analogical reasoning for diagnosing strategic issues...

Strategic Management JournalStrat. Mgmt. J., 36: 2000–2020 (2015)

Published online EarlyView 14 November 2014 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/smj.2335Received 30 January 2013; Final revision received 9 August 2014

ANALOGICAL REASONING FOR DIAGNOSINGSTRATEGIC ISSUES IN DYNAMIC AND COMPLEXENVIRONMENTS

KENT D. MILLER1 and SHU-JOU LIN2*1 The Eli Broad Graduate School of Management, Department of Management,Michigan State University, East Lansing, Michigan, U.S.A.2 Graduate Institute of Global Business and Strategy, National Taiwan NormalUniversity, Taipei, Taiwan

Organizations interpret their environments by categorizing strategic issues as either opportunitiesor threats. They make such categorizations as inferences drawn from analogies from pastexperience. The accuracy of issue interpretations turns on: (1) which analogy is used, (2) whatare the environment’s properties, and (3) what is the timeframe? A computational model allowsus to evaluate over time the accuracy of interpretations based on different forms of analogicalreasoning in environments that differ in variation (unpredictability and dynamism) and complexity(dimensionality and ruggedness). This study elaborates a contingency approach to assessinganalogical reasoning by organizations in which the form of analogical reasoning, environmentalproperties, and time all matter. Our findings indicate when particular forms of reasoning producerelatively more accurate inferences about opportunities and threats. Copyright © 2014 John Wiley& Sons, Ltd.

INTRODUCTION

Strategic managers face the ongoing challenge ofinterpreting a wide array of environmental signalsand ascertaining their implications for their orga-nization’s strategy and performance. Along withdata from the environment, managers’ interpre-tive schema figure prominently in the diagnosisof strategic issues (Dutton, Fahey, and Narayanan,1983). Managers often reduce complex situationsdown to simple interpretations such as whether thecurrent environment presents a threat or an oppor-tunity (Dutton and Jackson, 1987; Jackson andDutton, 1988; Julian and Ofori-Dankwa, 2008). The

Keywords: analogy; computational model; environmen-tal complexity; organizational decision making; strategicissue diagnosis*Correspondence to: Shu-Jou Lin, Graduate Institute of GlobalBusiness and Strategy, National Taiwan Normal University,Taipei, 10610, Taiwan. E-mail: [email protected]

Copyright © 2014 John Wiley & Sons, Ltd.

importance of interpretations of strategic issues liesin their connection to decision making and organi-zational action (Dutton, Stumpf, and Wagner, 1990;Gilbert, 2006; Julian and Ofori-Dankwa, 2008).

Ascertaining the strategic implications of com-plex organizational environments often calls forreasoning by analogy (Dutton et al., 1983). Ana-logical reasoning draws upon a past experience ina similar, but not identical, situation to make infer-ences about the present situation (Holyoak, 2005;Holyoak and Thagard, 1995). The process of ana-logical reasoning involves elaborating and evaluat-ing the mapping of a known source (a rememberedexperience) onto a partially-unknown target (a cur-rent situation). After identifying a suitable analogy,information about the source is transferred to thetarget as a plausible but fallible inference (Gentnerand Markman, 1997). Analogical reasoning is rel-evant when a new situation requiring interpretationhas both important similarities and dissimilar-ities relative to prior experiences. Real estate

Analogical Reasoning in Dynamic and Complex Environments 2001

appraisal, setting salaries, attorneys’ judgments,and investors’ predictions about companies’prospects all apply analogical reasoning in decisionmaking (Gregan-Paxton and Cote, 2000; Holyoakand Thagard, 1995). Another example comes fromthe use of financial reporting as an analogy sup-porting firms’ adoption of sustainability reporting(Etzion and Ferraro, 2010). Directly relevant to ourtopic is the observation from Gavetti, Levinthal, andRivkin (2005) and Gavetti and Rivkin (2005, 2007)that strategists make use of analogical reasoning innovel environments. Research on analogical reason-ing can enrich our understanding of how managersprocess environmental data to classify strategicissues.

Gavetti et al. (2005) contend that using analog-ical reasoning aids organizational adaptation incomplex environments. They developed a compu-tational model that identified some contingenciesdetermining the effectiveness of analogical reason-ing in novel situations. Our study extends their lineof inquiry and responds to the call for research toidentify the situational contingencies that affect theperformance of analogical reasoning in strategicdecision making (see Gary, Wood, and Pillinger,2012). In particular, we ask: (1) which analogy is inuse, (2) what is the nature of the environment, and(3) what is the relevant timeframe? We start fromthe premise that there are different ways to reasonanalogically and the efficacy of these differentforms of reasoning when applied to strategic issuesmay depend upon features of the environmentin which they are used. The timeframe mattersbecause organizations need time to accumulatea repertoire of experiences from which to drawanalogies. The set of possible analogies grows asan organization learns over time, and this may ormay not improve decision making.

We advance this contingency view by using com-putational modeling to evaluate different kinds ofanalogical reasoning under different environmentalconditions. We assume that learning and applyinganalogical reasoning occur together over time, thuswe adopt a dynamic perspective toward evaluatinghow well analogical reasoning performs. Ourmodel allows us to explore a wide variety of pos-sible environments and various forms of reasoning,thereby identifying influences on the effectivenessof reasoning by analogy over time. We model anorganization operating in a changing environment,not on a fixed landscape. Encompassing consid-erations raised by Wholey and Brittain (1989)

and Levinthal (1997), our depiction of the envi-ronment includes variation (unpredictability anddynamism) and complexity (dimensionality andruggedness). Computational modeling allows us toset up a full factorial experimental design to derivethe implications over time of different forms ofreasoning in environments that differ along variousdimensions.

We evaluate the accuracy of two forms of ana-logical reasoning—best available match and sat-isficing match—in a wide array of environments.Analogical reasoning using the best available matchdraws inferences on the basis of the most simi-lar past experience. Satisficing chooses an anal-ogy that is deemed adequate to the situation, butnot necessarily the best match (see Simon, 1956,1976). We compare the accuracy of interpretationsbased on these two different approaches to ana-logical reasoning and using only exact matchesfrom prior experience. We find, for example, thatsatisficing is advisable only in early periods inenvironments with low unpredictability or low fre-quency of change. Depending on the environment,organizational learning can have positive, neutral,or negative effects on the accuracy of satisfic-ing reasoning over time. Using the best availablematch generally outperforms exact match reason-ing, except when unpredictability and ruggednessare high.

This study advances understanding of the con-tingencies affecting the quality of strategic issuediagnoses. Evaluating empirically the accuracyof organizations’ diagnoses of strategic issues isproblematic because such diagnoses are usuallyunobservable to outsiders and whether a situationis an opportunity or a threat can be difficult toascertain ex post. Furthermore, natural experimentspresent incomplete designs with respect to theset of contingencies affecting the quality of issuediagnoses. Computational modeling bypassesthese limitations, thereby allowing us to addressan important topic that empiricists working onstrategic issue diagnosis have not taken up. Overall,our analyses show the relevance of a contingencyapproach to assessing the accuracy of analogicalreasoning for classifying strategic issues. Theform of analogical reasoning, environmentalproperties, and time all matter. The accuracy ofanalogical reasoning about strategic issues dependsupon complex interactions among environmentalproperties.

Copyright © 2014 John Wiley & Sons, Ltd. Strat. Mgmt. J., 36: 2000–2020 (2015)DOI: 10.1002/smj

2002 K. D. Miller and S.-J. Lin

INTERPRETING ENVIRONMENTS BYREASONING ANALOGICALLY

Managers face the ongoing challenge of inter-preting the changing environments in which theirorganizations operate. According to Dutton et al.(1983), strategic issue diagnosis is a processwhereby managers order and give meaning toa continuous stream of diverse and ambiguousdata that characterize their organization’s situa-tion. Whereas environmental scanning generatesdata, managerial interpretation gives meaning todata, thereby promoting learning and facilitatingorganizational action (Daft and Weick, 1984).Interpretation moves from myriad observationsabout aspects of the situation to an integrated,holistic understanding (Boland, Tenkasi, andTe’eni, 1994). Data from the current environmentand organizational members’ past experiences enterinto this interpretation process (Daft and Weick,1984). Strategic issue diagnosis applies inferentialreasoning to arrive at fallible hypotheses rather thandefinitive conclusions (Dutton et al., 1983). Suchhypotheses about organizational situations guideaction through decisions that exploit perceivedopportunities when they arise in the environment(Eisenhardt and Sull, 2001). Hence, the accuracyof managers’ inferences carries implications for thestrategic actions and performance of organizations.Managers’ biased interpretations of organizationalenvironments, expressed as excessive optimism orpessimism about the organization’s prospects, canlead to inappropriate strategic actions (Martins andKambil, 1999).

The interpretations that we consider here aresimple categorizations of complex environments asfavorable or unfavorable situations. Managers fre-quently use the labels “opportunity” and “threat”to distinguish these situations (Dutton and Jackson,1987; Jackson and Dutton, 1988). Threatening cir-cumstances present potential loss and reflect lack oforganizational control, whereas opportunities affordpotential gains that may be controllable (Jacksonand Dutton, 1988; Thomas and McDaniel, 1990).Such categorizations affect how organizations pro-cess and respond to strategic issues (Chattopadhyay,Glick, and Huber, 2001; Dutton and Jackson, 1987).Prior studies indicate that interpreting an issue asan opportunity is likely to lead to a more proactiveresponse than if the situation is considered threat-ening (Sharma, 2000; Thomas, Clark, and Gioia,1993). Managers’ interpretations of organizational

environments guide their strategic deployment oforganizational capabilities and moves into new ven-tures (Cornelissen and Clarke, 2010; Eggers andKaplan, 2013). Managers must detect and interpretenvironmental signals that, on the basis of priorexperience, identify a threat before they can initiateremediating responses (Kiesler and Sproull, 1982).

Barr’s (1998) investigation of pharmaceuticalfirms’ interpretations of a regulatory change pro-vides an in-depth look at the process of interest here.The environmental change featured in that studywas an increase in the evidence of product efficacyrequired by the U.S. Food and Drug Administration.While the proposed legislation was being devel-oped, pharmaceutical firms connected the unfoldinglegislative process with other experiences, such asprior government interventions or public relationsproblems. Such analogies to familiar past expe-riences guided their anticipation of eventual out-comes. Over time, as the nature of the emerginglegislation became clearer, firms were able to ascer-tain more clearly its implications for the industryas a whole, and for their firm-specific performance.Managers’ interpretations of the issue, in turn, drovestrategic changes involving R&D, marketing, for-eign expansion, and diversification.

Prior research connects the diagnosis of strate-gic issues to various organizational characteristics,such as strategy and information processing struc-ture and culture (Dutton and Ottensmeyer, 1987;Thomas and McDaniel, 1990), diversity of thetop management team (Dutton and Duncan, 1987;Plambeck and Weber, 2009), resource dependencies(Milliken, 1990), as well as experience, inertia, andavailable resources (Denison et al., 1996). Studiesreport the relevance of information gathering andprocessing (Anderson and Nichols, 2007; Kuvaas,2002) and past experience (Denison et al., 1996;Plambeck and Weber, 2010), but do not identify thespecific hermeneutics (i.e., interpretive schemas)used by managers. Research on analogical reason-ing helps fill this gap by clarifying how managersmake use of environmental data to classify strategicissues.

The act of classifying environments suggeststhat an organization’s interpretive process followssome explicit or implicit schema. The characteris-tics of the current situation are compared with thoseof schematically-organized knowledge from pastexperience to arrive at categorizations of situationsbased on their degree of similarity (Jackson andDutton, 1988). Judgments of similarity compare



features and determine the extent to which theymatch (Tversky, 1977). The suitability of an anal-ogy depends upon the extent to which its featuresmap onto the target (Gentner, 1983; Holyoak andThagard, 1989, 1995). Using an analogous situa-tion, an inference can be made about the outcomeof the presenting situation, and this inference guidescurrent action. Understanding novel situations inthe light of similar familiar situations is the basicstructure of reasoning by analogy. Understandingof the current situation follows from comparativereasoning reflecting the general “as-structure” ofinterpretation.

However, the general claim that managersreason by analogy lacks precision; it offers insuf-ficient detail about the nature of such reasoning.Analogical reasoning may take different forms(Weitzenfeld, 1984). Analogical reasoning in achoice situation involves specifying a considerationset (i.e., the set of possible analogies retrieved frommemory for consideration), comparing analogiesin this set to the target, selecting an analogy, andderiving inferences (Markman and Moreau, 2001).To advance a theory of analogical reasoning fordiagnosing strategic issues, we need to specify howmanagers choose relevant experiences and reasonfrom them to current decisions and actions (seeCornelissen and Clarke, 2010; Etzion and Ferraro,2010; Foss and Lorenzen, 2009). Furthermore,we need to understand the conditions that deter-mine when analogical reasoning—in its variousforms—promotes organizational adaptation tocomplex environments. Prior research leaves unan-swered whether we need a contingency perspective(i.e., different forms of analogical reasoning aresuperior for different environments) or whethersome forms of analogical reasoning robustlyproduce accurate interpretations across differentkinds of environments. Computational modelingprovides a method for assessing the accuracy ofdifferent forms of analogical reasoning in differentkinds of environments.

The process of analogical reasoning modeled byGavetti et al. (2005) involved the following steps:(1) observe a subset of the current situation’s char-acteristics; (2) find a known analogy that is the bestmatch for the observed dimensions of this situa-tion; (3) choose the best performing general policyresponse based on average outcomes for the bestmatching situation; (4) randomly choose a set ofdetailed choices consistent with this general pol-icy; and then (5) use incremental search to enhance

performance. This characterization of analogicalreasoning reflects some key implicit assumptions.Step 1 assumes a limited and comparable set ofobservable characteristics for the current and pastsituations. Step 2 requires extensive prior experi-ence to build a broad portfolio of possible analogies.The learning process generating the set of possibleanalogies precedes application to analogical reason-ing. Step 3 assumes knowledge of the performanceimplications of an extensive set of policy deci-sions for any given analogy, and optimization withrespect to this set. Steps 4 and 5 are implicationsof the incomplete guidance provided by simplifiedrepresentations of the environment. Reasoning byanalogy is used just once to find a starting pointfor subsequent trial-and-error search on a fixedlandscape.

The situation reflected in our model divergesfrom these assumptions. First, our focus is oncategorizations of the environment as favorable orunfavorable (Dutton and Jackson, 1987; Jacksonand Dutton, 1988), rather than choices amongmany possible multidimensional organizationalstrategies. The problems posed here and by Gavettiet al. (2005) are distinct but both are relevant toorganizations. Complex strategies respond to thedetails of environments, but managers’ decisionsabout whether to deploy, postpone, or discardsuch strategies follow from holistic assessments ofenvironmental favorability. Second, our modeledorganization possesses no prior bank of experiencesfrom which to draw analogies; rather, learning fromexperience and decision making through analogicalreasoning occur together over time (Cornelissenand Clarke, 2010). Organizations often must drawconclusions from limited histories (March, Sproull,and Tamuz, 1991). We assume analogies emergeout of learning over time. Modeling the learningprocess from the beginning of a de novo organiza-tion distinguishes our approach from prior modelsof environmental interpretation (cf. Carley and Lin,1997; Gavetti et al., 2005). Third, given our focus,we isolate analogical reasoning and do not augmentit with other search and decision processes, whichintroduce confounding effects on the organization’sperformance. Hence, analogical reasoning fullycharacterizes organizational decision making in ourmodel. Fourth, we consider environments wherechange is ongoing, rather than one-time movesfrom one fixed landscape to another. Ongoingchange allows the organization to build up itsrepertoire of experiences from which to draw



analogies and simultaneously presents the needfor repeated application of analogical reasoning tointerpret the environment over time.

Environmental variation is multidimensional.Based on a review of prior research on environ-mental instability, Wholey and Brittain (1989)concluded that environmental variation occursalong three dimensions: frequency, amplitude, andpredictability. Frequency refers to the time betweenenvironmental changes. Amplitude indicates theextent of change or the distance between suc-ceeding environmental states. Predictability is thedegree to which a future state can be anticipated.The higher the number of possible states and themore equal their probabilities of occurring, theless predictable is the environment. Frequencyand amplitude make up the rate of change (ordynamism), whereas predictability relates to howreadily people can foresee the particular directionof change (McCarthy et al., 2010). These threedimensions of environmental variability are inde-pendent (Wholey and Brittain, 1989). As such, wecan set up a series of experiments with our modelthat covers the range of possible combinations thatmake up distinct forms of environmental variation.

Our interest in characterizing the environmentextends beyond its variation over time to its com-plexity. The complexity of an environment dependson the number of relevant industry and general envi-ronmental features and the extent of their interactiv-ity in determining whether a situation presents anopportunity or threat for an organization. We referto these features as an environment’s dimensional-ity and ruggedness, respectively. NK models featureboth forms of complexity: N refers to the number ofenvironmental elements and the contribution of anyelement depends on K other elements, so rugged-ness increases with K (Kauffman, 1993; Levinthal,1997). The concept “ruggedness” reflects interestgoing back to Emery and Trist (1965) in the causaltexture of environments, i.e., the interdependenciesamong environmental components and their effectson organizations. Ruggedness refers to the extent towhich interactions among environmental elementsaffect the resulting state (see Kauffman, 1993;Levinthal, 1997). For a smooth landscape, similarconfigurations of elements tend to cause the samestate. By contrast, in a rugged environment, similarconfigurations of elements often produce differentstates. Such an environment is unstable and can eas-ily tip from being favorable to unfavorable, or viceversa, due to minor perturbations of its components.

Our model combines complexity together withfrequency, amplitude, and unpredictability ofchange to characterize a broad range of possibleenvironments in which to test the efficacy ofdifferent forms of analogical reasoning.

MODEL SPECIFICATION

Environment

Each period in a model run represents the occur-rence of a configuration of environmental elementsthat presents a strategic issue for organizationalanalysis.1 The environment consists of n elementsrepresenting industry forces and factors in thebroader context that jointly determine whether theorganization faces an opportunity or threat. Eachenvironmental element takes one of two possiblequalitative states in any period—represented bythe values 1 or −1. Each of the 2n possible config-urations of the environmental elements determinesan overall state that is either an opportunity (1) ora threat (−1). The set of possible 2n configurationstogether with their resulting states make up alandscape.

As discussed above, environments can be char-acterized in terms of their variation (includingunpredictability and dynamism) and complexity(consisting of dimensionality and ruggedness).Dynamism separates into two facets: the frequencyand magnitude of changes. In all, there are fiveenvironmental characteristics incorporated into themodel.

Dimensionality

The range and heterogeneity of environmentaldimensions that are relevant to an organizationmake up important aspects of complexity (Child,1972; Daft, Sormunen, and Parks, 1988; Miller,Ogilvie, and Glick, 2006). In our model, dimension-ality is the number of environmental elements, n.

Unpredictability

An environment is predictable if it always pro-duces the same state regardless of the configura-tion of its elements. By contrast, unpredictability

1 The passage of time should be understood in terms of thesuccession of strategic issue events, which may be separated bylulls of varying duration during which no strategic issue arises.



is highest when the two possible outcome statesoccur with equal probability. Let q be the probabil-ity of an opportunity state, so (1− q) is the prob-ability of a threat state. When q equals 0.5, thelandscape is most unpredictable, whereas it is com-pletely predictable if q= 0 or 1 because the statenever changes.2 In other words, unpredictability isa triangular function of q peaking at q= 0.5.

Dynamism

We model two facets of environmental dynamism:the frequency and magnitude of environmentalchange (see Kim and Rhee, 2009; Wholey and Brit-tain, 1989). The frequency of environmental change(f ) is the inverse of the number of times that an issuerecurs before an environmental change. If the envi-ronment presents the same issue repeatedly beforeswitching to a new configuration then change occurswith low frequency.

When environmental change occurs, the proba-bility of a state change (from 1 to −1 or vice versa)for each element is p. The higher p, the greater is theexpected magnitude of the environmental change.

Ruggedness

Our specification of environmental ruggedness fol-lows a general polynomial function. This functionis a sum of the values for the environmental vari-ables and their cross products, each weighted byits assigned coefficient. The multiplicative (inter-action) terms indicate interdependencies among theenvironmental elements that contribute to determin-ing whether the overall environmental state is anopportunity or a threat. The general polynomialfunction is:

P (n) =n∑

i=1

j(i)∑j=1

𝛼{i}j X{i}

j (1)

where X{i}j indicates the jth i-variable monomial

term, 𝛼{i}j is its corresponding coefficient, and j (i) =(

ni

). We assume that coefficients take values

2 Because q is the probability of an opportunity state, it reflectsthe general level of munificence of the environment. Whereasunpredictability peaks at q= 0.5, munificence increases linearlywith q.

of either 1 or 0, thereby determining whether amonomial term is included or not in the polyno-mial equation.3 Landscape ruggedness generallyincreases with the number of cross-product coeffi-cients (i.e., 𝛼{i}

j for i≥ 2) with a value of 1.We assign states to environmental configurations

based on the values generated by a polynomialfunction. We first rank the cases in the order oftheir function values. An opportunity state (value of1) is assigned to a case when its functional valueis ranked in top q proportion of all (2n) cases;otherwise it is classified as a threat (state of −1).4

Organizational reasoning and accuracy

Analogical reasoning

When confronting a strategic issue, the organiza-tion considers analogous cases from those storedin memory in order to posit an inference about thecurrent state. The process of analogical reasoningconsists of identifying possible analogies that makeup a consideration set, evaluating the commonali-ties and alignable differences among analogies andthe target, and drawing inferences about the targetbased on the analogies deemed suitable (Markmanand Moreau, 2001). Figure 1 provides an overviewof the modeled organizational reasoning process. Inthe example in Figure 1, the environment is a vec-tor with n (=5) elements of either 1 or −1, andexperiences stored in memory are partitioned vec-tors consisting of the remembered n environmentalelements and the associated state.

Identifying relevant analogies is based on judg-ments of similarity between prior experiences andthe current situation (Gentner and Markman, 1997).Similarity is inversely related to the difference ordistance in the two situations’ features (Goldstoneand Son, 2005). Different subjective criteria forassessing similarity call for different functionalforms (Gilboa and Schmeidler, 2001). We use Ham-ming (H) distance to measure alignable differencesbetween environments experienced in the past and

3 Using only the value of 1 for all nonzero coefficients doesnot limit the generality of our findings. Using values for nonzerocoefficients drawn at random from a uniform distribution on therange (0, 1) would produce the same results on average.4 Ranks for environments that share the same polynomial functionvalue are resolved by referring to a random order established at thebeginning of a model run.



Figure 1. Flowchart and example for the organizational reasoning process

the current situation.5 Such an approach is con-sistent with prior research indicating the relevanceof superficial similarities (i.e., one-to-one corre-spondence of attributes), rather than similarity ofstructural relations, in the retrieval of analogiesfrom long-term memory (Gentner, Rattermann, andForbus, 1993; Holyoak and Koh, 1987; Ross, 1989).

5 Hamming distance (H) is the number of elements that must bechanged to make one environmental configuration match another.In other words, Hamming distance is the number of nonidenticalelements in a pairwise comparison of environments, which canrange from 0 (all elements identical) to n (all elements different).

Reliance on corresponding attributes in the retrievalof analogies is particularly evident for repeatedexperiences within a domain (Ross, 1987) as is thecase when an organization interprets repeatedly theenvironment in which it operates and draws analo-gies from its own past experiences.

Two plausible matching heuristics are using (1)the best available match or (2) a satisficing match.The former approach optimizes over a limited set ofexperiences, whereas the latter approach satisfices(Simon, 1956). When using the best available matchheuristic, the organization considers the memorizedpattern(s) having the minimum Hamming distance



(MIHD) from the current configuration of the envi-ronment. When applying the satisficing heuristic,the organization includes in its consideration set allcases with Hamming distance less than a thresholdvalue (i.e., H ≤ h). The maximum acceptable Ham-ming distance (MAHD), h, is an integer between 1and n. If the cutoff is n, then the organization deemsall past cases to be relevant analogies.

If no experience fulfills the applied heuristic, theorganization makes a random guess about the cur-rent state.6 If just one experience fulfills a heuristic,then the organization infers that the current statecorresponds with the outcome of that prior experi-ence. If more than one experience fulfills a heuristic,then the organization makes a probabilistic choicebased on the frequency of each state among thecases in the consideration set.

Exact match

For comparison, we also examine exact matchreasoning, which draws an inference about thestate from a prior experience that correspondsexactly with the current environmental configu-ration (H = 0). An exact match is not, strictlyspeaking, an analogy because analogical reasoningrequires that the source not be identical to the target(Holyoak, 2005).

Accuracy

Organizational performance is measured as its per-centage of correct interpretations (i.e., inferencesthat correspond with the actual state). We report theaccuracy of the organization’s interpretations overtime averaged across runs for a given model.

Experimental design and parameters

Our model provides a rich set of possible environ-mental conditions for evaluating the accuracy ofanalogical reasoning. We chose two divergent val-ues for each environmental parameter to set up ourexperimental design (see Law and Kelton, 2000: ch.12). Table 1 lists the environmental parameters andthe values chosen. For environmental dimensional-ity, we chose a low value, n= 5, that affords a widerange of levels of ruggedness. We briefly discuss the

6 In the initial period, an organization has no prior experience andsimply makes a random guess regardless of its interpretation rule.

Table 1. Environmental parameters and cases

Parameter Values Interpretation

n 5 Dimensionality: number ofenvironmental elements

q 0.1, 0.5 Unpredictability: proportion ofpositive outcomes among theconfigurations for a landscape

f 0.1, 1 Dynamism frequency: inverseof the number of times that anissue recurs before anenvironmental change

p 0.1, 0.5 Dynamism magnitude:probability that anenvironmental elementchanges

r 5, 30 Ruggedness: number ofmonomial terms in thepolynomial function

implications of higher environmental dimensional-ity (n≫ 5) at the end of the findings regarding inter-pretation accuracy, rather than present additionalfigures with these results.

Values of q equal to 0.1 and 0.5 correspondto low and maximum unpredictability. Environ-mental change occurs either once every 10 issues(f = 0.1) or every issue (f = 1). At p= 0.1, the envi-ronment experiences limited change. At values ofp approaching 1, the environment tends to switchback and forth between two configurations. Atp= 0.5, the environment configuration has maxi-mum variability over time, so we use this as thehigh value for dynamism magnitude. The polyno-mial expressions with 5 and 30 terms produce lowand maximum ruggedness.7 For satisficing reason-ing, we chose the threshold value of h= 1 as themaximum acceptable Hamming distance (MAHD)for possible analogies.

We analyze the accuracy of the three forms of rea-soning under each of 16 (=24) possible variants ofthe environment. Because the organization learns byaccumulating experience, the accuracy of analogi-cal reasoning changes over time. The graphs in thefigures plot mean interpretation accuracy for eachexperiment over 500 strategic issues. The reportedmean accuracy values are based on 1,500 runs of

7 Our measure of ruggedness is the proportion of all possibleenvironmental configuration pairs in a landscape with Hammingdistance of one having opposite states. This novel measureindicates the uniqueness of configuration states when comparedto those of all immediately adjacent (H = 1) configurations.



each model. Lower case letters (q, f , p, r) andupper case letters (Q, F, P, R) distinguish envi-ronments characterized by low and high levels ofunpredictability, dynamism frequency, dynamismmagnitude, and ruggedness, respectively.

EVALUATION OF ANALOGICALREASONING

Comparisons of interpretation accuracy

Figure 2 plots interpretation accuracy for thethree forms of reasoning in environments withhigh-frequency change (f = 1). Several key findingsappear in these results. First, interpretation accuracyimproves with experience for best available matchand exact match reasoning. Eventually, cumulativelearning causes the accuracy of these two formsof reasoning to reach 100 percent. However, thepaths over time differ with the environmentalcharacteristics. Organizations in environmentscharacterized by incremental change (i.e., low p)have higher accuracy in early periods but slowerlearning processes over time than organizationsin environments experiencing changes of greatermagnitude (i.e., high p). The greater the magnitudeof change, the more difficult is analogical reason-ing, but learning opportunities are also greater.Second, reasoning according to the best availablematch generally outperforms exact match andsatisficing match reasoning. The key exceptionoccurs in highly-rugged environments, where exactmatch reasoning outperforms interpretations basedon the best available match. When ruggedness ishigh, reasoning on the basis of a proximate, butless-than-exact, match often is not good enoughto draw accurate inferences. Third, dependingon the nature of the environment, experientiallearning can have positive, neutral, or negativeeffects on the accuracy of satisficing reasoningover time. Considering the specific cases wherelearning causes improvement (QFPr, qFPr, qFPR),little effect (qFpr, qFpR), or deterioration (QFpr,QFpR, QFPR) of accuracy for satisficing matchreasoning, no pattern appears at the level of mainor two-way effects. Instead, these differing impli-cations of learning reflect three-way interactionsamong unpredictability, dynamism magnitude, andruggedness.

Low-frequency change causes volatility ininterpretation accuracy over time. For reasoning

according to best available match and exact match,environmental changes cause single-period dropsin interpretation accuracy, but in subsequent peri-ods of environmental stability accuracy goes to100 percent due to learning from prior experiencein the identical environment (so the best availablematch is an exact match). The result is generallyhigh performance for these two forms of reasoning,with punctuating drops in interpretation accuracywhenever the environment shifts. Because of thispattern, plotted results for best available and exactmatch reasoning tend to overlap so Figures 3 and4 separate them into two different columns—bestavailable match on the left and satisficing match onthe right. Figure 3 covers environments character-ized by low-frequency change and low ruggedness.Figure 4 presents plots for low-frequency changeand high ruggedness. For each environment, theleft-hand and right-hand plots for satisficing matchare identical and serve as a baseline for comparisonwith the other two forms of reasoning.

Figures 3 and 4 affirm some patterns alreadyobserved in Figure 2. First, average interpretationaccuracy improves over time for both best availablematch and exact match reasoning irrespective of theenvironment. Second, best available match gener-ally outperforms exact match reasoning. The keyexception is when both unpredictability and rugged-ness are high. In unpredictable and highly-ruggedenvironments, analogical reasoning from proximatepast experiences is prone to errors. Third, in envi-ronments with low-frequency change, learning overtime generally diminishes interpretation accuracyfor satisficing match reasoning. When environmen-tal change occurs infrequently, early experiencesprovide exact matches from which to draw infer-ences about the state. As time passes, the organiza-tion draws from a growing set of possible analogiesthat satisfice yet are less accurate predictors of theenvironmental state than recent experiences.

Figures 3 and 4 also reveal that satisficingmatch reasoning results in much lower performancevolatility than do the other two forms of reasoning inenvironments with low-frequency change. Further-more, performance volatility from using satisficingreasoning diminishes with learning over time. Thesuboptimization associated with satisficing dimin-ishes both volatility and performance relative to bestmatch and exact match reasoning.

The worst performing decision rule differsdepending on the time horizon and context. Inearly periods, the exact match rule is generally



0 100 200 300 400 5000

0.2

0.4

0.6

1(a) (b)

(c) (d)

(e) (f)

(g) (h)

t

Inte

rpre

tatio

n ac

cura

cy

QFpr

Satisficing MatchBest Available MatchExact Match

0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

QFPr


0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

qFpr


0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

qFPr


0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

QFpR


0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

QFPR


0 100 200 300 400 5000

0.2

0.4

0.6

0.8 0.8

0.8 0.8

0.8 0.8

0.8 0.8

1

t

Inte

rpre

tatio

n ac

cura

cy

qFpR


0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

qFPR


Figure 2. (a–h) Interpretation accuracy in environments with high-frequency change. Correction added on 8 January2015: each L should be R (uppercase) and each l should be r (lowercase).



0 100 200 300 400 5000

0.2

0.4

0.6

1(a) (b)

(c) (d)

(e) (f)

(g) (h)

t

Inte

rpre

tatio

n ac

cura

cy

Qfpr

Best Available MatchSatisficing Match

0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

Qfpr

Exact MatchSatisficing Match

0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

QfPr


0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

QfPr


0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

qfpr


0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

qfpr


0 100 200 300 400 5000

0.2

0.4

0.6

0.8

1

t

Inte

rpre

tatio

n ac

cura

cy

qfPr


0 100 200 300 400 5000

0.2

0.4

0.6

0.8

0.8 0.8

0.8 0.8

0.8 0.8

1

t

Inte

rpre

tatio

n ac

cura

cy

qfPr


Figure 3. (a–h) Interpretation accuracy in environments with low-frequency change and low ruggedness. Correctionadded on 8 January 2015: each L should be R (uppercase) and each l should be r (lowercase).



0 100 200 300 400 5000

0.2

0.4

0.6

1(a) (b)

(c) (d)

(e) (f)

(g) (h)

t

Inte

rpre

tatio

n ac

cura

cy

QfpR


0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

QfpR


0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

QfPR


0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

QfPR


0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

qfpR


0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

qfpR


0 100 200 300 400 5000

0.2

0.4

0.6

1

t

Inte

rpre

tatio

n ac

cura

cy

qfPR


0 100 200 300 400 5000

0.2

0.4

0.6

0.8 0.8

0.8 0.8

0.8 0.8

0.8 0.8

1

t

Inte

rpre

tatio

n ac

cura

cy

qfPR


Figure 4. (a–h) Interpretation accuracy in environments with low-frequency change and high ruggedness. Correctionadded on 8 January 2015: each L should be R (uppercase) and each l should be r (lowercase).



the inferior approach in environments with lowruggedness. Following exact match reasoning, aninexperienced organization often resorts to randomguessing because it has no memory that matchesexactly its environment’s current configuration.An exact match is the optimal choice to guidedecision making but it is not always an availablechoice. In such situations, either form of analogicalreasoning would improve the organization’s abilityto diagnose strategic issues arising in its envi-ronment. Exact match reasoning only dominatesin early periods when both unpredictability andruggedness are high. Satisficing can be a reasonableapproach in early periods in environments withlow unpredictability or low frequency of changebut it is always inferior to the other two forms ofreasoning in the long run. The short-run potentialand the long-run inferiority of satisficing reasoningare important implications of our analyses.

To limit the tables included, we presented resultsonly for a relatively simple environment with n= 5elements. Based on unreported supplemental analy-ses, we can offer a few conclusions about analogicalreasoning in higher dimensional environments. Inmost cases, as environmental dimensionality (n)increases, the challenges associated with reason-ing from analogies increase for an organizationdrawing only on its own limited experience. In ahighly predictable environment (q→ 0 or q→ 1),increasing dimensionality does not adversely affectthe accuracy of inferences about the environment’sstate drawn from past experience. Likewise, ina stable environment (f → 0, p→ 0), analogicalreasoning works well. However, as the environmentbecomes unpredictable, unstable, and rugged,the accuracy of analogical reasoning from lim-ited experience breaks down in high-dimensionenvironments. Because the conditions that causeanalogical reasoning to fail may characterize some(possibly, many) organizations’ environments, oursupplemental findings highlight a practical concernabout whether and how organizations can copewith such circumstances.

Analysis of variance

The analyses presented thus far compared theaccuracy of predicted states using two dif-ferent forms of analogical reasoning and theexact match rule in different environments.However, we may also be interested in posing acomplementary question: which environmental

characteristics—unpredictability (q), dynamismfrequency (f ), magnitude (p), and ruggedness(r)—account for the accuracy of each form ofreasoning? Because interaction effects occur, theanswer to this question is not directly evident fromFigures 2–4. To address this question precisely,we applied multifactor analysis of variance to thedata generated by our simulation runs. In thesemodels, each of the four variables is dichotomizedwith 1 indicating the higher value and 0 otherwise.Therefore, the coefficients are easily interpreted asthe incremental changes in the expected interpre-tation accuracy relative to the sum of the overallmean (𝛼) and the lower-order component terms.We estimated the following model:

y = 𝛼+𝛽1q + 𝛽2f + 𝛽3p + 𝛽4r + 𝛽5qf + 𝛽6qp

+ 𝛽7qr + 𝛽8fp + 𝛽9fr + 𝛽10pr + 𝛽11qfp

+ 𝛽12fpr + 𝛽13prq + 𝛽14rqf + 𝛽15qfpr + 𝜀,

(2)

where y is the interpretation accuracy and 𝜀 is theobservation-specific error term.

Table 2 presents the multifactor analysis of vari-ance results for the two forms of analogical rea-soning and the exact match rule over the short run(periods 1–10), long run (periods 491–500), andentire run (periods 1–500).8 Each analysis is basedon 24,000 observations because there were 1,500runs for each of the 16 environmental contexts.

Some consistent patterns appear in Table 2. First,the intercepts (𝛼) are all close to 1, with valuesranging from 0.899 to 0.997. When unpredictabil-ity, frequency, dynamism, and ruggedness are low,interpretations based on any of the three forms ofreasoning are quite accurate. Apart from the inter-cept, most significant effects and all of the strongeffects, defined as |𝛽k|> 0.1 (k= 1, … , 15), are neg-ative. Generally, the effects of unpredictability, fre-quency, dynamism, and ruggedness are detrimentalto interpretation accuracy over the short, long, andentire runs.

Our results show that the effects of the environ-mental factors differ depending on the time hori-zon considered and form of reasoning used. Forthe three time horizons examined, environmentalfactors have the least predictive power (i.e., lowadjusted R2 as well as near zero coefficients) for

8 The analyses used the Stata (version 9) command anova withthe regression option.



Tabl

e2.

AN

OV

Afo

rin

terp

reta

tion

accu

racy

Var

iabl

esC

onst

ant

qf

pr

qfqp

qrfp

frpr

qfp

fpr

prq

rqf

qfpr

Coe

ffici

ents

𝛼𝛽

1𝛽

2𝛽

3𝛽

4𝛽

5𝛽

6𝛽

7𝛽

8𝛽

9𝛽

10𝛽

11𝛽

12𝛽

13𝛽

14𝛽

15A

djR

2

(A)

Max

imum

acce

ptab

ledi

stan

ce=

1(s

atis

ficin

gm

atch

)Sh

ort-

run

accu

racy

0.99

1**

−0.

007

−0.

091*

*−

0.03

5**

−0.

001

−0.

117*

*0.

005

−0.

016*

−0.

129*

*−

0.01

0†0.

001

−0.

003

−0.

010

0.00

7−

0.24

2**

0.00

60.

71L

ong-

run

accu

racy

0.89

9**

−0.

184*

*−

0.01

2−

0.01

2−

0.02

9**

−0.

018†

0.00

5−

0.32

4**

0.01

20.

001

0.00

8−

0.00

9−

0.00

9−

0.07

6**

−0.

059*

*0.

066*

*0.

57E

ntir

e-ru

nac

cura

cy0.

916*

*−

0.14

6**

−0.

027*

*−

0.01

3**

−0.

027*

*−

0.04

2**

−0.

009*

*−

0.27

9**

0.00

6*0.

008*

*0.

005

0.00

1−

0.01

6**

−0.

035*

*−

0.10

9**

0.03

7**

0.94

(B)

Min

imum

dist

ance

(bes

tava

ilabl

em

atch

)Sh

ort-

run

accu

racy

0.99

4**

−0.

009*

−0.

046*

*−

0.01

0**

−0.

002

−0.

073*

*−

0.02

5**

−0.

015*

*−

0.06

9**

−0.

004

0.00

1−

0.10

5**

−0.

015*

0.01

6*−

0.11

7**

−0.

042*

*0.

70L

ong-

run

accu

racy

0.99

7**

−0.

005*

*0.

002*

*0.

000

−0.

002*

*0.

005*

*0.

001

−0.

010*

*0.

000

0.00

2*0.

001

−0.

001

−0.

001

−0.

001

0.00

9**

0.00

20.

09E

ntir

e-ru

nac

cura

cy0.

995*

*−

0.00

7**

−0.

005*

*−

0.00

3**

−0.

001*

*−

0.00

8**

−0.

005*

*−

0.01

1**

0.00

3**

−0.

000

−0.

0004†

0.00

5**

−0.

000

−0.

006*

*−

0.01

7**

0.01

1**

0.93

(C)

Max

imum

acce

ptab

ledi

stan

ce=

0(e

xact

mat

ch)

Shor

t-ru

nac

cura

cy0.

979*

*0.

000

−0.

156*

*−

0.02

8**

0.00

2−

0.00

4−

0.00

0−

0.00

3−

0.22

2**

−0.

008

−0.

001

0.01

8*0.

009

0.00

40.

015†

−0.

022*

0.70

Lon

g-ru

nac

cura

cy0.

991*

*0.

000

0.00

9**

−0.

000

−0.

002*

−0.

000

0.00

00.

000

0.00

10.

002

0.00

1−

0.00

0−

0.00

1−

0.00

1−

0.00

00.

001

0.05

Ent

ire-

run

accu

racy

0.98

5**

−0.

000

−0.

017*

*−

0.01

1**

0.00

00.

000

−0.

000

−0.

000

0.01

1**

−0.

000

−0.

000

0.00

10.

001*

0.00

1†0.

000

−0.

001*

0.64

†p<

0.1;

*p<

0.5;

**p<

0.01

long-run accuracy, particularly when using bestavailable match and exact match reasoning. Asreported earlier, long-run accuracy associated withthese two heuristics is nearly perfect regardlessof the environmental characteristics. Contrastingly,when using satisficing match, environmental influ-ences on interpretation accuracy continue in thelong run, with strong environmental influences(|𝛽k|> 0.1) associated with unpredictability (q) andits interaction with ruggedness (qr). This pattern oflowest overall model fit (adjusted R2) in the longrun is an implication of organizational learning. Asorganizations accumulate diverse experiences, theaccuracy of their interpretations depends less onthe environmental characteristics. At the extreme,an organization has experienced and rememberedall 2n environmental configurations, and reasonsfrom these. Hence, the time horizon figures promi-nently in whether the accuracy of analogical rea-soning depends on the unpredictability, frequency,dynamism, and ruggedness of the environment.

For short-run accuracy, environmental character-istics account for about 70 percent of total variancefor each form of reasoning, however, the stronginfluences differ across the reasoning approaches.When using analogical reasoning, the strong influ-encing factors (|𝛽k|> 0.1) are interactions involvingsubsets of all four variables (rqf , fp, and qf for satis-ficing match; rqf and qfp for best available match).In contrast, the short-run accuracy of exact matchreasoning depends most strongly on the effects ofonly dynamism frequency and magnitude (fp andf ). In early periods, an organization following exactmatch reasoning has not yet recorded enough expe-riences to interpret accurately frequently changingenvironments.

For satisficing match and best available matchreasoning, environmental characteristics explainmore than 90 percent of total variance in interpre-tation accuracy over the entire run. In contrast, theANOVA model explains only 64 percent of variancewhen the organization follows exact match reason-ing. Environmental ruggedness never significantlyinfluences the interpretation accuracy for organi-zations following exact match reasoning becausesuch organizations avoid reasoning from proxi-mate experiences. Dynamism frequency (f ) andmagnitude (p) account for most of the explainedvariation for exact match firms for the entire runthrough their influence on experience accumulationand the relevance of past experiences for cur-rent inferences. Reasoning from exact matches



benefits from rapid accumulation of diverse expe-riences in environments characterized by frequentand high-magnitude change.

It is noteworthy that not every coefficient is neg-ative. For the models related to best match and sat-isficing analogical reasoning, all of the significantpositive coefficients are associated with interactionsinvolving dynamism frequency (f ) and appear inthe long-run and entire-run models. For satisficing,the overall effect of the frequency of change onaccuracy is negative in the long run and entire run,despite the occurrence of some weak positive inter-action effects. In contrast, for reasoning accordingto the best available match, the overall effect offrequency is unambiguously positive in the longrun. Whereas frequent environmental changes posea challenge, they also help build up organizationalmemory that improves the accuracy of inferencesover time. In the long run, the benefits of diverseexperiences outweigh the challenges. The positivelong-run effect of frequent environmental changesholds for exact match reasoning as well.

DISCUSSION

Interpreting environments by reasoninganalogically

Our study advances prior research on how orga-nizations classify their contexts as favorable orunfavorable (e.g., Dutton and Jackson, 1987; Jack-son and Dutton, 1988). Research on strategic issuediagnosis draws attention to the role that man-agers play in “how data and stimuli get inter-preted and understood” in organizations (Duttonet al., 1983: 309) and suggests that managers arriveat interpretations by reasoning analogically (Jack-son and Dutton, 1988). However, studies in thisline of research rarely elaborate the reasoning pro-cess involved (e.g., Denison et al., 1996; Gilbert,2006; Martins and Kambil, 1999; Sharma, 2000;Thomas and McDaniel, 1990). Although learningfrom experience is the basis for subsequent interpre-tations of organizational environments, how experi-ence leads to interpretations is not explored in depthin research on diagnosing environmental opportu-nities and threats. For example, empirical studiesshow that prior relevant experiences increase thelikelihood of interpreting an issue as an opportunity(Anderson and Nichols, 2007; Denison et al., 1996;Plambeck and Weber, 2010) but they do not fully

specify the learning and reasoning process involvedin arriving at this classification.

Our study draws from research on analogical rea-soning to advance theoretical understanding of theprocess of environmental interpretation for strategicissue classification. By bringing analogy research tothe study of environmental interpretation, we shedlight on the nature of the learning and inferentialreasoning involved. Details of the analogical rea-soning process connect past experiences and orga-nizational memory to inferences about a strategicissue that managers face in the current environment.We see further opportunities for theory building andtesting at the intersection of analogical reasoningand strategic issue diagnosis research.

Prior research (e.g., Denison et al., 1996; Miller,1993; Starbuck and Milliken, 1988) observes thataccumulated experience in an issue domain shapesmanagers’ diagnoses. Without careful characteriza-tion of how prior experiences are processed andapplied, theorized implications of experience forissue diagnosis could be superficial and vague. Inresponse, we put forward a model that specifiesexplicitly the learning process and plausible deci-sion heuristics associated with analogical reason-ing in strategic issue diagnosis. Analyses using ourmodel demonstrate which contingencies make theaccumulation of experiences over time beneficial ordetrimental to interpretation accuracy.

Overall, our study demonstrates the relevance ofa contingency approach to assessing the implica-tions of analogical reasoning by organizations inwhich the form of analogical reasoning, environ-mental properties, and time all matter. A contin-gency approach maintains that organizations maybenefit from taking different approaches to reason-ing and decision making depending on the context.Such a perspective poses the “meta-decision” prob-lem of deciding how to decide (Grandori, 1984),which carries normative implications, but also mayhave relevance for predicting how organizationsactually behave.

Five principal conclusions arise from our analy-ses of the three alternative forms of reasoning.

1. Interpretation accuracy improves over time forbest available match and exact match reasoning,but often declines for satisficing reasoning. Oldmemories, which are best forgotten, often mis-lead a satisficing organization. In dynamic envi-ronments, organizations should either abandonsatisficing in favor of the best available match



form of analogical reasoning or they should pro-mote unlearning of distant past experiences (seeMartin de Holan and Phillips, 2004a, 2004b;Tsang and Zahra, 2008).

2. Environmental dynamism erodes performance inthe short run, but enhances experiential learn-ing, thereby improving the accuracy of analogi-cal reasoning from the best available match overthe long run. Consistent with this contention,Gary et al. (2012) found that exposure to sourcevariation produced higher performance on ananalogous task among participants in a labo-ratory experiment. New entrants into dynamicenvironments may invest heavily in learning butexperience poor results in the short run, yetthrough their participation in a challenging envi-ronment they gain unique knowledge over timethat enables them to cope with environmentaldynamism.

3. Best available match generally outperformsexact match reasoning, except when unpre-dictability and ruggedness are high. This findingprovides the general rationale and boundarycondition for analogical reasoning by decisionmakers with limited experience relative to thediverse possibilities within their environment.Given limited experience, decision makingsneed to deploy the best of what they do know,rather than wait to acquire what they wouldideally like to know.

4. In highly-rugged environments, satisficing canproduce results that are worse than randomguessing. Satisficing is advisable only in earlyperiods in environments with low unpredictabil-ity or low frequency of change. Outside of stableor predictable contexts, we would expect—andadvocate—that organizations abandon satisfic-ing in favor of other forms of inferential reason-ing.

5. Satisficing can reduce performance volatility(relative to using best available or exact matchreasoning) when environmental change occursinfrequently, but lower risk comes at a cost toperformance. In most environments, an organi-zation would have to be extremely risk averse toaccept the risk-return tradeoff involved in usingsatisficing reasoning over the long run.

Analyses of which environmental factors accountfor the variance in interpretation accuracy pointedto high-level (three-way and four-way) interactionsamong environmental unpredictability, frequency

and magnitude of change, and ruggedness. Effectsof the various environmental factors differ depend-ing on the time horizon and form of reasoning used.These findings support the view that the accuracyof analogical reasoning is contingent upon complexinteractions among environmental characteristics,and they strongly caution against broad gener-alizations about the validity of guidance drawnfrom analogies. Different environmental conditionscall for different approaches to organizationallearning (Gnyawali and Stewart, 2003). In gen-eral, unpredictability, dynamism, and ruggednessdetrimentally affect the accuracy of analogicalreasoning. These effects persist over time aslong as the environment is sufficiently complexthat the organization’s cumulative learning aboutpossible environmental configurations remainsnonexhaustive.

Strategic issue diagnosis can occur as a reflec-tive and active process or as an unreflective andautomatic process (Dutton, 1993). Prior researchindicates that analogical reasoning often occurswith little conscious awareness on the part ofdecision makers (Gary et al., 2012; Schunn andDunbar, 1996). The potential for improving strate-gic issue diagnosis as a process of analogicaldecision making turns on whether managers arecapable of reflecting upon the details of theirreasoning, and then evaluating those processes inlight of the contingencies that characterize theirdecision contexts. Opportunities for improvinganalogical reasoning can be found in (1) morefully and accurately recording past experiences, (2)systematically searching experiences in memory,and (3) selecting analogies on the basis of criteriathat suit the environment and level of organizationalexperience. To the extent that managers rely uponanalogical reasoning to make strategic decisionsthat affect the overall direction and performanceof organizations, making individual managers’ andcollective management teams’ analogical reasoningprocesses more explicit and reflecting criticallyon these processes to find ways to improve thembecome compelling pursuits.

Our study carries various implications forresearchers interested in strategic issue diagnosis.Overall, our study points out the need to considerthe form of analogical reasoning, environmentalproperties, and learning timeframe, and theirbearing on the accuracy of organizational diag-noses of strategic issues. Our framing encouragesa shift in focus toward issue diagnosis quality,



a managerially-relevant yet under-researchedoutcome. Because environmental classificationstrigger organizational responses to environmentalstimuli (Chattopadhyay et al., 2001; Dutton andDuncan, 1987; Dutton and Jackson, 1987; Milliken,1990), we can infer that organizational maladapta-tion could, at least partially, result from problematicissue diagnoses. Nevertheless, the evaluation ofdiagnosis quality has never been a significant topicin the strategic issue diagnosis literature. Albeitchallenging, researchers should consider pursuinglongitudinal research that examines organizationallearning and reasoning processes over a prolongedobservation window in order to evaluate thevariables affecting issue diagnosis quality.

Computational modeling

Beginning with Levinthal’s (1997) study of orga-nizational adaptation on a rugged landscape,application of Kauffman’s (1993) NK model tomanagement issues has spread (see Ganco andHoetker, 2009; Vidgen and Bull, 2011). NK modelspresent static landscapes in which 2N configurationsare available to an organization as possible choicesat any point in time. The organization’s challengeis to find fitness-enhancing configurations througha process of searching and learning from perfor-mance feedback. The environment remains fixedand, although it determines the fitness outcomesfor alternative organizational configurations, theorganization never observes the environment.

We modeled a different problem—one involvingan organization’s interpretations of an observableand changing environment. Our approach assumesthat the environment is analyzable (Daft and Weick,1984); in other words, an organization can accessinformation about the many facets of its environ-ment. The organizational problem is to categorizethe current state as favorable or unfavorable on thebasis of the observed environmental data. Hence,our model places the environment in the foreground,but removes the details of the organization’s adap-tive response featured in prior NK models.

The organizational problem framed in thisstudy resembles that found in the computationalmodels of environmental interpretation developedby Carley (1992) and Carley and Lin (1997).Those studies modeled an organization that facesa series of multidimensional and varying situationsthat require simple classification decisions withonly two or three possible inferential conclusions.

Hence, the organization’s interpretations map ahighly complex situation onto a simple dichoto-mous or trichotomous conclusion. Organizationalmembers learn from feedback received in responseto their prior decisions. In particular, they learnthe true outcome of the situation, which remainedunknown at the time of their decision. Organiza-tional members relate remembered experiences tosubsequent situations, thereby improving decisionaccuracy over time.

Gavetti et al. (2005) provide an important prece-dent for the use of computational modeling tounderstand the nature and implications of analog-ical reasoning in organizations. Our study extendstheir effort in three key ways. First, we acknowl-edge that there are various ways to draw analo-gies (Weitzenfeld, 1984). The form of analogicalreasoning used in different organizational settingsremains an open question that empirical researchersare just beginning to address (Kim, 2011). Our studyconsidered two rules for identifying analogouscases—satisficing and best available match. Thesetwo specifications fit within a general understand-ing of analogical reasoning in choice situations asdrawing potential analogies from memory and eval-uating them on the basis of their similarities andalignable differences relative to the target to arrive atcandidate inferences applied to the target (Gentnerand Markman, 1997; Markman and Moreau, 2001).Our modeled decision process involves identify-ing a consideration set from organizational memoryand then drawing an inference based upon the fre-quency of experienced states. Lovallo, Clarke, andCamerer (2012) argue that using multiple sourcecases together can improve the accuracy of man-agers’ analogical reasoning for strategic decisions.

Second, we specified various environmentalproperties that affect the accuracy of reasoningfrom analogies. These include variation (unpre-dictability, frequency of change, and magnitudeof change) and complexity (dimensionality andruggedness). This set of environmental propertiescombines the key aspects of environmental change(Kim and Rhee, 2009; Miller et al., 2006; Wholeyand Brittain, 1989) with the emphases on dimen-sionality and ruggedness in NK modeling research(Kauffman, 1993; Levinthal, 1997). Gavetti et al.(2005) modeled analogical reasoning for one-timemoves from one static environment to another. Incontrast, we consider organizations in environmentsexperiencing ongoing change. Combining variationand complexity reflects the distinctiveness of these



constructs and the need for both to characterize fullyorganizational environments (Miller et al., 2006).

Third, we assume that learning and analogicalreasoning occur together over time. This assump-tion contrasts with that of some prior computationalmodels (e.g., Carley and Lin, 1997; Gavetti et al.,2005) that separate into distinct phases (1) organiza-tional learning through repeated experiences and (2)application of learning to subsequent experiences.Such a separation makes learning an off-line activ-ity (Gavetti and Levinthal, 2000), rather than keep-ing learning and sensemaking together. Assumingan on-line view (i.e., learning through interpreta-tion experiences) makes the accuracy of analogicalreasoning depend upon the accumulation of experi-ences in the organization’s memory over time (Cor-nelissen and Clarke, 2010).

Computational modeling facilitates theorydevelopment by permitting comprehensive andlongitudinal experimental designs that are not fea-sible in an empirical setting (Burton, 2003; Davis,Eisenhardt, and Bingham, 2007). We specifiedan original computational model that allowed usto set up a full factorial experimental design foralternative landscapes. This design generated datato assess various forms of analogical reasoningover time across a broad range of environments.By requiring precise specification of alternativeanalogical reasoning processes and deriving theirperformance implications over time, computationalmodeling contributes to theory building in a waythat complements empirical research on man-agers’ use of analogical reasoning in laboratoryand field settings. Furthermore, computationalmodeling calls for careful operationalization ofenvironmental constructs in a field in which verbalarguments are not always clear about the relevantenvironmental dimensions and their definitions(Miller et al., 2006).

We see various ways to extend our model infuture research. Modelers could, for example, con-sider alternative assumptions regarding organiza-tional memory and how organizations draw uponremembered past experiences when selecting analo-gies. The model presented here assumes that orga-nizations remember all past experiences and drawequally upon those experiences irrespective of theirage. Relative to what goes on in actual organiza-tions, this may be an excessively rational form oflearning and reasoning. We have not taken intoaccount organizational forgetting or the possibilitythat organizations can exhibit biases in choosing

among possible analogies. Rather than weighting allexperiences equally, organizations may favor earlycases (based on imprinting following founding) orrecent cases (due to their salience). Analyses of suchalternatives informed by research on organizationalmemory and decision making would extend furtherour explanation of the accuracy of analogical rea-soning.

Organizations’ possible responses to dynamicand complex environments extend beyond whatwe have considered and modeled in this study.Organizations adapt to complex, unpredictable, andunstable environments by specializing in ways thatlimit their exposure to a subset of the environment’sdemands. They both choose and enact environ-ments that simplify the set of environmental factorsto which they must attend. When possible, theyseek to control critical environmental contingencies(Pfeffer and Salancik, 1978). Organizations fosterintraorganizational specialization by creating dif-ferentiated divisions and roles that focus members’attention on distinct aspects of their complexenvironment (Lawrence and Lorsch, 1967). Suchspecialization helps to rationalize the allocationof individuals’ limited attention (Ocasio, 1997;Simon, 1971, 1976) and information-processingcapacity (Galbraith, 1977; March and Simon,1958). Environmental interpretation is a social andpolitical process, as well as a cognitive process(Dutton, 1997).

Following through with this richer conceptu-alization of organizations requires a shift from asingle-agent model of analogical reasoning to amultiple-agent model in which analogical reason-ing is decentralized. Whereas our study followsthe single-agent approach of Gavetti et al. (2005)to model analogical reasoning, Carley (1992) andCarley and Lin (1997) set up the multiple-agentproblem. In a multiple-agent model, individualswith distinct, limited focuses of attention useanalogical reasoning to draw inferences about theoverall environmental state from limited informa-tion, then they share their inferences with othersto generate an organization-level conclusion. Howagents allocate their attention across environmentalelements and other agents holds important implica-tions for the accuracy of organizations’ collectiveprocess of issue diagnosis. Furthermore, in ahierarchical organization, agents also actively exertupward influences through issue selling (e.g., Dut-ton and Ashford, 1993; Dutton et al., 1997, 2002)or downward influence through leader sensegiving



(e.g., Gioia and Chittipeddi, 1991; Gioia andThomas, 1996). Building on the current study,future research can address the multiple-agentproblem using different specifications of analogicalreasoning and our extended portrayal of the envi-ronment that includes variation and complexity.Such models should advance research on organi-zations as interpretation systems (Daft and Weick,1984; Maitlis, 2005; Weick, 1995) by makingexplicit the social networks and decision processesinvolved, and their performance implications.

We see opportunities for further advances inunderstanding strategic issue diagnosis as analogi-cal reasoning in organizations. Advances can comeas computational modelers and empiricists workmore closely together. Empirical researchers shouldseek access to organizations to identify how man-agers acquire and use analogies in actual practice.Such research can study individuals but also shouldseek to understand the social processes and envi-ronmental contingencies related to analogical rea-soning in teams and organizations. Remaining to beexplored are the critical connections between ana-logical reasoning, decision making, organizationalaction, and performance. Empirical research canground computational modelers’ specifications. Asillustrated in the present study, computational mod-eling can contribute to developing a process theorythat explains the dynamics of learning and applyinganalogical reasoning in organizations over time.

ACKNOWLEDGMENTS

Our thanks go to Richard Bettis, MartaGeletkanycz, June-Young Kim, and DanielLevinthal for helpful comments on earlier drafts.

REFERENCES

Anderson MH, Nichols ML. 2007. Information gatheringand changes in threat and opportunity perceptions.Journal of Management Studies 44: 367–387.

Barr PS. 1998. Adapting to unfamiliar environmentalevents: a look at the evolution of interpretation andits role in strategic change. Organization Science 9:644–669.

Boland RJ Jr, Tenkasi RV, Te’eni D. 1994. Designinginformation technology to support distributed cogni-tion. Organization Science 5: 456–475.

Burton RM. 2003. Computational laboratories for organi-zation science: questions, validity and docking. Com-putational & Mathematical Organization Theory 2:91–108.

Carley K. 1992. Organizational learning and personnelturnover. Organization Science 3: 20–46.

Carley KM, Lin ZA. 1997. A theoretical study of organiza-tional performance under information distortion. Man-agement Science 43: 976–997.

Chattopadhyay P, Glick WH, Huber GP. 2001. Organiza-tional actions in response to threats and opportunities.Academy of Management Journal 44: 937–955.

Child J. 1972. Organizational structure, environment andperformance: the role of strategic choice. Sociology 6:1–22.

Cornelissen JP, Clarke JS. 2010. Imagining and rationaliz-ing opportunities: inductive reasoning and the creationand justification of new ventures. Academy of Manage-ment Review 35: 539–557.

Daft RL, Sormunen J, Parks D. 1988. Chief Executivescanning, environmental characteristics, and companyperformance: an empirical study. Strategic Manage-ment Journal 9: 123–139.

Daft RL, Weick KE. 1984. Toward a model of organiza-tions as interpretation systems. Academy of Manage-ment Review 9: 284–295.

Davis JP, Eisenhardt KM, Bingham CB. 2007. Developingtheory through simulation methods. Academy of Man-agement Review 32: 480–499.

Denison DR, Dutton JE, Kahn JA, Hart SL. 1996.Organizational context and the interpretation ofstrategic issues: a note on CEOs’ interpretations offoreign investment. Journal of Management Studies33: 453–474.

Dutton JE. 1993. Interpretations on automatic: a differentview of strategic issue diagnosis. Journal of Manage-ment Studies 30: 339–357.

Dutton JE. 1997. Strategic agenda building in organiza-tions. In Organization Decision Making, Shapira Z (ed).Cambridge University Press: Cambridge, UK; 81–107.

Dutton JE, Ashford SJ. 1993. Selling issues to topmanagement. Academy of Management Review 18(3):397–428.

Dutton JE, Ashford SJ, Lawrence KA, Miner-Rubino K.2002. Red light, green light: making sense of theorganizational context for issue selling. OrganizationScience 13(4): 355–369.

Dutton JE, Ashford SJ, O’Neill RM, Hayes E, Wierba EE.1997. Reading the wind: how middle managers assessthe context for selling issues to top managers. StrategicManagement Journal 18(5): 407–423.

Dutton JE, Duncan RB. 1987. The creation of momentumfor change through the process of strategic issue diag-nosis. Strategic Management Journal 8: 279–295.

Dutton JE, Fahey L, Narayanan VK. 1983. Toward under-standing strategic issue diagnosis. Strategic Manage-ment Journal 4: 307–323.

Dutton JE, Jackson SE. 1987. Categorizing strategicissues: links to organizational action. Academy of Man-agement Review 12: 76–90.

Dutton JE, Ottensmeyer E. 1987. Strategic issue manage-ment systems: forms, functions, and contexts. Academyof Management Review 12: 355–365.

Dutton JE, Stumpf SA, Wagner D. 1990. Diagnosingstrategic issues and managerial investment of resources.Advances in Strategic Management 6: 143–167.



Eggers JP, Kaplan S. 2013. Cognition and capabilities:a multi-level perspective. Academy of ManagementAnnals 7: 295–340.

Eisenhardt KM, Sull DN. 2001. Strategy as simple rules.Harvard Business Review 79(1): 106–116.

Emery FE, Trist EL. 1965. The causal texture of organiza-tional environments. Human Relations 18: 21–32.

Etzion D, Ferraro F. 2010. The role of analogy in the insti-tutionalization of sustainability reporting. OrganizationScience 21: 1092–1107.

Foss N, Lorenzen M. 2009. Towards an understandingof cognitive coordination: theoretical developmentsand empirical illustrations. Organization Studies 30:1201–1226.

Galbraith JR. 1977. Organization Design. Addison-Wesley: Reading, MA.

Ganco M, Hoetker G. 2009. NK modeling methodology inthe strategy literature: bounded search on a rugged land-scape. In Bergh DD, Ketchen DJ Jr. (eds.), ResearchMethodology in Strategy and Management, EmeraldGroup: Bingley, UK (Volume 5: 237–268.

Gary MS, Wood RE, Pillinger T. 2012. Enhancing mentalmodels, analogical transfer, and performance in strate-gic decision making. Strategic Management Journal33: 1229–1246.

Gavetti G, Levinthal D. 2000. Looking forward andlooking backward: cognitive and experiential search.Administrative Science Quarterly 45: 113–137.

Gavetti G, Levinthal D, Rivkin JW. 2005. Strategy makingin novel and complex worlds: the power of analogy.Strategic Management Journal 26: 691–712.

Gavetti G, Rivkin JW. 2005. How strategists really think:tapping the power of analogy. Harvard Business Review83(4): 54–63.

Gavetti G, Rivkin JW. 2007. On the origin of strategy:action and cognition over time. Organization Science18: 420–439.

Gentner D. 1983. Structure-mapping: a theoretical frame-work for analogy. Cognitive Science 7: 155–170.

Gentner D, Markman AB. 1997. Structure mapping inanalogy and similarity. American Psychologist 52:45–56.

Gentner D, Rattermann MJ, Forbus KD. 1993. Theroles of similarity in transfer: separating retrievabilityfrom inferential soundness. Cognitive Psychology 25:524–575.

Gilbert CG. 2006. Change in the presence of residual fit:can competing frames coexist? Organization Science17: 150–167.

Gilboa I, Schmeidler D. 2001. A Theory of Case-basedDecisions. Cambridge University Press: Cambridge,UK.

Gioia DA, Thomas JB. 1996. Identity, image, and issueinterpretation: Sensemaking during strategic changein academia. Administrative Science Quarterly 41(3):370–403.

Gioia DA, Chittipeddi K. 1991. Sensemaking and sense-giving in strategic change initiation. Strategic Manage-ment Journal 12(6): 433–448.

Gnyawali DR, Stewart AC. 2003. A contingency per-spective on organizational learning: integration envi-ronmental context, organizational learning processes,

and types of learning. Management Learning 34:63–89.

Goldstone RL, Son JY. 2005. Similarity. In The CambridgeHandbook of Thinking and Reasoning, Holyoak KJ,Morrison RG (eds). Cambridge University Press: NewYork; 13–36.

Grandori A. 1984. A prescriptive contingency view oforganizational decision making. Administrative ScienceQuarterly 29: 192–209.

Gregan-Paxton J, Cote J. 2000. How do investors make pre-dictions? Insights from analogical reasoning research.Journal of Behavioral Decision Making 13: 307–327.

Holyoak KJ. 2005. Analogy. In The Cambridge Handbookof Thinking and Reasoning, Holyoak KJ, MorrisonRG (eds). Cambridge University Press: New York;117–142.

Holyoak KJ, Koh K. 1987. Surface and structural simi-larity in analogical transfer. Memory & Cognition 15:332–340.

Holyoak KJ, Thagard P. 1989. Analogical map by con-straint satisfaction. Cognitive Science 13: 295–355.

Holyoak KJ, Thagard P. 1995. Mental Leaps: Analogy inCreative Thought. MIT Press: Cambridge, MA.

Jackson SE, Dutton JE. 1988. Discerning threats andopportunities. Administrative Science Quarterly 33:370–387.

Julian SD, Ofori-Dankwa JC. 2008. Toward an integra-tive cartography of two strategic issue diagnosis frame-works. Strategic Management Journal 29: 93–114.

Kauffman SA. 1993. The Origins of Order: Self-organization and Selection in Evolution. OxfordUniversity Press: Oxford, UK.

Kiesler S, Sproull L. 1982. Managerial response to chang-ing environments: perspectives on problem sensingfrom social cognition. Administrative Science Quar-terly 27: 548–570.

Kim JY. 2011. How adaptive firms innovate: ananalogy-based view of organizational learning andinnovation. In Paper Presented at the OrganizationScience Winter Conference, Steamboat Springs, CO.

Kim T, Rhee M. 2009. Exploration and exploitation:internal variety and environmental dynamism. StrategicOrganization 7: 11–41.

Kuvaas B. 2002. An exploration of two competing per-spectives on informational context in top managementstrategic issue interpretation. Journal of ManagementStudies 39: 977–1001.

Law AM, Kelton WD. 2000. Simulation Modeling andAnalysis (3rd edn). McGraw-Hill: Boston, MA.

Lawrence PR, Lorsch JW. 1967. Organization and Envi-ronment: Managing Differentiation and Integration.Harvard Business School Press: Boston, MA.

Levinthal DA. 1997. Adaptation on rugged landscapes.Management Science 43: 934–950.

Lovallo D, Clarke C, Camerer C. 2012. Robust analogizingand the outside view: two empirical tests of case-baseddecision making. Strategic Management Journal 33:496–512.

Maitlis S. 2005. The social processes of organizationalsensemaking. Academy of Management Journal 48(1):21–49.



March JG, Simon HA. 1958. Organizations. John Wiley:New York.

March JG, Sproull LS, Tamuz M. 1991. Learning fromsamples of one or fewer. Organization Science 2: 1–13.

Markman AB, Moreau CP. 2001. Analogy and analogicalcomparison in choice. In The Analogical Mind: Per-spectives from Cognitive Science, Gentner D, HolyoakKJ, Kokinov BN (eds). MIT Press: Cambridge, MA;363–399.

Martin de Holan P, Phillips N. 2004a. Organizational for-getting as strategy. Strategic Organization 2: 423–433.

Martin de Holan P, Phillips N. 2004b. Remembrance ofthings past?: the dynamics of organizational forgetting.Management Science 50: 1603–1613.

Martins LL, Kambil A. 1999. Looking back and thinkingahead: effects of prior success on managers’ interpre-tations of new information technologies. Academy ofManagement Journal 42: 652–661.

McCarthy IP, Lawrence TB, Wixted B, Gordon BR.2010. A multidimensional conceptualization of envi-ronmental velocity. Academy of Management Review35: 604–626.

Miller D. 1993. The architecture of simplicity. Academy ofManagement Review 18: 116–138.

Miller CC, Ogilvie DT, Glick WH. 2006. Assessing theexternal environment: an enrichment of the archival tra-dition. Research Methodology in Strategy and Manage-ment 3: 97–122.

Milliken FJ. 1990. Perceiving and interpreting environ-mental change: an examination of college admini-strators’ interpretation of changing demographics.Academy of Management Journal 33: 42–63.

Ocasio W. 1997. Towards an attention-based view of thefirm. Strategic Management Journal 18: 187–206.

Pfeffer J, Salancik GR. 1978. The External Control ofOrganizations: A Resource Dependence Perspective.Harper & Row: New York.

Plambeck N, Weber K. 2009. CEO ambivalence andresponses to strategic issues. Organization Science 20:993–1010.

Plambeck N, Weber K. 2010. When the glass is halffull and half empty: CEOs’ ambivalent interpretationsof strategic issues. Strategic Management Journal 31:689–710.

Ross BH. 1987. This is like that: the use of earlier problemsand the separation of similarity effects. Journal ofExperimental Psychology 13: 629–639.

Ross BH. 1989. Distinguishing types of superficial simi-larities: different effects on the access and use of earlierproblems. Journal of Experimental Psychology: Learn-ing, Memory, and Cognition 15: 456–468.

Schunn CD, Dunbar K. 1996. Priming, analogy, and aware-ness in complex reasoning. Memory and Cognition 24:271–284.

Sharma S. 2000. Managerial interpretations and organiza-tional context as predictors of corporate choice of envi-ronmental strategy. Academy of Management Journal43: 681–697.

Simon HA. 1956. Rational choice and the structure of theenvironment. Psychological Review 63: 129–138.

Simon HA. 1971. Designing organizations for aninformation-rich world. In Communications, Comput-ers, and the Public Interest, Greenberger M (ed). JohnsHopkins Press: Baltimore, MD; 37–52.

Simon HA. 1976. Administrative Behavior: A Study ofDecision-making Processes in Administrative Organi-zation (3rd edn). Free Press: New York.

Starbuck WH, Milliken FJ. 1988. Challenger: fine-tuningthe odds until something breaks. Journal of Manage-ment Studies 25: 319–340.

Thomas JB, Clark SM, Gioia DA. 1993. Strategic sense-making and organizational performance: linkagesamong scanning, interpretation, action, and outcomes.Academy of Management Journal 36: 239–270.

Thomas JA, McDaniel RR Jr. 1990. Interpreting strate-gic issues: effects of strategy and the information-processing structure of top management teams.Academy of Management Journal 33: 286–306.

Tsang EWK, Zahra SA. 2008. Organizational unlearning.Human Relations 61: 1435–1462.

Tversky A. 1977. Features of similarity. PsychologicalReview 84: 327–352.

Vidgen R, Bull L. 2011. Applications of Kauffman’scoevolutionary NKCS model to management and orga-nization studies. In The Sage Handbook of Complexityand Management, Allen P, Maguire S, McKelvey B(eds). Sage: London, UK; 201–234.

Weick KE. 1995. Sensemaking in Organization. Sage:Thousand Oaks, CA.

Weitzenfeld JS. 1984. Valid reasoning by analogy. Philos-ophy of Science 51: 137–149.

Wholey DR, Brittain J. 1989. Characterizing environmen-tal variation. Academy of Management Journal 32:867–882.


analogical reasoning for diagnosing strategic issues...

Documents