Interpreting culture as a cognitive phenomenon shifts emphasis in assessing cul- ture from latent values (Hof stede 1980) to that of a socially constructed ontology of rationalized ideologies, interests, and ideas (Chia 1995, Abramson et al. 1996). This shift recasts a cultural group, such as managers or scholars, in terms of a par- ticular ontology that outlines the assumptions that harmonize thought processes (Weick 1979, Daniels 1991), moderates the use of intuitive versus analytical al- mir vol. 40, 2000/3 271 This content downloaded from 139.182.75.138 on Thu, 25 Apr 2019 06:51:48 UTC All use subject to https://about.jstor.org/terms Daniel P. Sullivan/Gary R. Weaver gorithms (Talmy 1995), shapes the mental structures used to perceive, process, and retrieve information (Abramson et al. 1996, DiMaggio 1997), explains the world through cause and effect relations (Maruyama 1984), and develops the heur- istics that mediate internal consistency versus external validity (Redding 1994, Chia 1995). The recurring use of this ontology sanctions its own rules, beliefs, propositions, and theories in ways that enact what has been variously called an “overall interpretive system schema,” (D’Andrade 1992), “mindset” (Maruyama 1984), or “dominant logic” (Prahalad/Bettis 1986). Thus, our study assumes that a culture incorporates distinct ontological traits that systematically shape the at- tributes of its “overall interpretive system schema.” Hence, we agree with DiMag- gio (1997, pp. 269-273) that “it may be useful to treat the schema as a basic unit of analysis for the study of culture . . . Individ Cultural Cognition in International Business Research Research supports this stance. The cognitive sciences report that people ex- press their logic of interpretation in the form of visual imagery (Shepard/Metzler 1971) or other non- verbal/non-textual mental representations (Abramson et al. 1996, Fiol 1994) that we can construe as an “orienting schema” (Weick 1979, Bou- gon 1992) or “cognitive maps” (Eden et al. 1992, Fiol 1994). Romney et al. (1996) and D’ Andrade (1995) note that the structure of an orienting schema, as much as it is abstract, spatial, and infused with meaning, reveals a person’s ontological and epistemological precepts. Furthermore, people use orienting schémas to make sense of events and information (Eden et al. 1992, D’ Andrade 1995), frame the reason- ing of decision-making (Abramson et al. 1996), and highlight objects, events, and priorities (Fiol/Huff 1992). Others report that people externalize, construe, or „re- represent” the elements and relationships that comprise their orienting schema through graphical expression, figure construction, map drawing, orientation judg- ments, and similar modes that represent knowledge in schematic, spatial forms (Fiol 1994, Markoczy/Goldberg 1995). Related research notes more precise links between culture and orienting schema, Levinson (1996, p. 363) for instance, re- port that spatial language and cognition vary across culture. In sum, evidence suggests a scholar’s schematic representations (i.e., explana- tory diagrams and figures) embody his or her interpretive frameworks just as tex- tual narratives do. A schematic representation, therefore, is an informative marker of the ontological assumptions of a researcher’s interpretive framework. Impor- tantly, it imbues a cognitive framework with ontological significance and serves as a proxy for more fundamental cognitive tendencies that are less easily accessible to researchers. Consequently, we hold that one can study the influence of culture on cognition by evaluating cultural variations in the schematic representations of ideas. Categories and Qualities of Schematic Representations Cognition research suggests three categories of schematic representation: Ana- log, composite, or propositional (cf. McNamara 1986, Fiol/Huff 1992, Eden et al. 1992). These categories have been used in prior research that specifically evalu- ated the ontology of interpretative frameworks reported in scholarly articles (Sul- livan 1998 A). An analog schema represents phenomena in a hierarchical order those lacks feedback loops or other means for “returning” to previously traversed levels of the hierarchy (Fiol/Huff 1992, Bougon 1992). An analog schema dis- plays unidirectional, step-by-step, and determinate progress through a hierarchy of states or processes as exemplified by the typical “box and arrows” schematic representation often found in management journals (see Exhibit 1). A propositional schema requires neither hierarchical structuring nor unidirec- tional transitions. Transitions or causal relationships may be effected in multiple ways that are non-recursive (i.e., no fixed step-by-step process is involved). End- mir vol. 40, 2000/3 273 This content downloaded from 139.182.75.138 on Thu, 25 Apr 2019 06:51:48 UTC All use subject to https://about.jstor.org/terms Daniel P. Sullivan/Gary R. Weaver Exhibit 1. Generic Forms of Schematic Representation of Interpretative Frameworks 274 mir vol. 40, 2000/3 This content downloaded from 139.182.75.138 on Thu, 25 Apr 2019 06:51:48 UTC All use subject to https://about.jstor.org/terms Cultural Cognition in International Business Research states may be reached by multiple paths; states that on one occasion function as “ends” in the schema may, on another occasion, function as the starting point (Weick/Bougon 1986, Sullivan 1998A). Exhibit 1 profiles the general form of the propositional representations often found in management journals. Finally, a com- posite schema (Fiol/Huff 1992, Sullivan 1998 A) represents an intermediary cat- egory that lacks the equifinality and multidirectionality of propositional catego- ries, but nevertheless allows idiosyncratic departures from analog standards. The composite form can appear as an analog form with added feedback linkages yet without the reciprocity across states that is manifest in a propositional schema (see Exhibit 1). Cognition research also specifies the schematic properties that apply to the forms of schematic representation (Eden et al. 1992, Markoczy/Goldberg 1995, Sullivan 1998 A). We can describe a schema in terms of its (1) comprehensive- ness, i.e., number of different entities or ideas they incorporate that in a typical diagram is represented by the number of nodes, (2) connectedness, i.e., the num- ber of distinct linkages in a schema that denotes the density of relationships among the entities or ideas in the schema, and (3) complexity, the ratio of connectedness to comprehensiveness within a schema that denotes the overall intricacy of the interaction represented between ideas and relationships (Bougon 1992, Fiol/Huff 1992, Sullivan 1998A). Variations in the characteristics of comprehensiveness, connectedness, and complexity fundamentally distinguish schematic representa- tions. Eden et al. (1992, p. 313) note, for example, that “a higher ratio (of links to nodes) indicates a densely connected map and supposedly a higher level of cog- nitive complexity.” Methodology We analyzed the influence of culture on cognition in IB research by examining variation in the spatial forms and structural features of schematic representations in premier IB journals regarding four proxies for culture. The applied stages of the study involved (1) identifying and coding the schematic representations in IB research publications, (2) synthesizing meaningful measures of culture, and (3) analyzing the relationship of these representations to the authors’ cultural con- texts. Domain The universe for this study is every “Figure” associated with any Article, Note, Comment, or Reply published in MIR and JIBS from January 1970 through De- ntir vol. 40, 2000/3 275 This content downloaded from 139.182.75.138 on Thu, 25 Apr 2019 06:51:48 UTC All use subject to https://about.jstor.org/terms Daniel P. Sullivan/Gary R. Weaver cember 1997. We focus on MIR and JIBS given their stature as the preeminent outlets for IB research (Ebrahimi/Ganesh-Copala/Chandy 1991, Lohrke/B niton 1997). The start of publication overlap between MIR and JIBS sets the period of study. We do not include data tables in that they are distinct from Figures and, per convention, presented in a standard form. From 1970 through 1997, MIR published 811 articles that contained 813 figures while the corresponding counts for JIBS are 734 articles and 45 1 figures. Review of the population found that some figures were charts or graphs generated by secondary data or statistical findings. Their descriptive intent fundamentally differs from those figures that schematically represent an interpretive framework. Thus, we exclude any representation of secondary or statistical data in the form of a scatter plot, ordinal interaction effect, histogram, pie chart, step function, bar graph, time-series, probability distribution, regression line etc. Too, some figures presented information in tabular or classification form – i.e., a literature review or a summary of hypotheses. Eliminating these figures reduced the universe of 1,264 figures to a population of 371 figures. We used the authors’ formal declaration of the purpose of the figure, as ex- pressed in keywords in the title of the figure, to identify those figures that for- mally and explicitly designate a schematic representation of the authors’ interpre- tation of the phenomena in question. The cognitive (D’ Andrade 1995), sensemak- ing (Chia 1995), and linguistics (Barley et al. 1988) literatures suggested three keywords: Model, Framework, or Conceptualization. The 149 figures whose ti- tle includes at least one of these keywords make-up our sample. Research Participants We used two sets of participants that we labeled “Academic” and “External.” The Academic set includes two researchers with terminal degrees. They inde- pendently rated the spatial form of the 149 Figures as either analog, composite, or propositional. We were averse to sampling additional fellow academics to val- idate their ratings. Research notes that organizational interest groups develop mutually shared orienting schémas (Kuhn 1962, Daniels 1991) as their members identify with a common interpretive framework that normalizes cognitive pro- cesses. Controlling for this effect among similarly trained individuals is often in- effectual, Finney and Mitroff (1986) add, because such “consensual cognitive schémas and scripts” often are embedded in an institution’s culture and used un- consciously by its members. Therefore, three business executives evaluated the spatial form of the 149 Figures. Two executives had global operating responsibil- ities (one in finance, one in marketing) in a Fortune 20 industrial MNC while the other had global staff (treasury) responsibilities in a Fortune 100 service MNC. 276 mir vol. 40, 2000/3 This content downloaded from 139.182.75.138 on Thu, 25 Apr 2019 06:51:48 UTC All use subject to https://about.jstor.org/terms Cultural Cognition in International Business Research Rating Schematic Representations We photocopied each figure onto a separate, letter-sized sheet of paper. We removed each title to preclude potential systematic bias – e.g., grouping of schémas by titles. We organized the figures chronologically, numbered them se- quentially, and then used a random number generator to set the sequence. The interaction between the administrator and a participant began with a twenty-min- ute talk about orienting schema and interpretative frameworks. The administra- tor noted the potential for different graphical forms and general attributes of an- alog, propositional, and composite forms but shared no schematic exemplars. Each participant then sorted the figures into one of three piles based on schematic sim- ilarity. Measures of Culture We considered culture from four viewpoints: Nation, Region, Cross-Cultural, and Disciplinary Perspective. We operationalized national culture with the author’s self-reported home-country institutional affiliation. This proxy, although simplis- tic, fits with theory and practice. Cultural cognition theory suggests that an individual’s effort to reduce cognitive dissonance motivates an affiliation marked by fit with the consensual schémas at play in the institutional environment and, by extension, the nation (DiMaggio 1997, Talmy 1995, Newman/Nollen 1996). As to practice, scholars routinely categorize an individual’s culture by the nation- ality of organizational affiliation (e.g., Hitt et al. 1997, Calori et al. 1994, Sharp/ Salter 1997). The second proxy of culture is Region – namely, North America, Europe, and Asia. We consider geographic regions given Hitt et al. ‘s (1997, p. 121) emphasis of the “importance of understanding strategic orientations in multiple regions of the world.” Procedurally, nations were aggregated into regional blocs based on geographical proximity. The third proxy of culture follows from the fact that many articles were jointly or multi-authored. If all authors listed affiliations with insti- tutions domiciled in the same nation or region, then we placed it in its correspond- ing category. Several articles listed authors of different national (e.g., Britain and Germany) and/or regional (e.g., North America and Europe) affiliations. We treat these articles as a proxy of Cross-national or Cross-Regional effects. Our fourth proxy of culture is the disciplinary perspective within IB, opera- tionalized in terms of whether the Figure appeared in MIR or JIBS. Although a loosely-coupled paradigm, IB manifests ontological and epistemological norms that allude to active “consensual cognitive schémas and scripts” (e.g., Schollham- mer 1994, Daniels 1991, Vernon 1994). The premier research outlets within the discipline, however, are formally located in different national and regional cul- mir vol. 40, 2000/3 277 This content downloaded from 139.182.75.138 on Thu, 25 Apr 2019 06:51:48 UTC All use subject to https://about.jstor.org/terms Daniel P. Sullivan/Gary R. Weaver tures and also have different editorial histories: JIBS is North American, with pre- dominately US editorship over its history, while MIR ‘s editorial history is largely German. Conceivably, these different cultural settings influence editorial review or authorial self-selection of the more appropriate outlet for particular types of interpretations. Hence, each journal may represent a different disciplinary per- spective that functions as a cultural moderator of cognition (e.g., the “North Amer- ican” effect, Teagarden et al. 1995). Also, estimating effects of disciplinary per- spective by controlling for the journal source should clarify the cultural cogniti- vism of IB studies as an intellectual construction plus help benchmark national and regional effects. Measuring Properties of Figures Two people independently counted the structural characteristics of each Model. We estimated comprehensiveness by counting the total number of nodes within a figure (Fiol/Huff 1992). We followed Stubbart’s (1989) statement that the textual designation of an object within a map signifies a node. We estimated connected- ness by counting the number of links between nodes within a figure. We followed reports that the “head” of a connection (the terminus of a line originating in one node that links to another) signifies a link between nodes (Eden et al. 1992, p. 313, Calori et al. 1994). If the arc or vector is reciprocal, then we counted it as two links (Bougon 1992). Finally, we followed precedent (e.g., Eden 1992, Eden et al. 1992, Fiol 1994, Laukkanen 1994) and estimated complexity as the ratio of connectedness (number of links) to comprehensiveness (number of nodes) within a figure. Temporal Issues Our first inclination for setting periodicity was to follow convention (e.g., Ink- pen/Beamish 1994) and segment the 1970-97 sample period by fixed-year peri- ods. An absolute specification of time, though, seemed prone to systematic error. Studies of the diffusion and use of knowledge suggest that research thought and output is marked by a continuous interweaving of cognition and action in a so- called process of “reflection in action” (Schon 1983). Practically, the rate of pub- lication of articles by MIR and JIBS changed over the sample period. Therefore, we control for possible interaction among cognition, diffusion, and time by set- ting periodicity in terms of “Publication Clusters.” Procedurally, we segment the 1 ,545 MIR and JIBS articles published from 1970 through the third quarter of 1997 into five Publication Clusters of 309 articles. Publication Cluster membership is a function of the article’s date of publication – e.g., the chronologically first 309 278 mir vol. 40, 2000/3 This content downloaded from 139.182.75.138 on Thu, 25 Apr 2019 06:51:48 UTC All use subject to https://about.jstor.org/terms Cultural Cognition in International Business Research articles from the combined MIR and JIBS pool comprise Publication Cluster One, the next 309 articles comprise Publication Cluster Two, and so forth. Reliability Axelrod (1976) advises interpreting reliability as a high degree of response con- sistency. We used Tinsley and Weiss’s (1975) measure of rater reliability – calcu- lated as the number of figures that each participant classified in the same way as a percentage of the total number of figures with both sums adjusted for the prob- ability that the participants agree simply by chance. Rater agreement was 82% among the five participants. Test-retest scores averaged 85% per participant. The intraclass correlations among the two independent counts of nodes and links – calculated with Formula ICC (2,k) per the recommendation of Shrout and Fleiss (1979) – was 0.90 for nodes and 0.86 for links. The raters resolved contrasting classifications and counts before analyzing the data. Analyzes We recorded the frequency of schematic representations classified as analog, prop- ositional, or composite. We also measured the structural features – connectedness, comprehensiveness, and complexity – of each representation. We then considered these measures in relation to our proxies of culture over time. We look at longi- tudinal effects as to the frequency of analog, propositional, and composite sché- mas by Publication Clusters. We analyze the nominal data with Student t-tests and ANO VA procedures. Post-hoc comparisons use Scheffe’s test. Results Authorship of the article, by national affiliation, from which we extracted a sche- matic representation included 15 British, 11 Canadian, 6 French, 12 German, 71 US, and 23 Cross-National observations. Small counts for several nations: e.g., 3 Swedish, 2 Norwegian, and 1 Austrian, Israeli, Korean, Chinese, and USSR, ex- cluded them from national-level analyses. Article authorship by region included 43 European, 77 North American, and 27 Cross-Regional observations. Article authorship by journal included 84 and 65 observations, respectively, from MIR and JIBS. We found few articles of exclusively Asian origin, whether by nation or region that included a schematic representation of an interpretative framework. Thomas et al. (1994) and Inkpen and Beamish (1994) note similar trends for sam- pling and authorship. mir vol. 40, 2000/3 279 This content downloaded from 139.182.75.138 on Thu, 25 Apr 2019 06:51:48 UTC All use subject to https://about.jstor.org/terms Daniel P. Sullivan/Gary R. Weaver Cross-Sectional Analysis Table 1 reports the average number and relative share of Figures classified by the Academic and External participants as an analog, composite, or propositional form per nation, region, journal, and disciplinary perspective. Overall, the participants classified the analog form without much indecision. Post-sort discussions found they noted properties (e.g., “easy-to-follow,” “linear flow,” and “single, straightfor- ward path”) that fit general reports (cf., Fiol 1994, Eden 1992, Bougon 1992). The participants noted that the absence of these features made classification less clear. In particular, the presence of non-recursive causality (i.e., without a strict, sequen- tial causal order) in a schematic representation suggested to some a different inter- pretive framework than in Figures that included unidirectional paths. Classification, though, seemed to tap common heuristics. The participants noted that if a Figure showed a formal start- and end-point and conveyed an orientation of general to spe- cific, they classified it as an analog form. Conversely, if the Figure did not pinpoint a start- or end-point but still suggested an orientation of general to specific or vice versa, then they classified it as a composite form. Finally, those Figures that out- lined multiple connections among multiple points – what Weick (1979) terms equiv- ocality – led the participants to classify it as a propositional form. The data indicate national and regional effects (Table 1). Canadian scholars typically encoded their interpretative frameworks in a composite form while their Table 1. The Form of Orienting Schema by Culture, Region, Journal, and Discipline N Analog Composite Propositional Absolute Relative Absolute Relative Absolute Relative Nation Canada 11 4 36% 7 64% 0 0 France 6 3 50% 0 0 3 50% Germany 12 6 50% 5 42% 1 8% UK 15 8 53% 3 20% 4 27% USA 71 36 51% 20 28% 15 21% Cross-National 23 17 74% 5 22% 1 4% Region Europe 43 23 53% 11 26% 9 21% US 77 39 51% 25 32% 13 17% Cross-Regional 27 19 70% 7 26% 1 0.04% Journal MIR 84 35 42% 29 35% 20 24% JIBS 65 40 62% 19 29% 6 9% Discipline 149 75 51% 48 32% 26 17% 280 mir vol. 40, 2000/3 This content downloaded from 139.182.75.138 on Thu, 25 Apr 2019 06:51:48 UTC All use subject to https://about.jstor.org/terms Cultural Cognition in International Business Research French counterparts relied on analog and propositional schémas. In contrast, Ger- man, British, and US researchers balanced the use of the three forms. Cross-Na- tional collaborations converged on one form; nearly 70% of this set used an ana- log representation to portray their model. The regional analyzes elaborate these effects. European and North America scholars showed equivalent inclination to use an analog, composite, or propositional form. Cross-Regional collaborations treated the analog form as the default and the propositional form as the exception. The data show divergence by disciplinary perspective. The distribution of orient- ing schémas in MIR shows a rough balance among analog, composite, and a prop- ositional schema whereas those in JIBS strongly favored the analog form to the particular deemphasis of the propositional form. Table 2 reports the comprehensiveness, connectedness, and complexity meas- ures partitioned per nation, region, and discipline. Scholars from different nations conceived structurally different schémas to represent their models. For instance, US scholars specified Figures with the most nodes (u = 8. 1 ; sd. = 5.98) while their British counterparts stipulated the most links (u = 13.4; sd. = 10.64). The data show that French and British scholars synthesized the most structurally complex interpretative frameworks while German, American, and Canadians converged at lower degrees of complexity. The Cross-National set had notably lower counts on comprehensiveness (u = 7.13, sd. = 3.38), connectedness (u = 7.62, sd. = 7.62) and complexity (u = 1.09; Table 2. The Structural Elements of Orienting Schema by Culture, Region, Journal, and Discipline N Comprehensiveness Connectedness Complexity ‘i sd fi sd fi sd Nation Canada 11 8.1 5.98 11 8.64 1.31 0.40 France 6 6.0 1.09 10.6 7.65 1.73 1.25 Germany 12 8.66 4.59 10.75 5.72 1.34 0.72 UK 15 9.66 5.06 13.4 6.41 1.51 0.71 USA 71 10.04 7.09 12.37 10.64 1.22 0.78 Cross-National 23 7.13 3.38 8.56 7.62 1.01 0.53 Region Europe 43 9 4.94 . 13.23 8.48 1.53 0.80 US 77 9.71 6.85 12.06 10.22 1.23 0.72 Cross-Regional 27 7.31 3.73 7.7 5.3 1.008 0.42 Journal MIR 84 9.49 5.98 8.1 6.1 1.31 0.85 JIBS 65 8.21 4.53 9.34 7.25 1.08 0.44 Discipline 149 9.02 5.76 11.61 8.93 1.29 0.72 mir vol. 40, 2000/3 281 This content downloaded from 139.182.75.138 on Thu, 25 Apr 2019 06:51:48 UTC All use subject to https://about.jstor.org/terms Daniel P. Sullivan/Gary R. Weaver sd. = 0.53). The marked differences between Cross-National and other nations prompted testing whether the collaboration process confounded these outcomes. Unlike observations in the exclusively national or regional sets, Cross-National and Cross-Regional observations always included at least two authors. Therefore, we assessed the mean scores on comprehensiveness, connectedness, and complex- ity between Figures drawn from single-authored versus multi-authored articles within each national and regional class. Student t-tests found no significant dif- ferences in terms of national or regional classes. Regional analyses elaborate these effects. Table 2 shows that European schol- ars specified models of higher complexity, North American researchers outlined more comprehensive models but with fewer connections, and Cross-Regional col- laborations specified models with the fewest concepts and fewest connections. Fi- nally, analysis by disciplinary perspective found that interpretative frameworks reported in MIR specified higher degrees of comprehensiveness and particularly higher degrees of connectedness and complexity than those reported in JIBS. We clarified these findings through Student t-tests and an ANO VA. Small cell counts for several nations precluded an ANO VA at the national level. Student t-tests by nation, though, found significant effects. Regarding connectedness, for example, England and Cross-National (p = 0.06) and US and Cross-National (p = 0.04) significantly differed; regarding complexity, France and US (p = 0.03), England and US(p = 0.08), and France and Cross-National (p = 0.02) significantly differed. The cell counts, normality of the distributed observations, and homogeneity of the variance for the dependent variables for “Regional” satisfied the model re- quirements of ANOVA. Therefore, we set Region as the independent variable and comprehensiveness, connectedness, and complexity as the dependent variables. This ANOVA model found no significant difference among European, North American, and Cross-Regional for comprehensiveness (F2, 145 = 1.85, p = 0.24) but significant differences for connectedness (F2f 145 = 2.95, p = 0.05) and com- plexity (F2, 145 = 4.41, p = 0.01). Post-hoc comparisons found significant differ- ences in mean scores between North American and Cross-Regional (p = 0.04) for comprehensiveness, between European and Cross-Regional (p = 0.02) for con- nectedness, and between European and North American (p = 0.02) and European and Cross-Regional (p = 0.006) for complexity. Finally, Student t-tests found sig- nificant differences between the interpretative frameworks reported in MIR and JIBS in terms of connectedness (p = 0.01) and complexity (p = 0.001) but none for comprehensiveness (p = 0.19). Longitudinal Analysis The small cell counts at the national level led to considering longitudinal trends for discipline and regional levels. Figures 1 and 2 show the absolute and relative 282 mir vol. 40, 2000/3 This content downloaded from 139.182.75.138 on Thu, 25 Apr 2019 06:51:48 UTC All use subject to https://about.jstor.org/terms Cultural Cognition in International Business Research Figure 1. Absolute Breakdown of International Business Models by Figure Form and Publication 35 Cluster, 1970 through 1997 25 | 20 Z 15 E / io s -^ 0 -J | – + – Analog Form – – M- – Composite Form – -A- – Propositional Form | Figure 2. Relative Breakdown of International Business Models by Figure Form and Publication 0.8 -i Cluster, 1970 through 1997 0.6 y – -^ 0.5 | 0.4 1 0.3 – * ^7>