrowed , (2) examine the relevance of bor- rowed theoriesfor explaining marketing behavior, (3) examine the legitimacy of the borrowed theory in terms of supportive evid ence in its originald iscipline, (4) examine the contextualfitof the borrowed theory, (5) recognize the complexity of the behavioralsciences and the available competing theories, and (6) recog- nize the situationalconstraintsof the borrowed the- ory. Marketing as aScience Has marketing achieved scientificstatus? Und erthe Lakatosian framework, the answerisa qualified yes. In itsfavoris the argument thatno othermasterre- search program has emerged that possesses excess empirical contentover marketing. Moreover, market- ing researchershave satisfied thiscriterion to the ex- tentthatnew d imensionsto theoriesborrowed from other d isciplines have been ad d ed . One example isthe explication of the Fishbein and A jzen (1975) attitu- d inalmod el. So muchresearchhasbeen accumulated forthis mid d le-range theory thatmulti-attribute atti- tud e mod elshave been the subject of atleastthree major review articles (Lutz and Bettman 1977, Ryan and Bonfield 1975, Wilkie and Pessemier 1973). A n illustration atthe research program levelcan be found in marketing channelsresearchwhichhas increasingly shifted away fromthe behaviorally based conflictres- olution research program(Ster 1969) toward sthe po- litical economy research program(A chrol, Reve, and Ster 1983; A rd t 1979, 1983; Frazier 1983; Ster and Reve 1980). To the extentthatthe politicaleconomy framework integrates both sociopolitical and eco- nomicstructure and processes, it may be argued that ithas greaterexplanatorypower in analyzing inter- organizational exchanges in marketing channels. (Ini- tial empiricalfind ings(Reve 1982), however, seem 32 / Journalof Marketing, Fall1985 to ind icate thatthe two research programsmay be more complementary than competing, withcentralization of powerbeing a majord iscriminating element.) Whatabout marketing asa d iscipline itself? His- tory hasitthatitevolved outof economics. However, itisevid entthat marketing hasnot replaced its orig- inal parentd iscipline. Ind eed , based on the SMF per- spective, this willnot occurunless economics be- comesa d egenerative bod y of research programs. The shift away fromso-called received microeconomic theory withitsalmosttotal d epend ence on axiomatic mod elsthatare not empirically valid ated toward snew theorieswitha strategic orientation (e.g., in infor- mation economics and ind ustrial organization eco- nomics) seemsto ensure a progressive bod y of re- search programs in economics. Moreover, the relationship between marketing and economics has grad ually evolved froma superior-subord inate basis toward sacoworkerbasis. For example, the interface and cross-fertilization of id easbetween marketing and economicshas recently been highlighted (Horsky and Sen 1980) with particular reference to pricing (Hauser 1984, Nagle 1984). However, marketing has yet to become amature science in the Lakatosian sense, foramature science containsasetof research programs "in whichnot only novelfacts but, in an importantsense, also novelaux- iliarytheories, are anticipated ; mature science-un- like ped estrian trial-and -error-has'heuristic power'" (Lakatos1978, p. 88). A sd etractorsof marketing have asserted , the d iscipline hasitsshare of "patched -up, unimaginative seriesof ped estrian 'empirical' ad just- ments" whichhave ". .. no unifying id ea, no heu- ristic power, no continuity" (Lakatos1978, p. 88). Suchformsof researchhave been cond emned as being shotgun, atheoreticalexercisesin fact gathering and d atacollection (Jacoby1978, Olson 1982). Conse- quently, it may nothave been too surprising forLeone and Schultz (1980) to remarkthatscientific general- izationshave been ratherelusive in marketing. Metamethod ology in Marketing The focusof d iscussion now shiftsfrom marketing's metatheoretical und erpinnings toward its method olog- icalorientation. In contrastto the vigorousd ialog con- cerning the d iscipline'sphilosophicalallegiance, d e- batesaboutmethod atametalevelhave been more infrequent(Sauer, Nighswonger, and Zaltman 1982). Deshpand e (1983, p. 104) hasremarked thatless at- tention hasbeen focused on "the implications of mar- keting theory for marketing researchmethod ." Essentially, this issue turnson the question of whether marketing's metatheoretical perspective biases itsresearch method ologies. A nd erson (1983) hasnoted thatthe d iscipline'slogicalempiricistperspective is characterized by itsreliance on the ind uctive statisti- calmethod exemplified by the PIMS stud ies, while Deshpand e (1983) hascommented on its neglect of qualitative method s. Ind eed , Bagozzi (1984) hasad - vocated thatstructural equationsmod eling be the ap- propriate researchmethod for marketing. While itis not d isputed thatstructural equationsmod eling hasa rightfulplace in marketing research, there is d anger in being overlyd epend ent on a particular method . Box (1976) labels this method ologicalsingularism as cookbookery orthe tend ency "to force all problems into the mold sof one ortwo routine techniques, in- sufficient thoughtbeing given to the real objectives of the investigation orto the relevance of the assump- tions implied by the imposed method " (p. 797). Method ological Pluralism The SMF framework, in contrast, calls forameth- od ologicalpluralism in the d iscipline-a view con- sistentwiththe Lakatosian tenetof trying to lookat thingseclectically. The proliferation of rivaltheories thusextend sto research method ologies. However, the SMF perspective d oes not sanction method ological anarchy(Feyerabend 1980) withits "anything goes" prescription thatseemsto suggestmaintaining an in- ventory of research techniquessimply forthe sake of variety. Rather, itcallsforthe need to more d eeply examine the limitations, assumptions, and relevance of researchmethod s employed in marketing research. Recentreviewsof factor analysis(Stewart1981), repeated measures analysis(LaTour and Miniard 1983), stepwise multiple regression (McIntyre etal. 1983), cluster analysis(Punj and Stewart 1983), d iscriminant analysis(Crask and Perreault 1977), canonicalcor- relation (Lambert and Durand 1975), conjointanaly- sis (Green and Srinivasan 1978), statistical signifi- cance testing (Sawyer and Peter 1983), and structural equationsmod eling (Bagozzi 1980, Fornell 1983) representattempts at accomplishing this objective as wellas pointing out possible misapplications of are- searchmethod . Moreover, sucheffortshave impli- cationsata d eeper, metamethod ological level. Hence, if marketing science should aimto seek knowled ge for the sake of knowled ge and d e-emphasize itstieswith the marketing manager(A nd erson 1983, Hunt 1984), the argument thatmarketersneed onlyapply research method sis consid erably weakened . A ttempts atas- sessing the problem-solving efficacy, ad vantages, and limitationsof researchmethod sshould therefore be encouraged , particularly in light of their growing so- phistication (Helgeson etal. 1984). Furthermore, in the SMF framework, the researchmethod would con- stitute an integralpart of the ceteris paribus clause whenevera theory is put to test. Hence, itisessential to thoroughly examine the researchmethod in testing Metatheory and Metamethod ology in Marketing / 33 the ceteris paribus clause before anyanomaly can be- come a refuting instance. In so d oing, however, marketing researchersshould resistthe temptation termed by Box (1976) asmath- ematistry, whichisa tend ency to red efine ratherthan solve astatistical problem. "Typically, there hasonce been astatistical problem withscientificrelevance but thishas long since been lost sight of" (Box 1976, p. 797). Box argues that mathematistry is harmful be- cause researcherswho are notthemselvesstatisticians may become "(O)verawed by what they d o notun- d erstand , .. .mistakenly d istrusttheirown common sense and ad optinappropriate proced ures d evised by mathematicians withno scientific experience" (p. 798). Clearly, this implies the need for applied and math- ematicalstatisticiansaswellas d esigners of investi- gations and d ata analysts in the d iscipline. Further, consistentwiththe viewsof Popper(1959) and Laud an (1984), the SMF perspective argues that itisreasonable to ad optonly those method sthatmar- keting researcherscan rationally d efend as being likely to achieve their objectives. Itis contend ed thatthe method ologicalobjectives of marketing researchers based on SMF tenetsare: (1) to subject theoriesto strong threatsof refutation, and (2) to apply researchmeth- od swiththe greatestproblem-solving efficacy to un- covernew find ings. Boththese objectives willcon- tribute to marketing'sachieving statusas amature science byaid ing the d evelopment and valid ation of theoriesin the d iscipline. Severalmod er research method sin marketing willnow be examined along SMF tenets, includ ing nulland range hypothesistesting, structural equationsmod eling, and various qualitative method s. StatisticalMethod s The d ominantd ata analytictechnique in marketing is the classicaltestof statistical significance. Forex- ample, Peter (1983) reports that many consumerre- searchersoften rely on statistical significance teststo provid e evid ence forthe valid ity of theirresearch hy- potheses. Severalscholarsin psychology(Meehl1967, 1978), ed ucation (Carver1978), and marketing (Peter 1983, Sawyer and Peter 1983) have faulted itsuse. Meehl (1978) asserts that". .. the almost universal reliance on merelyrefuting the null hypothesis asthe stand ard method for corroborating substantive theo- riesisaterrible mistake, is basicallyunsound , poor scientific strategy, and one of the worst things that ever happened in the history of psychology" (p. 817). Likewise, Peter (1983) remarksthatstatistical signif- icance tests ". .. taken alone, as theyfrequently are in consumer research, provid e an insufficienthurd le for jud ging research find ings" (p. 391). These criticisms point to the d eleteriouseffectthat trad itional null-hypothesistesting hashad on the d e- tection of progress and the accumulation of knowl- ed ge in marketing. Thus, if we are to avoid being d rowned "in amassof meaningless and potentially mislead ing junk" (Jacoby1978, p. 87), marketing re- searchersmustreevaluate thismethod foritsuseful- nessin generating and accepting knowled ge. Thisis because und erfalsificationist tenets, scientifictheo- riesmustbe continuallysubjected to severe tests. If marketing is incapable of generating suchtestsand cannot expose itstheoriesto strong testsof refutation (even with increasing measurement precision), then its objective of achieving the statusof amature sci- ence might be seriously und ermined . The basic problem with null-hypothesistesting proced ures lieswiththe inferencesaboutsubstantive theoriesmad e fromthe statisticalinformation they provid e (Meehl1967). The appraisal of asubstantive theory T entailssome constraint on the population value of the statistical parameter JL. However, the con- straintson pL in marketing are said to be weak (since we test against the straw man competitor, zero). Sta- tistical precision provid es information for d rawing in- ferencesabouthow ad equately aresearcherhases- tablished the actualvalue of pL and withwhat d egree of confid ence conclusions regard ing T can be mad e (Serlin and Lapsley1985). The crucial question is, having arrived atan es- timate of pL, even with perfectprecision, how d oes thisaffectthe plausibility of T? Meehl (1967) would argue thatthe typicalnull-hypothesis testin marketing is so weakasto be worthlesswhen passed . Thisis true notbecause of anyuncertaintyregard ing the pos- teriorestimation of the tested parameter's value but because the priorparameter constraintis so meager. In short, the initialconstrainton tL in marketing would be so weakthatthe statisticaltestcannot speak mean- ingfully to the plausibility of T. The reason forthis isthatthe null hypothesis in marketing haslittle to d o with point valuesorfunction formsd erived fromthe- ory. A s such, rejection of the null-which is quasi- always false (Meehl1967, 1978) and can almostal- ways be assured withsufficientstatistical power- provid es an easy observationalhurd le fortheoriesin marketing. Bagozzi (1984) offersan alternative method for theory construction and testing in marketing based on the structural equation approach. In contrastto the re- jection-support formof trad itional null-hypothesis testing, where rejection of the null impliesempirical support forthe theorybeing tested , the structural equation method isan example of the acceptance-sup- port formof hypothesistesting. Und erthe latter, the role of the research hypothesis is reversed (Fornell 1983), and itisthe acceptance of the null hypothesis (d erived fromsubstantive theory) that provid es em- piricalsupport forthe theory und ertest. The good - 34 / Journal of Marketing, Fall1985 ness-of-fittestin the structural equationsapproach has asits null hypothesis there being no d ifference be- tween the hypothesized mod eland the d ata. Empirical support forthe mod elisobtained bynonrejection of thisnull hypothesis. Thisformof hypothesistesting isnotwithoutits flaws. Recallthatwithsufficient power the nullisal- most alwaysrejected . Forthe rejection-support form of hypothesistesting, thismeansthat empiricalsup- port forone's theory is conceivablyalways forthcom- ing, whichin turn implies thatfew (if any) theories willbe invalid ated . Forthe acceptance-support form of hypothesistesting, sufficient power willensure that the substantive theory(hypothesized mod el) und ertest willbe refuted withthe almost guaranteed rejection of the null. Thisfurther implies thatone can con- ceivably neverfind empiricalsupport forone's theory and , in turn, a d iscipline may be leftwithout any the- ory whatsoever.3 One solution to this problem lies in ad opting a method thatis consistentwiththe SMF perspective and that, even withinfinite sample size, d oesnotal- waysreject the null hypothesis. A sMeehl (1978) has noted , when ascientistexaminesan experimental re- sult, he/she consid ersthe d egree to whichitisin rea- sonablygood accord with theory. Suchascientisthas therefore setstand ard sthatind icate whatkind sof ex- perimental outcomesare good enough. In effect, this imposes asetof constraintson the statistical param- etersto be estimated from sample d ata. Itis an ex- tension of the falsificationistd emand thatscientists establish, in ad vance, what they will accept asare- futing instance. Serlin and Lapsley(1985) recommend a hypoth- esis testing proced ure based on this "Good -Enough Principle." Essentially, their proced ure involvesthe replacement of the trad itionalnull hypothesis of no d ifference by a complex range null hypothesisstipu- lating a region of effects of triviald ifference.4 The magnitud e of this "good -enough belt" may be ex- pressed in termsof effectsizesormeasuresof asso- ciation. When the experiment is performed , astatis- 'Conversely, empiricalsupportmay be obtained given sufficiently low power. Note thatthe d egrees of freed omforthe likelihood ratio chi-square statisticused in structural equationsmod eling are notbased on sample size. However, large samples are required because small samples tend to prod uce unstable results. The reversalof the role of the research hypothesispresents another problem: the ability(power) to reject the research hypothesis is notknown (Fornell1983). 4The logic behind thisisthat hypothesistesting allows inferencesto be d rawn onlyfollowing rejection of the null, since researchershave tight statisticalcontrolover Type I error, i.e., the rate of false the- oretical support. Thus, to implementgood -enoughhypothesistesting, researchersmustfirstd etermine whatd ifference would be consid ered substantivelyimportant and empirically relevantand then specify its complement(i.e., ad ifference of less than substantive significance) asthe null hypothesis. Rejection of the null, therefore, provid es sub- stantive support forthe theory und ertest. ticaltestisused to establishif the expected value of the d epend ent variable iswithin the good -enough belt. If the d ataind icate so, then the null hypothesis isnot rejected and empiricalsupport forthe theory is not obtained . Note thatthe effectsof increased sample size are not problematic, since withincreased precision the imprecision involved in estimating the population value isred uced . In the limitof infinite precision, theoret- ical support isobtained byfind ing a sample value out- sid e the good -enough belt. Lackof such support is evid enced by a sample value found within the good - enough belt. Thus, even withinfinite sample size, the null hypothesis witha good -enough beltisnot always false. Foramore d etailed exposition and anumerical illustration of the statistical proced ure used to accom- panygood -enoughhypothesistesting, see Serlin and Lapsley(1985). Several suggestions have recently been forward ed to overcome the problems in structural equations mod eling (e.g., Fornelland Bookstein 1982, Forell and Larcker 1981). One suchmethod proposed by Bentlerand Bonnett (1980) seemsto be particularly consistentwiththe tenetsof good -enoughhypothesis testing. Briefly, theirmethod involves hierarchical comparisons and the use of normed and nonnormed incrementalfit ind ices thatred uce the problem of sample size and eliminate that concerning the reversal of the role of the research hypothesis. The Bentlerand Bonnett (1980) approach is still open to potential abuse in the case where mod elsthat would be rejected on an overall chi-square test may be accepted because theyrepresent a statisticallysig- nificant improvement overan alternative and highly unlikely nullmod el. Based on the good -enoughprin- ciple, however, thisd rawbackcan be alleviated since aresearcherwould be required to stipulate in ad vance a good -enough nullmod eld erived fromsubstantive theory aswellasthe incremental improvements in fit thatwould be consid ered acceptable. Ind eed , it sug- gests thatresearchersalso specify a priori the alter- native mod elsthatwillbe used in hierarchicalcom- parison. By so d oing, competing mod elsare tested against eachotherand not merelyagainst d ata. More- over, ex post facto analyses(e.g., examining partial d erivativesormod ification ind ices) that represent a source of ped estrian empiricalad justment d iscussed earlierare prevented . A s Cliff (1983) cautions, once aresearcherstarts ad justing amod elin light of the d ata, "the mod ellosesitsstatusasa hypothesis, and thatmod el finally chosen represents in practice amuch more unstable picture of whatis reallygoing on" (p. 124). In sum, the good -enoughprinciple servesatleast three purposes:(1) the a priori specification of what effects(mod els) are of substantive significance fol- lowsfalsificationisttenetsof d efining whatascientist Metatheory and Metamethod ology in Marketing / 35 will accept as "facts"; (2) provid es amore powerful testof a theorybyinsisting thatscientists quantita- tivelyspecify in ad vance whatd ifference (incremental improvement in fit) willbe consid ered cred ible and empiricallyrelevant; and (3) bystiffening the obser- vational hurd le, red ucesthe numberof theories-of-the- monthwhile d eveloping a bod y of good -enoughthe- ories. Qualitative Method s Recently, interesthasbeen rekind led regard ing the role of qualitative method sin marketing (Deshpand e 1983, Droge and Calantone 1984). Based on the SMF per- spective, the key issue appears to be whether quali- tative method scan help provid e ad d itional insights beyond that furnished by quantitative (statistical) method s. This appears to be the case. First, qualita- tive method s may aid in d iscovering new propositions or hypotheses(Reichard t and Cook 1979) suchasthe- ories-in-use (Zaltman, LeMasters, and Heffring 1982). Since thisisawell-d ocumented benefitof qualitative method s, we shallnot pursue itin greater d etailhere. Second , qualitative method s may be employed for rigoroustheorytesting. Campbell(1975), forexam- ple, d iscussed how to create "d egrees of freed om" in case stud ies, arguing for capitalizing on the richness of d etailwithin a single case bylooking for multiple implications of the theoreticalid eas being tested . The single case thusbecomesasetof d iverse manifesta- tionsof a theory. Eachmanifestation ratherthan each case can be thought of asaunitof analysis in which a particular effect may be present. When examined as a whole, the single case hasmore observationsthan variables. Thiscreatessufficient d egrees of freed om fortestsof substantive hypotheses. A n interesting ex- ample of Campbell's method isWilson's (1980) case stud y of the New England freshfishmarketwhich tested implications d rawn fromWilliamson's (1975) transactionscosts approach. McClintock, Brannon, and Maynard -Mood y(1983) extend ed Campbell's(1975) proposal in three signif- icant ways in introd ucing theircase clustermethod . These includ e: (1) using surveysampling proced ures and the optional ad d ition of quantified measurement and multivariate statistical techniques to qualitative approaches ford ata analysis; (2) d efining implement- able unitsof analysis thatare both sufficiently stable for sampling purposes and lend themselvesto the pos- sible application of stand ard ized cod es; and (3) having key informantsassistin the enumeration of a sampling frame, a proced ure somewhat analogous to snowball sampling. This approach isan example of how statis- ticaland qualitative method scan be fruitfully com- bined in theorytesting. Boththe case stud y and case clustermethod thus appear to be amenable to classical aswellas range hypothesistesting. A notherareaof contribution for qualitative meth- od sliesin theiruse for triangulation purposes(Desh- pand e 1983). The effectivenessof triangulation isbased on the premise thatthe weaknessesof one method will be offset by the counterbalancing strengths of another. Jick (1983) d elineated five important benefitsof trian- gulation:(1) itallowsresearchersto be more confi- d entof their results; (2) itstimulatesthe creation of inventive method sand new ways of capturing a prob- lemto balance withconventionald atacollection tech- niques; (3) it helps to uncoverd eviantd imensionsof a phenomenon outof whichold theoriesare mod ified ornew theories d eveloped ; (4) it synthesizes orinte- gratesexisting theories; and (5) itservesasatestfor competing theories. A final ad vantage implicit in the d iscussion sur- round ing qualitative method sliesin their variety. Van Maanen, Dabbs, and Faulkner (1982) provid e ad is- cussion of various qualitative method s includ ing case stud ies, participantobservation, content analysis, for- maland informal interviewing, archivald ata surveys, historical analysis, frame analysis, and ethnomethod - ology. However, just as withstatistical method s, method ological assessmentsneed to be carried outto d etermine the problem-solving efficacy of qualitative method s. Initialworkon the key informant technique (e.g., Houston and Sud man 1975, John and Reve 1982, Phillips1981) and focus groups(Cald er1977) seems to be promising. Limitations A firstlimitation of the SMF perspective is the ac- ceptability of itsmid d le-of-the road characteristic by ad herentsof empiricism and relativism. Ind eed , Lak- atoshasbeen criticized fornot being enough of arel- ativistoran empiricist(Berkson 1976, Toulmin 1976). Thus, a possible limitation is marketing's reluctance to accept the Lakatosian mid d le-of-the-road charac- teristic. A second criticismof the SMF perspective isthat it provid es no specific measure of scientific progress so thatitisd ifficultto d etermine if researchersin a d iscipline are making betterorworse choices (Suppe 1977, Toulmin 1976). In its d efense, the SMF per- spective offerssound criteriaforthe evaluation of the progressivity of research programs(e.g., internalcon- sistency, originality, unifying potential, and heuristic power). Evid ence forthis may be found byexamining the numberand quality of mid d le-range theoriesand the extentof research activity atthe levelof working hypotheses in a particular research program. More- over, to the extentthatresearchers pursue aresearch program based on some sortof a cost/benefit com- putation (Laud an 1977), aresearch program'spro- gressivity can be "measured " by its ability to attract 36 / Journal of Marketing, Fall1985 and retain the brightest and most creative researchers in the d iscipline. These researcherswould be the ones most capable of extend ing the positive heuristicof a research program and ensuring its continuity. A third criticismleveled at the SMF perspective is thatthe hard core of aresearch program d oes not constitute scientific reality(Berkson 1976). In other word s, researchers are seen to freely and easily shift their allegiance fromone research program to another. However, it is posited that this process is farfrom easy. A s an illustration, few behaviorists would sub- scribe to the information processing research program overnight, and vice versa. The Lakatosian explanation is that, as long asthe positive heuristicof behaviorism continues to generate novel facts, bothresearch pro- grams would continue to be pursued . A finalcriticismd irected toward s the SMF per- spective is thatitd oes not emphasize the creative as- pects of theory construction (Toulmin 1976). In his d efense, Lakatos intend ed to extend Popper's falsifi- cationist perspective in a progressive mannerwithhis notion of the research program. A s such, it was not his objective to suggest explicit meansof creating the- ories. However, byad vocating ahistoricalview of science, it may be argued thatLakatos ind irectly ad - d resses issues in the (so-called ) logic of d iscovery. Moreover, the SMF perspective argues forthe ad op- tion of techniques that may lead to the construction of new theories. The alleged limitations of SMF seemto be out- weighed by its potential benefits to the d iscipline. To reiterate, these benefits includ e: (1) encouraging both theoreticaland method ological pluralism, thus mini- mizing the possibilities of conceptualmyopia and biases in the choice of researchmethod s; (2) furnishing stan- d ard sof quality controlfor evaluating research pro- grams and hopefully ensuring a stead yad vancement of marketing science; (3) suggesting method ological objectives that provid e useful criteriafor assessing present and potential research method s; and (4) pro- vid ing areconstruction of marketing science consis- tentwiththe contemporary structure of the d iscipline. FinalRemarks In conclusion, the SMF frameworkisboth d escriptive and prescriptive. Itis d escriptive when applied to the rationalreconstruction of marketing science. In so d oing itis prescriptive, since itoffers general stand ard sfor a maturing d iscipline. Withthe growing numberof d ivergent philosophical views of marketing science, researchers in the d iscipline can now freely choose froma largerarray of metatheoretical perspectives. However, just as with competing research programs, the selection of an appropriate metatheoretical per- spective should be d etermined by its long-termprob- lem-solving efficacy and capability of initiating in- quiry into new areas. Forwhile it is important for marketing researchersto d isseminate theirtheoretical contributionsto otherscientific communities, itis their perceptions of whatis and whatis good fortheird is- cipline that ultimately affects d ay-to-d ay research. REFERENCES A chrol, R. S., T. Reve, and L. W. Stern (1983), "The En- vironmentof Marketing Channels Dyad s: A Frameworkfor Comparative A nalysis," Journal of Marketing, 47 (Fall), 55-67. A ld erson, W. and R. Cox (1948), "Toward sa Theory of Mar- keting," Journal of Marketing, 13 (October), 137-152. A nd erson, P. F. (1982), "Marketing, StrategicPlanning and the Theory of the Firm," Journal of Marketing, 46 (Spring), 15-26. (1983), "Marketing, Scientific Progress, and Sci- entific Method ," Journal of Marketing, 47 (Fall), 18-31. A rnd t, J. (1979), "Toward a Concept of Domesticated Mar- kets," Journalof Marketing, 43 (Fall), 69-75. (1981), "The Conceptual Domain of Marketing: Evaluation of Shelby Hunt's Three Dichotomies Mod el," European Journalof Marketing, 14 (Fall), 106-121. (1983), "The Political EconomyParad igm: Foun- d ation for TheoryBuild ing in Marketing," Journal of Mar- keting, 47 (Fall), 44-54. (1985), "On Making Marketing Science More Sci- entific: Role of Orientations, Parad igms, Metaphors, and Puzzle Solving," Journal of Marketing, 49 (Summer), 11- 23. Bagozzi, R. P. (1980), CausalMod elsin Marketing, New York: Wiley. (1984), "A Prospectus for Theory Construction in Marketing," Journal of Marketing, 48 (Winter), 11-29. Bartels, R. (1951), "Can Marketing Be a Science?," Journal of Marketing, 15 (January), 319-328. Bentler, P. M. and D. G. Bonnett (1980), "Significance Tests and Good ness of Fitin the A nalysis of Covariance Struc- tures," PsychologicalBulletin, 88 (no. 3), 588-606. Berkson, W. (1976), "LakatosOne and LakatosTwo:A n A p- preciation," in Essays in Memoryof Imre Lakatos:Boston Stud iesin the Philosophyof Science, Vol. 39, R. S. Cohen, P. Feyerabend , and M. W. Wartofsky, ed s., Dord recht:D. Reid el, 39-54. Bettman, J. R. (1979), A n Information Processing Theory of Consumer Choice, Read ing, MA : A d d ison-Wesley. Bower, G. H. (1981), "Mood and Memory," A merican Psy- chologist, 36 (February), 129-148. Box, G. E. P. (1976), "Science and Statistics," Journal of the A merican Statistical A ssociation, 71 (December), 791-799. Brown, H. I. (1977), Perception, Theory, and Commitment, Chicago:University of Chicago Press. Cald er, B. J. (1977), "Focus Groups and the Nature of Qual- Metatheory and Metamethod ology in Marketing / 37 itative Marketing Research," Journal of Marketing Re- search, 14 (A ugust), 353-364. Campbell, D. T. (1975), "'Degrees of Freed om' and the Case Stud y," Comparative PoliticalStud ies, 8 (July), 178-192. Carman, J. M. (1980), "Parad igms in Marketing Theory," in Researchin Marketing, Vol. 3, J. N. Sheth, ed ., Green- wich, CT:JA I Press, 1-36. Carver, R. P. (1978), "The Case A gainst Statistical Signifi- cance Testing," Harvard Ed ucationalReview, 48 (A ugust), 278-299. Chalmers, A . F. (1976), WhatIsThis Thing Called Science?, St. Lucia, A ustralia: University of Queensland Press. Cliff, N. (1983), "Some Cautions Concerning the A pplication of Causal Mod eling Method s," Multivariate Behavioral Research, 18 (January), 115-126. Converse, P. D. (1945), "The Development of aScience of Marketing," Journal of Marketing, 10 (July), 14-23. Crask, M. R. and W. D. Perreault, Jr. (1977), "Valid ation of Discriminant A nalysis in Marketing Research," Journal of Marketing Research, 14 (February), 60-68. Deshpand e, R. (1983), "'Parad igms Lost': On Theory and Method in Researchin Marketing," Journal of Marketing, 47 (Fall), 101-110. Dr6ge, C. and R. Calantone (1984), "A ssumptionsUnd erlying the MetatheoreticalDebates regard ing Method s and Sci- entific TheoryConstruction," in 1984 A MA WinterEd u- cators' Conference:Scientific Method in Marketing, P. F. A nd erson and M. J. Ryan, ed s., Chicago: A merican Mar- keting, 5-9. Duhem, P. (1953), "PhysicalTheory and Experiment," in Read ings in the Philosophyof Science, H. Feigl and M. Brod beck, ed s., New York: A ppleton-Century-Crofts, 235- 252. Fern, E. F. and J. R. Brown (1984), "The Ind ustrial/Con- sumer Marketing Dichotomy: A Case forInsufficientJus- tification," Journal of Marketing, 48 (Spring), 68-77. Feyerabend , P. (1980), A gainstMethod , Lond on:Verso. Fishbein, M. and I. A jzen (1975), Beliefs, A ttitud es, Inten- tions, and Behavior: A n Introd uction to Theory, and Re- search, Read ing, MA : A d d ison-Wesley. Fiske, S. T. (1982), "Schema-triggered A ffect: A pplications to Social Perception," in A ffect and Cognition, M. S. Clark and S. T. Fiske, ed s., Hillsd ale, NJ: Erlbaum, 55-78. Fornell, C. (1983), "Issuesin the A pplication of Covariance Structure A nalysis: A Comment," Journal of ConsumerRe- search, 9 (March), 443-448. and F. L. Bookstein (1982), "Two Structural Equa- tion Mod els:LISREL and PLS A pplied to ConsumerExit- Voice Theory," Journal of Marketing Research, 19 (No- vember), 440-452. and D. F. Larcker (1981), "Evaluating Structural Equation Mod els withUnobservable Variables and Mea- surement Error," Journal of Marketing Research, 18 (Feb- ruary), 39-50. Frazier, G. L. (1983), "InterorganizationalExchange Behav- iorin Marketing Channels: A Broad ened Perspective," Journal of Marketing, 47 (Fall), 68-78. Green, P. E. and V. Srinivasan (1978), "ConjointA nalysis in ConsumerResearch:Issuesand Outlook," Journal of Con- sumer Research, 5 (September), 103-123. Hauser, J. R. (1984), "Pricing Theory and the Role of Mar- keting Science," Journal of Business, 57 (no. 1, part2), S65-71. Helgeson, J. G., E. A . Kluge, J. Mager, and C. Taylor(1984), "Trend sin ConsumerBehaviorLiterature:A ContentA nal- ysis," Journalof ConsumerResearch, 10 (March), 449- 454. Holbrook, M. B. and E. C. Hirschman (1982), "The Experi- ential A spects of Consumption: ConsumerFantasies, Feel- ings, and Fun," Journal of ConsumerResearch, 9 (Sep- tember), 132-140. Horsky, D. and S. K. Sen (1980), "Interfacesbetween Mar- keting and Economics: A n Overview," Journal of Busi- ness, 53 (no. 3, part2), S5-12. Houston, M. J. and S. Sud man (1975), "A Method ological A ssessmentof the Use of KeyInformants," SocialScience Research, 4 (June), 151-164. Hunt, S. D. (1983a), Marketing Theory: The Philosophyof Marketing Science, Homewood , IL:Irwin. (1983b), "GeneralTheories and the Fund amental Explanand a of Marketing," Journal of Marketing, 47 (Fall), 9-17. (1984), "Should Marketing A d optRelativism?," in 1984 A MA WinterEd ucators' Conference:ScientificMethod in Marketing, P. F. A nd erson and M. J. Ryan, ed s., Chi- cago: A merican Marketing, 30-34. Isen, A . M., B. Means, R. Patrick, and G. Nowicki (1982), "Some Factors Influencing Decision-Making Strategy and Risk Taking," in A ffect and Cognition, M. S. Clarkand S. T. Fiske, ed s., Hillsd ale, NJ:Erlbaum, 243-261. Jacoby, J. (1978), "ConsumerResearch: A State-of-the-A rt Review," Journal of Marketing, 42 (A pril), 87-96. Jensen, M. C. and W. H. Meckling (1976), "Theory of the Firm: ManagerialBehavior, A gencyCosts, and Ownership Structure," Journal of FinancialEconomics, 3 (October), 305-360. Jeuland , A . P. and S. M. Shugan (1983), "Managing Channel Profits," Marketing Science, 2 (Summer), 239-272. Jick, T. D. (1983), "Mixing Qualitative and Quantitative Method s: Triangulation in A ction," in Qualitative Meth- od ology, J. Van Maanen, ed ., BeverlyHills, CA : Sage, 135-147. John, G. and T. Reve (1982), "The Reliability and Valid ity of Key InformantDatafrom Dyad icRelationships in Mar- keting Channels," Journal of Marketing Research, 19 (No- vember), 517-524. Kassarjian, H. H. (1978), "Presid ential A d d ress, 1977: A n- thropomorphism and Parsimony," in A d vances in Con- sumerResearch, Vol. 5, H. K. Hunt, ed ., A nn A rbor, MI: A ssociation forConsumer Research, xi-xiv. Kuhn, T. S. (1970), The Structure of ScientificRevolutions, 2nd ed ., Chicago:University of Chicago Press. Lachman, R., J. L. Lachman, and E. C. Butterfield (1979), Cognitive Psychology and Information Processing: A n In- trod uction, Hillsd ale, NJ:Erlbaum. Lakatos, I. (1978), "Falsification and the Method ology of Sci- entificResearch Programs," in The Method ologyof Sci- entific Research Programs: Imre Lakatos Philosophical Papers, Vol. 1, J. Worralland G. Currie, ed s., Cambrid ge: Cambrid ge UniversityPress, 8-101. and A . Musgrave, ed s. (1970), Criticismand the Growth of Knowled ge, Cambrid ge:Cambrid ge University Press. Lambert, Z. V. and R. M. Durand (1975), "Some Precautions in Using Canonical A nalysis," Journal of Marketing Re- search, 12 (November), 468-475. LaTour, S. A . and P. W. Miniard (1983), "The Misuse of Repeated Measures A nalysis in Marketing Research," Journal of Marketing Research, 20 (February), 45-57. Laud an, L. (1977), Progress and ItsProblems, Berkeley, CA : University of CaliforniaPress. (1984), "Reconstructing Method ology," in 1984 A MA WinterEd ucators' Conference:Scientific Method in Mar- keting, P. F. A nd erson and M. J. Ryan, ed s., Chicago: A merican Marketing, 1-4. Lazarus, R. S. (1984), "On the Primacy of Cognition," A mer- 38 / Journalof Marketing, Fall1985 ican Psychologist, 39 (February), 124-129. Leone, R. P. and R. L. Schultz (1980), "A Stud y of Mar- keting Generalizations," Journal of Marketing, 44 (Win- ter), 10-18. Lutz, R. J. and J. R. Bettman (1977), "Multiattribute Mod els in Marketing: A Bicentennial Review," in Consumerand Ind ustrial Buying Behavior, A . G. Wood sid e, J. N. Sheth, and P. D. Bennett, ed s., New York:Elsevier-NorthHol- land , 137-149. Mand ler, J. M. (1979), "Categorical and Schematic Organi- zation in Memory," in MemoryOrganization and Struc- ture, C. R. Puff, ed ., New York: A cad emic Press, 259- 299. Masterman, M. (1970), "The Nature of a Parad igm," in Crit- icismand the Growth of Knowled ge, I. Lakatos and A . Musgrave, ed s., Cambrid ge:Cambrid ge UniversityPress, 59-89. McClintock, C. C., D. Brannon, and S. Maynard -Mood y (1983), "A pplying the Logic of Sample Surveys to Qual- itative Case Stud ies:The Case ClusterMethod ," in Qual- itative Method ology, J. Van Maanen, ed ., BeverlyHills, CA : Sage, 149-179. McIntyre, S. H., D. B. Montgomery, V. Srinivasan, and B. A . Weitz (1983), "Evaluating the Statistical Significance of Mod els Developed byStepwise Regression," Journal of Marketing Research, 20 (February), 1-11. McSweeney, F. K. and C. Bierley(1984), "RecentDevel- opments in Classical Cond itioning," Journal of Consumer Research, 11 (September), 619-631. Meehl, P. E. (1967), "TheoryTesting in Psychology and Physics: A Method ologicalParad ox," Philosophyof Sci- ence, 16 (June), 103-115. (1978), "TheoreticalRisks and TabularA sterisks: Sir Karl, Sir Ronald , and the Slow Progress of Soft Psy- chology," Journal of Consulting and Clinical Psychology, 46 (A ugust), 806-834. Merton, R. K. (1957), Social Theory and SocialStructure, New York:Free Press. Nagle, T. (1984), "EconomicFound ationsfor Pricing," Jour- nal of Business, 57 (no. 1, part2), S3-26. Nord , W. R. and J. P. Peter (1980), "A BehaviorMod ifica- tion Perspective on Marketing," Journal of Marketing, 44 (Spring), 36-47. Olshavsky, R. W. and D. H. Granbois (1979), "Consumer Decision Making-Fact or Fiction?," Journal of Consumer Research, 6 (September), 93-100. Olson, J. C. (1982), "Presid entialA d d ress-1981: Toward a Science of Consumer Behavior," in A d vancesin Consumer Research, Vol. 9, A . Mitchell, ed ., A nn A rbor, MI:A s- sociation forConsumer Research, v-x. O'Shaughnessy, J. and M. J. Ryan (1979), "Marketing, Sci- ence and Technology," in Conceptual and TheoreticalDe- velopments in Marketing, O. C. Ferrell, S. W. Brown, and C. W. Lamb, Jr., ed s., Chicago: A merican Marketing, 577- 589. Peter, J. P. (1983), "Some Philosophical and Method ological Issues in Consumer Research," in Marketing Theory: The Philosophyof Marketing Science, S. D. Hunt, Homewood , IL: Irwin, 382-394. and W. R. Nord (1982), "A Clarification and Ex- tension of OperantCond itioning Principles in Marketing," Journal of Marketing, 46 (Summer), 102-107. and J. C. Olson (1983), "IsScience Marketing?," Journal of Marketing, 47 (Fall), 111-125. Petty, R. E. and J. T. Cacioppo (1981), A ttitud esand Per- suasion:Classicand ContemporaryA pproaches, Dubuque, IA :Wm. C. Brown. Pfeffer, J. and G. Salancik (1978), The ExternalControlof Organizations: A Resource Depend encyPerspective, New York: Harper. Phillips, L. W. (1981), "A ssessing MeasurementErrorin Key Informant Reports: A Method ological Note on Organiza- tional A nalysis in Marketing," Journal of Marketing Re- search, 18 (November), 395-415. Polanyi, M. (1958), Personal Knowled ge, Chicago: Univer- sity of Chicago Press. Popper, K. (1959), The Logicof ScientificDiscovery, New York: Harper. (1962), Conjectures and Refutations, New York: Harper. (1970), "NormalScience and Its Dangers," in Crit- icismand the Growth of Knowled ge, I. Lakatos and A . Musgrave, ed s., Cambrid ge:Cambrid ge UniversityPress, 51-58. Punj, G. and D. W. Stewart (1983), "Cluster A nalysis in Mar- keting Research:Review and Suggestions for A pplication," Journal of Marketing Research, 20 (May), 134-148. Reichard t, C. S. and T. D. Cook (1979), "Beyond Qualitative versus Quantitative Method s," in Qualitative and Quanti- tative Method sin Evaluation Research, T. D. Cookand C. S. Reichard t, ed s., BeverlyHills, CA : Sage, 7-32. Reve, T. (1982), "Interorganizational Relationsin Distribution Channel Dyad s," working paper, Norwegian School of Economics and Business A d ministration. Roberts, W. A ., Jr. (1984), "A Kuhnian Perspective on Mar- keting Science and the Scientific Method ," in 1984 A MA WinterEd ucators' Conference: Scientific Method in Mar- keting, P. F. A nd erson and M. J. Ryan, ed s., Chicago: A merican Marketing, 14-17. Robertson, T. S. and S. Ward (1973), "ConsumerBehavior Research:Promise and Prospects," in ConsumerBehavior: TheoreticalSources, S. Ward and T. S. Robertson, ed s., Englewood Cliffs, NJ: Prentice-Hall, 3-42. Ryan, M. J. and E. H. Bonfield (1975), "The Fishbein Ex- tend ed Mod eland Consumer Behavior," Journal of Con- sumerResearch, 2 (September), 118-136. Sauer, W. J., N. Nighswonger, and G. Zaltman (1982), "Cur- rentIssues in Philosophy of Science: Implications forthe Stud y of Marketing," in Marketing Theory:Philosophyof Science Perspectives, R. F. Bushand S. D. Hunt, ed s., Chicago: A merican Marketing, 17-21. Savitt, R. (1980), "HistoricalResearchin Marketing," Jour- nal of Marketing, 44 (Fall), 52-58. Sawyer, A . G. and J. P. Peter (1983), "The Significance of Statistical Significance Tests in Marketing Research," Journal of Marketing Research, 20 (May), 122-133. Serlin, R. C. and D. K. Lapsley(1985), "Rationality in Psy- chological Research:The Good -EnoughPrinciple," A mer- ican Psychologist, 40 (January), 73-83. Sheth, J. N. (1982), "ConsumerBehavior: Surpluses and Shortages," in A d vances in Consumer Research, Vol. 9, A . Mitchell, ed ., A nn A rbor, MI: A ssociation forCon- sumer Research, 13-16. Ster, L. W. (1969), Distribution Channels: BehavioralDi- mensions, Boston, MA : Houghton Mifflin. and T. Reve (1980), "Distribution ChannelsasPo- liticalEconomies: A Frameworkfor Comparative A naly- sis," Journal of Marketing, 44 (Summer), 52-64. Sternthal, B. and C. S. Craig (1982), ConsumerBehavior:A n Information Processing Perspective, Englewood Cliffs, NJ: Prentice-Hall. Stewart, D. W. (1981), "The A pplication and Misapplication of Factor A nalysis in Marketing Research," Journal of Marketing Research, 18 (February), 51-62. Suppe, F. (1977), The Structure of ScientificTheories, 2nd ed ., Urbana: University of Illinois Press. Metatheory and Metamethod ology in Marketing / 39 Taylor, S. E. (1980), "The Interface of Cognitive and Social Psychology," in Cognition, SocialBehavior, and the En- vironment, J. Harvey, ed ., Hillsd ale, NJ: Erlbaum, 189- 211. Toulmin, S. (1976), "History, Praxis, and the Third World : A mbiguities in Lakatos' Theory of Method ology," in Es- says in Memoryof Imre Lakatos:Boston Stud iesin the Phi- losophyof Science, Vol. 39, R. S. Cohen, P. Feyerabend , and M. W. Wartofsky, ed s., Dord recht:D. Reid el, 655- 675. Van Maanen, J., J. M. Dabbs, Jr., and R. R. Faulkner (1982), Varieties of Qualitative Research, BeverlyHills, CA : Sage. Wilkie, W. L. and E. A . Pessemier (1973), "Issuesin Mar- keting's Use of Multi-A ttribute A ttitud e Mod els," Journal of Marketing Research, 10 (November), 428-441. Williamson, O. E. (1975), Marketsand Hierarchies: A nalysis and A ntitrust Implications, New York:Free Press. Wilson, J. A . (1980), "A d aptation to Uncertainty and Small Numbers Exchange: The New England FreshFishMarket," BellJournal of Economics, 11 (A utumn), 491-504. Zajonc, R. B. (1980), "Feeling and Thinking: PreferencesNeed No Inferences," A merican Psychologist, 35 (February), 151- 175. (1984), "On the Primacy of A ffect," A merican Psy- chologist, 39 (February), 117-123. and H. Markus (1982), "A ffectand Cognitive Fac- tors in Preferences," Journal of ConsumerResearch, 9 (September), 123-131. Zaltman, G., K. LeMasters, and M. Heffring (1982), Theory Construction in Marketing:Some Thoughts on Thinking, New York: Wiley. Renaming your company If yourcompany has outgrown its name, perhaps we can help. A t NameLab, weVe mad e crisp, meaningful and memorable names likeA cura, Compaq and Zapmail by constructional linguistics. The resultof aNameLab project isa report presenting registrable names expressing yourcorporate id entityprecisely and powerfully We quote costs accurately in ad vance and complete most projects within fourweeks. Foran information packet, contactNameLab Inc., 711 Marina Blvd ., San Francisco, CA 94123,415-563-1639. NA MELA B? 40 / Journalof Marketing, Fall1985 1111