Endorsing the notion that simple rules, ideas, and theories are better and explain as much or more than complex ones (e.g., Einstein’s famous equation E ¼ mc2), we wish to formulate theories with a minimal number of constructs that explain the behaviors and actions of entities or systems. In fact, the absolute minimum (simple) specification for a ‘‘theory’’ is two constructs, with a relationship specified between them, about something (i.e., entities or objects of study), as exemplified by Einstein’s equation for energy.
Our four proposed theories have these characteristics and are notions that have ‘‘deep roots,’’ as they are linked to traditional or classical ideas. We begin this discussion with a focus on the entities or levels of analysis.
Levels of Analysis
Levels of analysis issues and multiple-level approaches are becoming increasingly important in many areas of OB and closely related research (see
Dansereau & Yammarino, 1998a, 1998b, 2003, 2005, 2007; Yammarino &
Dansereau, 2002a, 2004, 2006). Various scholars (Dansereau et al., 1984, 1999;House, Rousseau, & Thomas-Hunt, 1995;Klein, Dansereau, & Hall, 1994;Rousseau, 1985;Yammarino et al., 2005) have noted the importance of clearly specifying the levels of analysis at which phenomena are expected to exist theoretically, and have stated that it is critical to ensure that the measurement of constructs and data-analytic techniques correspond to the asserted levels of analysis, so that inference drawing is neither misleading nor artifactual.
Levels of analysis are inherent in theoretical formulations. In some formulations, the levels of analysis are implicit or assumed. In other cases, levels of analysis are used to formulate the boundary conditions under which a theory is expected to hold. In still other instances, theories, propositions, and hypotheses explicitly incorporate levels of analysis as an integral component of the formulation. Understanding how and if levels are specified permits an examination of the potential for or degree of prevalence of theoretical misspecification. Moreover, identification of relevant levels of analysis issues may help account for mixed, inconsistent, and contradictory findings in prior research. Without explicit incorporation of levels of analysis issues, incomplete understanding of a construct or phenomenon may lead to faulty measures, inappropriate data-analytic techniques, and erroneous conclusions.
Theoretical revolutions in science often emerge when other levels of analysis are considered. For example, a revolution in biology occurred when some theorists suggested, and subsequently demonstrated, that evolution can occur at a level of analysis higher than the organism level. Likewise, a well-known revolution in physics arose when some theorists asserted, and subsequently demonstrated, that quantum mechanics operate at a level of analysis lower than the atomic level. In this same way, OB theory building can advance when we include lower and higher levels of analysis in theory development and hypothesis generation.
Levels of analysis are the entities or objects of study. In the current work, we are interested in human beings in work organizations. Entities are typically arranged in hierarchical order such that higher levels (e.g., groups) include lower levels (e.g., persons), and lower levels are embedded in higher levels (seeDansereau et al., 1984; Yammarino, 1996;Yammarino & Bass, 1991). In the various areas of OB research, four key levels of analysis of human beings are relevant: individuals or persons (independent human beings), dyads (two-person groups and interpersonal relationships), groups (workgroups and teams), and organizations (collectives larger than groups
and groups of groups) (see Dansereau et al., 1984; Yammarino, 1996;
Yammarino & Bass, 1991).
First, human beings in organizations can be viewed as individuals or persons, independent of one another. In this case, we can focus on an employee, a manager, a leader, or a follower/subordinate, or how these individuals differ from one another. Individual differences are of interest here.
Second, human beings in organizations can be viewed as dyads, or two individuals who are interdependent on a one-to-one basis. A dyad is a special case of groups – that is, a two-person group. In this case, we can focus on superior–subordinate dyads, leader–follower dyads, peer–peer dyads, coworker–coworker dyads, or interpersonal relationships, indepen-dent of the formal workgroup.
Third, human beings in organizations can be viewed as groups or teams.
While there are some potential differences between groups and teams, we view them similarly here – as a collection of individuals who are interdependent and interact on a face-to-face or ‘‘virtual’’ (non-colocated) basis with one another. Formal workgroups or teams generally consist of a leader or a manager and his or her immediate followers or direct reports.
Fourth, human beings in organizations can be viewed as collectives. In this case, the focus is on clusterings of individuals that are larger than groups and that are interdependent based on a hierarchical structuring or a set of common or shared expectations. Collectives include groups of groups, departments, functional areas, strategic business units, and organizations.
They often do not involve direct interaction among people (as in groups), but rather are held together by echelons or hierarchies.
These four levels of analysis – person, dyad, group, and collective – represent different perspectives on the human beings who make up organizations. In this sense, they can be thought of as different lenses through which human beings can be observed. A key characteristic of these levels is their embeddedness; for example, two persons make up a dyad, multiple persons make up a group, multiple dyads make up a larger group, and multiple groups make up a collective. In other words, as one views human being from increasingly higher levels of analysis, the number of entities decreases (e.g., there are fewer collectives than groups in an organization), and the size of the entities increases (e.g., collectives include a larger number of human beings than do groups).
Wholes and Parts
In our approach, there are four alternatives to consider for each level of analysis (seeTable 2). Two of these alternatives are plausible views of the
focal entities (parts and wholes as units of analysis), and two of them indicate that focal entities are not relevant but that other entities may be plausible (equivocal and inexplicable).
We distinguish conceptually between two different views of any level of analysis (also seeLerner, 1963). A wholes view is defined as a focus between entities but not within them; differences between entities are viewed as valid, and differences within entities are viewed as error (random). This perspective can be described as a between-units case (Glick & Roberts, 1984; Pedhazur, 1982). In this instance, members of a unit are homo-geneous, the whole unit is of importance, and relationships among members of units with respect to constructs of a theory are positive. Relationships among theoretical constructs are a function of differences between units.
A parts view is defined as a focus within entities but not between them;
differences within entities are valid, and differences between entities are erroneous. This perspective can be described as a within-units case (Glick &
Roberts, 1984;Pedhazur, 1982) or a frog pond effect (Firebaugh, 1980). In this instance, members of a unit are heterogeneous, a member’s position relative to other members is of importance, and relationships among members of units with respect to constructs of a theory are negative. Relationships among theoretical constructs are a function of differences within units.
These two views – wholes and parts – are conceptually different ways to indicate that a particular level of analysis is relevant for understanding constructs and variables of interest. In addition to permitting effects at a particular level of analysis, various authors indicate that effects may not be evidenced at that level (Lerner, 1963;Miller, 1978;Pedhazur, 1982). Thus the focal level is considered not relevant, and other levels must be considered.
In one case, there is a focus both between and within entities at a focal level. Determining whether a wholes or parts view is occurring is difficult because both between- and within-entities differences are valid. Thus the
Table 2. Summary of Single-Level Formulations.
Alternative Views of Entities
Members of Units
Associations among Unit Members
Between-Entities Differences
Within-Entities Differences
Wholes Homogeneous Positive Systematic Error
Parts Heterogeneous Negative Error Systematic
Equivocal Independent Independent Systematic Systematic
Inexplicable Not relevant Not relevant Error Error
focal level of analysis does not clarify our understanding of the constructs and variables of interest. Consequently, other levels must be considered.
If the assumption is made that only one level of analysis can be considered, then seemingly both conditions (wholes and parts) are occurring. Because there are always other levels of analysis to consider (Miller, 1978), however, this condition must be viewed as equivocal – between- and within-entities differences are equally likely. The more parsimonious conclusion is that neither wholes nor parts views at the focal level are appropriate (Dansereau et al., 1984; Yammarino & Markham, 1992). In this instance, members of a unit are independent, members are free of the unit’s influence, and relationships among members of units with respect to constructs of a theory are independent. In short, relationships among theoretical constructs are a function of differences between members (e.g., persons) independent of higher-level units (e.g., groups).
Another possibility – namely, error or lack of focus between and within entities – is an inexplicable or traditional null view of a focal level. In this case, the focal level also is not relevant for understanding the theoretical constructs of interest; instead, other levels of analysis should be specified conceptually.
In summary:
Wholes are homogeneous entities that display similarity among members, where between-entities differences are systematic and within-entities differences are error.
Parts are heterogeneous entities that display complementarity among members, where within-entities differences are systematic and between-entities differences are error.
Equivocal reflects independence among members, where between- and within-entities differences are systematic and other entities should be considered.
Inexplicable indicates a null case where between- and within-entities differences are error and other entities should be explored.
Given these alternatives, it becomes a matter of selecting among them based on theory and data analysis.
Multiple Levels
Beyond these single levels of analysis (i.e., individuals, dyads, groups, or collectives viewed separately), a key issue is that of multiple levels of analysis.
In other words, levels can be viewed in combination or simultaneously
(seeTable 3). In these cases, we are concerned with multi-level or cross-level effects, as well as with mixed determinants and mixed-level effects (for details and a review, see Dansereau et al., 1984; Dansereau & Yammarino, 2000;
Klein et al., 1994;Rousseau, 1985).
For us, multi-level or meso formulations (theories, propositions, and hypotheses) are explanations linking variables, which operate at different levels of analysis (e.g., person-level X is positively related to group-level Y) (see Behling, 1978). Rousseau states that such theories specify ‘‘relationships between independent and dependent variables at different levels’’ (1985, p. 20).
(Rousseau calls these cross-level – not multi-level or meso – formulations.) Models of this type provide among-level explanations because they link variables in terms of multiple levels of analysis. Included here are mixed-effects models, in which a single variable of interest may have effects at multiple levels with multiple criteria of interest, as well as mixed-determinants models, in which multiple predictor variables at various levels of analysis affect a single criterion at a single level of analysis.
For us, cross-level formulations (theories, propositions, and hypotheses) are statements about relationships among variables that are likely to hold equally well at a number of levels of analysis (e.g., X and Y are positively related for individuals and for groups) (seeBehling, 1978;Dansereau et al., 1984; Miller, 1978). Rousseau notes that such cross-level formulations
‘‘specify patterns of relationships replicated across levels of analysis’’ (1985, p. 22). (Rousseau, however, calls these multi-level – not cross-level – formulations.) Models of this type are uniquely powerful and parsimonious (simple) because the same effect is manifested at more than one level of analysis (e.g., E ¼ mc2, which holds at multiple levels of analysis).
Table 3. Summary of Multiple-Level Formulations.
Multiple-Level Formulation Lower-Level View Higher-Level View
Cross-level wholes Wholes Wholes
Cross-level parts Wholes Parts
Level-specific wholes Wholes Equivocal
Level-specific parts Parts Inexplicable
Emergent wholes Equivocal Wholes
Emergent parts Equivocal Parts
Equivocal Equivocal Equivocal
Inexplicable Inexplicable Inexplicable
Source:SeeDansereau et al. (1984, p. 186)for eight additional (null) alternatives and their interpretation.
Assuming only one level of analysis in a study, or choosing only one level without consideration of other levels, can either mask effects or indicate an effect when none truly exists (Lerner, 1963;Miller, 1978; Pedhazur, 1982;
Roberts, Hulin, & Rousseau, 1978). These issues are especially important when individuals are embedded within larger units such as dyads, groups, and collectives in organizations. Thus considering only one level of analysis is insufficient. Instead, multiple levels should be identified in combination.
Regarding the particular formulations in Table 3, relationships among constructs may be hypothesized to hold at a lower (e.g., person) level but not at a higher (e.g., group) level. These relationships are discussed as a discontinuity thesis (Miller, 1978), as level-specific formulations (Dansereau et al., 1984; Miller, 1978), or empirically as disaggregated, individual, or level-specific effects (Pedhazur, 1982; Robinson, 1950). In these cases, the higher level of analysis is not relevant for understanding the theoretical constructs.
In contrast, relationships among constructs may not be asserted at a lower level but may be hypothesized to manifest themselves at a higher level of analysis. These relationships also are discussed as a type of discontinuity thesis (Miller, 1978), as emergent formulations that hold at a higher (e.g., group) level after not being asserted or found to hold at a lower (e.g., person) level (Dansereau et al., 1984; Miller, 1978), empirically as a higher-level effects that do not disaggregate, or as emergent effects (Miller, 1978;
Robinson, 1950). In these cases, the lower level of analysis is not relevant for understanding the theoretical constructs.
Thus, in the case of level-specific and emergent formulations, even though a single level of analysis is of primary concern, other levels are considered but defined as not relevant. Alternatively, relationships among constructs may be hypothesized to hold at higher (e.g., collective) and lower (e.g., group) levels of analysis. These relationships are discussed as a homology thesis (Miller, 1978) or empirically as aggregated or ecological effects (Glick & Roberts, 1984; Pedhazur, 1982; Robinson, 1950). As noted pre-viously, they are of two types:
Cross-level explanations (Behling, 1978; Miller, 1978; Dansereau et al., 1984) specify relationships among theoretical constructs that hold equally well, or are replicated across, higher and lower levels.
Multi-level or meso explanations (Behling, 1978) assert relationships among theoretical constructs (independent and dependent variables) that operate at different levels of analysis (e.g., individual-level X is related to group-level Y).
We focus on multiple levels of analysis simultaneously and in conjunction with wholes and parts views of levels. Using such specifications for identifying multi-level views, eight cases can be developed.
First, wholes at a lower level can aggregate or manifest themselves as wholes at a higher level. This cross-level wholes formulation means that members are homogeneous with respect to the constructs of interest in all entities (e.g., groups and collectives) at both levels of analysis, but the entities (e.g., groups and collectives) differ from one another.
Second, wholes at a lower level can aggregate or manifest themselves as parts at a higher level. This cross-level parts formulation means that members are homogeneous with respect to the constructs of interest in all the lower-level entities (e.g., groups), and that these differ from one another.
In contrast, in all higher-level entities (e.g., collectives), there is hetero-geneity as members within the entities differ from one another.
Third, wholes at a lower level may not aggregate or manifest themselves at a higher level (equivocal). This level-specific wholes formulation means that members are homogeneous with respect to the constructs of interest in all lower-level entities (e.g., groups), but that higher-level entities (e.g., collectives) are not relevant.
Fourth, parts at a lower level may not aggregate or manifest themselves at a higher level (inexplicable). This level-specific parts formulation means that members are heterogeneous with respect to the constructs of interest in all lower-level entities (e.g., groups), but that higher-level entities (e.g., collectives) are not relevant.
Fifth, for an emergent wholes formulation, constructs are expected to hold at a higher (e.g., group) level where members are homogeneous with respect to the constructs after not having been expected or observed at a lower level (equivocal).
Sixth, for an emergent parts formulation, constructs are expected to hold at a higher (e.g., group) level where members are heterogeneous with respect to the constructs after not having been expected or observed at a lower level (equivocal).
Finally, multi-level formulations can be specified as equivocal (case 7) or inexplicable (case 8) where both higher and lower levels are asserted to be equivocal or inexplicable, respectively, and not relevant for the constructs of interest. In these two cases, additional (other) levels of analysis must be considered for understanding the phenomena of interest.
These eight formulations are presented in Table 3 for sake of comple-teness, but only some are relevant here to specify our four simple theories.
Having outlined an approach for understanding entities in terms of single and multiple levels of analysis, it is now possible to consider variables and constructs and their relationships in conjunction with the entities to specify four simple, yet comprehensive theories for a new kind of OB. As noted earlier, each theory consists of two variables, an association between them, and a specification of the levels of analysis at which they are expected to operate (i.e., entities for which they are asserted to hold). These ‘‘little ideas that can be well tested’’ are simple ‘‘rules’’ [inWolfram’s (2002)terms] that can account for or explain a variety of behaviors and actions in OB.