Reviews of research on the attitude–behavior relationship and their implications for future environmental education research

Abstract

Since the early days of the field, attitude–behavior (A–B) relationships have received sustained attention in the evaluation and researching of environmental education (EE). This level of interest extends beyond the field’s scope though, in part due to a certain reliance on theoretical and empirical assertions which claim that attitudes serve as a strong precursor and/or predictor of behavior. In this paper, we consider reviews of studies on the A–B relationship in EE and other fields that routinely challenge such assumptions, leading to the questioning of corresponding foci and commitments for research, evaluation, practice, and development. With key findings from these studies to hand, we highlight several insights that may be useful for dispelling some of the folklore about what matters in, and what can be argued with, A–B studies, if we are to develop the design, conduct, reporting, and critique of studies and practices of EE reliant on such relationships.

KEYWORDS:

• Attitudes and behavior
• research design principles
• research findings
• research claims
• quality use of evidence

Introduction

Two of the more persistent questions within the field of environmental education (EE) and its research pertain to: (a) the relationship between attitudes (A) and behaviors (B), and (b) the value of educators and researchers focusing their efforts on understanding this relationship (Reid and Dillon 2016a).

On the one hand, many of the goals of the field frequently cite the need to change both attitudes and behaviors, argued from a range of perspectives and positions on what EE is (really) for. On the other, there are questions about what EE is (actually) able to achieve, through ideal, innovative, current, or past forms of its provision.

Put starkly, on both sides, EE is understood as a tool that variously enables others to become something other than they were, are or might become, which in this case, usually means using theory and evidence to argue for deepening or reworking the relationship between particular attitudes and behaviors, and/or adopting or promoting more environmentally responsible ones and patternings than those in existence before.

Unsurprisingly, hundreds of assessment, evaluation, and research studies in EE have explored and assessed the status and significance of A-B relationships in theory and practice, including some of the most cited and discussed papers and reviews in this field (Iozzi 1981, 1984; Rickinson 2001; Volk and McBeth 1997).

Yet often overlooked in readings and critiques of reviews of such programs and studies, is the attention authors give to the nature and strength of an attitude–behavior (A–B) relationship. Some insights into such relationships and their features have been presented in meta-analyses, literature reviews and syntheses of research relevant to EE (Bamberg and Moser 2007; Hines, Hungerford and Tomera 1987; Zelezny 1999; Zint 2012).

The purposes of this paper are to refocus attention on this ‘blindspot’ (Reid 2019) by: (1) summarizing results presented in wider reviews of A–B research; (2) exploring arguments and concerns raised in/by these reviews; and (3) begin to explore the implications of these reviews, including for how they might inform future practices and studies of EE designed to address such relationships.

Before undertaking this though, it is equally appropriate to identify what is not possible here, because numerous topics pertaining to A–B relationships lie beyond the paper’s purpose and scope.

First, the views expressed herein are not based on our own large-scale of assessment, research, and evaluation studies in the field that investigated A–B relationships; rather, significant results pertinent to A–B relationships in prior reviews such as those cited above will be noted and discussed.

Second, although arguments, concerns, and implications will be introduced, we do not provide an in-depth analysis of these for research, theory, policy, and/or practice in this field, noting too that various analyses of these are already available in the literature (Kollmuss and Agyeman 2002; Simmons 1991, 1995; Wals and Dillon 2012; Zint 2012).

Third, this paper is not written to review the wide range of theories of behavior and behavior change, as that too has been done elsewhere (Darnton 2008; Heimlich and Ardoin 2008; Jackson 2005). Consequently, it is beyond the scope of this paper to analyze the implications of these results of reviews of research for those theories.

Fourth, although we do comment on A–B relationships largely within the context of the Knowledge–Attitude–Behavior (K–A–B) model, there is no attempt here to review and Critiquing K‐A relationship either, for example, articulating or critiquing key assumptions, presenting results from reviews of research on K–A relationships, or discussing implications (we invite others to do that).

Fifth, we do not use this paper to analyze the relationship of attitude to other psychological constructs that have some documented relationship to behavior, for example, environmental sensitivity, place attachment, and associated significant life experiences; self-efficacy and locus of control; personal and social norms; and intention; but of these, intention will be discussed, given its demonstrable import for A–B relationships. Again, we invite others to consider these other relationships.

Finally, and more positively, our hope is that in light of these delimitations, readers will find ample opportunity, and therefore are encouraged, to review pertinent literature and explore the implications of the research results presented in this paper to the depth that they deserve, including for their own studies of EE, and its relation to the wider arguments, key findings and possible implications noted below.

Definition of key terms

To continue, for the key terms at work here, a brief accounting is needed of attitude and behavior. Historically, much of the scholarly work studying both terms and their relations has sprung from ideas that can be traced to psychology (e.g. social, applied, and educational).

Positions therein, such as those typified by Allport (1935), characterize attitude as a ‘mental and neural state of readiness’ and argue that it is ‘the most distinctive and indispensable concept [construct] in contemporary social psychology’ (p. 798). Indeed, over the last 50 years, many theorists and researchers both within and beyond the field of EE have come to view attitudes largely in such terms, where it is primarily evidenced in relation to a person’s evaluative dispositions and judgments about an ‘object’ (e.g. a being, thing, event, idea, issue, or action) that are derived, at least in part, from their experience or situation.

A further commonplace is that attitudes have cognitive components (e.g. beliefs or knowledge), affective components (e.g. feelings or emotions), and behavioral components (e.g. a predisposition that may influence whether and how to act). The behavioral component is central to our concerns here, given such arguments as: ‘The ability of attitudes to predict behavioral intentions or overt behavior continues to be a major focus of theory and research’ (Ajzen 2001, p. 42). Thus for Ajzen (2001), the extent to which attitudes predict or influence behavior depends on a number of factors, including the nature and strength of underlying beliefs and emotions, the co-existence of positive and negative dispositions, and situational factors.

Equally, for the purposes of this paper, Stern’s (2000) two-part definition of environmentally significant behavior is worth recalling when we try to tease out the role(s) of, and possible interactional effects of senses of ambivalence, hedonism, impulsivity/self-control/delayed gratification, social norms, educational processes, immediacy of horizons and timelines of anticipatory behavior, opportunities for sustained sense of connectedness to nature, and so forth. First, environmentally significant behavior may be defined ‘from the actor’s standpoint as behavior that is undertaken with the intention to change (normally, to benefit) the environment’, thereby highlighting ‘intent as an independent cause of behavior’ (p. 408). However, as Stern indicates, this is insufficient because ‘intent may fail to result in an environmental impact’ (p. 408). Thus, Stern differentiates between a person’s intent to act and behaviors based on their intent. For this reason, Stern also differentiates a pro-environmental behavior as that which ‘may be defined by its impact’ or consequences (p. 408), including direct and broader influences on biodiversity, natural resources, and pollution and waste (e.g. direct: household waste disposal; broader: national and international policies on development, pricing, and taxation). Furthermore, we must also note behaviors fostered or shaped through EE may still have a positive impact on the environment, despite the absence of any direct intention by an individual (or feature of the EE programming) to do so immediately (Gould et al. 2019; Heimlich et al. 2012).

Why this focus in EE and its research?

It is relatively easy to trace the first flourishing of interest in attitudes, behavior, and their relationship, in the EE literature (Disinger 1983; Hart 1981; Harvey 1977; Schmeider, 1977; Stapp 1974).

An early paper, influential in both the U.S.A. and emerging international literature, identified ‘attitudes of concern for the quality of the biophysical environment’ as one of major objectives of EE: ‘it implies a combination of factual knowledge and motivating emotional concern which result in a tendency to act’ (Stapp et al. 1969, p. 31).

Later, both Attitudes and Participation were presented as categories of objectives at the international Belgrade Workshop on EE (UNESCO 1977), and then revised and endorsed at the subsequent Tbilisi Intergovernmental Conference (UNESCO 1978). The Tbilisi Recommendations have been reaffirmed at subsequent UN-sponsored EE meetings and conferences, as well as adopted and adapted by countries around the world (see Reid and Dillon 2016b, for key examples).

Within the U.S.A., many national and state EE frameworks also gave prominent attention to attitudes and behavior (Simmons 1995), as have environmental literacy frameworks (Hungerford and Volk 1990; Hollweg et al. 2011; McBride et al. 2013; Roth 1992; Wilke 1995). These and other sources are indicative of the broad and enduring level of attention that has been given to both attitudes and behavior in EE.

Nonetheless, some have been skeptical, even critical, of the focus on and use of attitudes as a meaningful objective and outcome in EE. An early paper in The Journal of Environmental Education, illustrates such concerns. Hendee (1972, p. 20) suggested that ‘[e]nviromental educators should cease their preoccupation with attitudes as a criterion to evaluate their efforts,’ because there are significant problems both with ‘the validity and feasibility’ of tests to measure them, and then using these tests for evaluating the effects of an educational program. Furthermore, Hendee opined, ‘[a]ttitudes respectful of nature are needed but must accrue as a result of relevant and complete information’ (ibid.), that is, arise from efforts that give precedence to developing personally relevant knowledge to various groups of people.

Yet despite such cautions and appeals to address what have been regarded as aspects of the wishful thinking, myths and folklore of the field¹, uncritical attention to attitudes has persisted. In a review of experimental studies 25 years later, Volk and McBeth (1997) found that ‘the most frequently measured variables were attitudinal in nature, with 76% of the studies’ (p. 42). Again, careful reading and reflection were invited by the authors; while they found that ‘positive changes in attitude were observed in 48% of the studies which measured that variable’ (p. 43), they also invited colleagues to be mindful of the tallies of non-significant, mixed or questionable results for any variable included in their review (table 19, p. 42, and commentary thereon).

Lucas (1981) raised similar concerns about what might be found in, and interpreted from, the evidence base at that time. Referencing large-scale surveys of secondary students in Australia, England, and the U.S.A., he concluded that ‘[i]n all three countries the pupils’ environmental attitudes tended to be positive, except when the object of concern impinges on their own lives’ (p. 35). Lucas interpreted these data as showing: 'the difficulty of reconciling different attitudes (toward full employment, environmental conservation, and personal freedoms, for example), [which] make it very difficult to rely on measures of environmental attitudes as an indicator of the success of environmental education programs … It is possible that well developed environmental attitudes may produce the opposite effect to the one desired, even in knowledgeable people'. (p. 35)

For Lucas, such survey findings spoke trouble to ‘the belief that education for the environment entails focusing on the development of attitudes as the major goal [which] is probably the most pervasive position in the literature’ (1981, p. 35). They cast doubt on, for example, such positions as those noted by Ramsey and Rickson (1976, p. 10) who had suggested that ‘[i]ncreased knowledge leads to favorable attitudes … which in turn lead to action promoting better environmental quality,’ and went unheeded in related projects, such as that of Birch and Schwaab (1983), who had offered a similar viewpoint:

the assumption must be made that informed attitudes will lead to subsequent water conservation behavior … Further research should attempt to offer empirical evidence that knowledge and attitude gains resulting from the water conservation unit will influence a student’s behavior regarding efficient water use. (p. 30)

Such quotations also serve to illustrate what was, at that time, a prevalent (if not problematical) line of thinking that associated the provision of educational experiences of particular kinds with generating or imbuing certain knowledge and attitudes, and attitudes with corresponding behaviors (see Colwell, 1976). In the 1990s, this often became known as the Knowledge-Attitude- Behavior or K-A-B Model (Hungerford and Volk 1990; Kollmuss and Agyeman 2002; Simmons 1991).

Yet rather than assume this model has become ‘old hat,’ it is important to recognize that aspects of it echo through to the field’s most recent work, even to ‘new hats’ that have been recommended since then. For example, key elements of this model are apparent in particular programs In EE that are designed to influence knowledge, attitudes and behaviors, as well as surveys designed to assess knowledge, attitudes, and behaviors, from a wide variety range of educational and environmental perspectives. Whether it is to foster contemporary forms of environmental consciousness, social justice or climate action of particular kinds through EE—be that from a positivist to post-positivist, constructivist to post-constructivist, or humanistic to post-humanistic perspective—researchers and educators continue to advocate particular experiences to change and link knowledge, attitudes and/or behaviors, consciously and unconsciously (Wals and Dillon 2012).

With this in mind, and our brief sketch of the support for and concerns about role of attitudes and the A-B relationship in EE complete, we now turn to what meta-analyses and the wider research literature might have to say about such relationships.

Evidence on the A-B relationship in reviews of EE research

Unfortunately, evidence overwhelmingly indicates that practices based on a K–A–B model are not as well founded on research evidence as some of the quotations above might imply. For example, one of the first meta-analyses of EE research on factors associated with behavior found only a moderate relationship of knowledge to behavior (r = .299), and of A-B (r = .347) (Hines et al. 1987, p. 3). They also found that other psychological constructs, notably verbal commitment (intention) and locus of control, had stronger relationships to behavior than did attitude. Later, Bamberg and Moser (2007, p. 20) drew similar conclusions, stating that:

for the association between these four psych-social variables [attitudes, moral norms or personal responsibility, self-efficacy/locus of control, and intention] and pro-environmental behavior our meta-analysis results in pooled mean correlations very similar to those reported by the Hines et al. meta-analysis 20 years ago.

These meta-analyses, along with other sources of evidence and critique, have led some in EE to question the value and use of any K–A–B model, including its implications for theory, research, policy and practice in EE (see a range of positions on this, in Hungerford and Volk 1990; Kollmuss and Agyeman 2002; Marcinkowski 2004). The studies also serve to highlight the importance of continuing to negotiate the value and status of any positing of an A–B relationship since then, whether that be in the context of advocating broader theoretical models in general, or in relation to the varieties, qualities and quality uses of evidence (Rickinson et al. 2017) for specific educational frameworks, engagements or experiences in EE. But before we consider aspects of that, we must consider wider perspectives on this relationship.

Additional evidence on the A–B relationship

For quite some time, developing a critical perspective to assess the evidence base on any purported relationship—direct to indirect, strong or weak, simple to complex—between an attitude and behavior has been of interest in many academic fields, including on a wide range of issues well beyond those of interest to EE (Ajzen and Fishbein 1977; Fishbein and Ajzen 1975; Wicker 1969, 1971).

In 2015, Marcinkowski conducted a literature search in the English language archives to locate significant meta-analyses of research on A–B relationships presented and/or published since the early 1990s (i.e. those analyses that have been positively cited, checked, replicated or extended). Consistent with the purposes for this paper, we summarize key features of each one below, to illuminate some of the methodological and substantive findings about A–B relationships, and the degrees of optimism and pessimism that might be had, in relation to the evidence base and claims concerning magnitude, mediation and predictive power.

To begin, we note the work of Harvard University’s Stephen Kraus (1990). Kraus presented a meta-analysis at the 1990 Annual Convention of the American Psychological Association, stating in his abstract that:

The relationship between attitudes and behavior has been a topic of considerable debate. Accordingly, this paper reports a meta-analysis of 83 attitude-behavior studies. The analysis suggests that attitudes significantly predict future behavior (combined p < .000000000001). The average attitude-behavior correlation (ABC) is r = .38. Methodologically, ABCs are higher for studies using (a) self-reported measures of behavior, (b) primarily non-students as subjects, or (c) attitude and behavior measures of corresponding levels of specificity. Substantively, ABCs tend to be higher when (a) the attitude is formed by direct experience, (b) the attitude is held with certainty, (c) the subject is a low self-monitor, or (d) the situation increases self-focused attention. (p.1, emphases added)

Several years later, Kraus (1995) polished and published his well-received review. The revised abstract noted:

The relationship between attitudes and behavior has been the topic of considerable debate. This article reports a meta-analysis of 88 attitude-behavior studies that reveals that attitudes significantly and substantially predict future behavior (mean r = .38; combined p <. 000000000001). Relatively large and significant moderating effects were found for the attitudinal variables of attitude certainty, stability, accessibility, affective-cognitive consistency, and direct experience (mean q = .39). A smaller but significant moderating effect was found for self-monitoring (mean q = .29). Methodological factors associated with high attitude-behavior correlations included self-report measures of behavior (q =. 22), the use of nonstudents as subjects (q =. 17), and corresponding levels of specificity in the attitude and behavior measures (mean q = .47). The practical magnitude of attitude-behavior correlations is considered, as are the future directions of attitude-behavior research. (p. 58, emphases added)

In the interval between Kraus’ conference presentation (1990) and publication (1995), Kim and Hunter (1993a, 1993b) also conducted meta-analyses of studies on the A–B relationship. They controlled for some of the methodological factors identified by Kraus, including factors linked to behavioral intention, self-report and overt behavior, that may confound, moderate or mediate the outcomes. Their abstract (Kim and Hunter 1993a) notes:

The difficulty of finding a relationship between attitudes and behavior is one of the greatest controversies in recent social science research. The purpose of this study was to determine whether attitudinal relevance substantially affects the magnitude of the correlation between attitudes and behavior, and whether the effects are content free. Using meta-analysis, we integrated findings from 138 attitude-behavior correlations with a total sample size of 90,908. The behaviors we studied ranged over 19 different categories and a variety of miscellaneous topics. Our results showed a strong overall attitude–behavior relationship (r = .79) when methodological artifacts were eliminated. As predicted, the higher the attitudinal relevance, the stronger the relationship between attitudes and behavior. This effect held true across diverse content domains. (p. 101, emphases added)

However, Kim and Hunter did not stop there. Rather, they extended their work (Kim and Hunter 1993b), arguing in a follow-up analysis that:

The trend in A–B research, however, is to conceive of behavioral intentions (BI) as a mediator between attitudes (A) and behaviors (B). In this study, it is hypothesized that (a) A–BI correlation would be higher than A–B correlation, (b) BI–B correlation would be higher than A–B correlation, (c) A–BI correlation would be higher than BI-B correlation, (d) the variation in BI-B correlations would be greater than that of A–BI, and (e) attitudinal relevance would affect the magnitude of the A–BI correlation. A series of meta-analyses, integrating the findings of 92 A–BI correlations (N = 16,785) and 47 B–BI correlations (N = 10,203) that deal with 19 specified categories and a variety of miscellaneous topics was performed. The results were consistent with all five hypotheses. (p. 331, emphases added)

A third meta-analysis of the A-B research literature was reported by Wallace et al. (2005), based on a very large collection of psychological studies. Among other things, they flagged the significance of motivation (sources, forms, and effects) as a factor, while in summary, their abstract noted:

A meta-analysis of 797 studies and 1001 effect sizes tested a theoretical hypothesis that situational constraints, such as perceived social pressure and perceived difficulty, weaken the relationship between attitudes and behavior. This hypothesis was confirmed for attitudes toward performing behaviors and for attitudes toward issues and social groups. Meta-analytic estimates of attitude–behavior correlations served to quantify these moderating effects. The present results indicated that the mean attitude–behavior correlation was .41 when people experienced a mean level of social pressure to perform a behavior of mean difficulty. The mean correlation was .30 when people experienced social pressure 1 standard deviation above the mean to perform a behavior that was 1 standard deviation more difficult than the mean. The results suggest a need for increased attention to the ‘behavior’ side of the attitude-behavior equation. Attitudes predict some behaviors better than others. (p. 214, emphases added)

To end our brief tour of some of the most significant meta-analyses of studies in the A-B research literature, we note the findings reported by Glassman and Albarracin (2006). These further underscore the significance of methodological factors and their possible interplay with substantive findings:

A meta-analysis (k of conditions = 128; N = 4598) examined the influence of factors present at the time an attitude is formed on the degree to which this attitude guides future behavior. The findings indicated that attitudes correlated with a future behavior more strongly when they were easy to recall (accessible) and stable over time. Because of increased accessibility, attitudes more strongly predicted future behavior when participants had direct experience with the attitude object and reported their attitudes frequently. Because of the resulting attitude stability, the attitude-behavior association was strongest when attitudes were confident, when participants formed their attitude on the basis of behavior-relevant information, and when they received or were induced to think about one- rather than two-sided information about the attitude object. (p. 778, emphases added)

From ‘so what’ to ‘now what’?

While studies of A–B relationships in EE may be compared against such findings and considerations, for our purposes here, we now highlight a range of points that emerge from these meta-analyses, focusing on matters of how research is designed, reported and interpreted in EE. These points are crystallized as follows:

1. The strength that can be attributed to research claims matters, as do the limitations of such claims. A key finding apparent in the meta-analyses by Kraus (1995) and Wallace et al. (2005) is that while an A-B relationship may be statistically significant, it is often of relatively moderate strength only (Cohen, 1988). Their findings are consistent with results of meta-analyses of EE studies that analyzed A-B relationships reported by Hines et al. (1987), and Bamberg and Moser (2007). The noteworthy exception to this were the findings reported by Kim and Hunter (1993a), who found a stronger A-B relationship, but only after controlling for several of the methodological factors that influence the strength of this relationship (see also point #2 on relevance, below).

2. The valency and relevance of referents can’t be ignored for findings to matter to researchers and others. A key finding in the meta-analysis by Wallace et al. (2005) is that the relative strength of an A–B relationship varies with different attitudinal objects or referents, such as attitudes towards various social groups, issues, and future behaviors. This matter of situatedness resurfaces in discussion of moderator variables and dynamics too, such as in relation to #4. But it is especially relevant to thinking and commentary on the goals and expectations of maintaining or developing diverse traditions and approaches to environmental education in light of the evidence base, as in parsing the differences between, interactions, and immediate- to longer-term effects of pursuing education about/through/for the environment (see also Rickinson 2006).

3. The degrees of generality and specificity of measures supporting evidence-based claims matter too. A third key point concerning the relative strength of an A–B relationship is the influence of a number of methodological factors, and this should also be acknowledged when scoping or claiming transferability, including as to any positive and/or negative spillover effects (Nilsson et al. 2017). The theory and research literature pertaining to this reflects an awareness of the influence of certain methodological factors on the A-B relationship that date back more than half a century (e.g. Ajzen and Fishbein 1977; Wicker 1969); for example:

   Conceptual and methodological analysis has focused on two principles: the principle of aggregation, founded on the psychometric fact that single items on any test are unreliable; and the principle of compatibility, which holds that substantial correlations between attitudinal and behavioral measures will only be found if these constructs are assessed at the same level of generality (or specificity). (Manstead 2001, p. 909; emphases added)

4. Moderators always matter. A fourth point is that the relative strength of an A-B relationship can be moderated by a number of factors that may be operating in or beyond the research setting, particularly among study participants. For environmental educators and researchers, we find points by Kraus (1990) worth reiterating: that these factors often include the existence or effects of whether 'the attitude is formed by direct experience, the attitude is held with certainty, the subject is a low-self-monitor, or the situation increases self-focused attention' (p.1). Similarly, Wallace et al. (2005) highlight the significance of personal moderators (e.g. perceived difficulty) and situational moderators (e.g. social pressures experienced and expressed through the interaction of subjective and social norms or expectations). Others have noted the role of affect (e.g. the emotions we feel before, during or after we have engaged in a behavior—e.g. Fazio 1986), perceived behavioral control, and the normative expectations of important others, such as peers, parents, and educators (Ajzen 2012)².

   In light of these considerations, as Steg and Vick (2009) argue, researchers should be prepared to document statistically the relative influence of personal moderators on an A–B relationship if they are to safely attribute any influence to the power of such moderators on the A and B and their relation (e.g. peer group effects, age, gender, or ethnicity; see also Muthén 2002). Similarly, in terms of ‘event-based’ situational moderators (such as those arising from participation in educational interventions or one-off programs), researchers may also need to recognize that attitudes may only change over considerable time, don’t always coincide with behavioral changes, and that the speed, scale or directionality in which either change are unlikely to be stable or permanent, including because of these or other ‘moderators’ and ‘mediators’.

5. It matters that we continue to recognize that human conduct—and education—are not solely amenable to indexes of attitudes that predict a future behavior. The preceding points give rise to our fifth: the importance of placing an A-B relationship in the context of a wider set of theoretical relationships and dynamics that are thought or considered to serve as precursors or predictors of behavior, and as of constraints, on it.

The location of the attitude‐behavior relationship within a more general theoretical framework has been most successfully achieved by the theory of reasoned action and the theory of planned behavior. As well as considering the role of attitudes in determining intentions, and thereby behavior, these models give recognition to the role played by the perceived expectations of others (i.e. subjective norms). (emphasis added)

The location of the attitude–behavior relationship within a more general theoretical framework has been most successfully achieved by the theory of reasoned action and the theory of planned behavior. As well as considering the role of attitudes in determining intentions, and thereby behavior, these models give recognition to the role played by the perceived expectations of others (i.e. subjective norms). (emphasis added)

By extension, this point can have a number of implications for EE. For example, on a direct basis, to what extent has the design of our EE programming been informed (or modified) by theories consistent with or departing from findings and observations such as these? Further, rotating this point to raise other questions, we might ask, in the context of developing climate change education, outdoor education or environmental citizenship education for these times: what might be learned from analysis of environmentally ‘insignificant’ or even ‘irresponsible’ attitudes and behaviors? How might these come to matter, especially to whose educational expectations are enacted, the evidence mobilized (or ignored), and to what these are expected to be, as ‘determiners’, working with or against the grain of learner intentions? Finally, what of ways to address these matters through education and educational research (Reid 2019), particularly if we were to more fully account for the various ‘lacks’ in statistical strength of some of the correlations noted above, or how these are theorized by researchers within or outwith this field?

Implications for EE theory, research, and practice

Many of our points are not actually that new. For example, Kraus (1995, p.71) summarized several implications for theory and research nearly 25 years ago:

… since Allport’s time, there has been a definite shift in general approaches to thinking about the attitude-behavior relationship. The ‘crisis’ was in large part predicated on the conception of behavior as a criterion variable against which the validity of the attitude concept could be tested; a lack of extremely strong and consistent attitude–behavior correlations was thought to invalidate the concept. An increasing number of analytic studies were conducted, and these indeed contributed to the refinement and preservation of the concept. Clearly, attitudes are not synonymous with behavior; attitudes should not be used as easily measured substitutes for behavior measures, nor does attitude theory suggest that attitudes will be the sole determinant of behavior.

In fact, well before the first of these meta-analyses, Hendee (1972, p. 22) highlighted a wide range of implications for research and practice in EE, including that:

We are too willing to believe that favorable environmental attitudes are a result of specific environmental education efforts. It may be that other socializing influences are more important to the development of favorable attitudes. Studies are needed that compare the effects on attitudes of exposure to specific environmental education programs with parental and home influences, personal attributes, extracurricular activities such as Boy Scouts or 4-H, and other factors that may individually or collectively account for environmental attitude differences. Again, such research is not to deprecate the merits of environmental education but to measure its relative impact on attitudes and values compared to other factors. It will help us determine who needs environmental education; who doesn’t? On whom does it have the greatest and least impact? Only with the answers to these questions can we design optimum environmental education strategies to achieve social environmental objectives.

Nearly five decades later, we trust our summary of and reflections on meta-analyses of research on A–B relationships offer more recent but equally challenging insights for contemporary theory and research in EE.

First, that there is reasonably clear, consistent, and substantial evidence, both within and beyond EE research, regarding the relatively moderate strength to many A-B relationships.

Second, that there may be sound, even compelling, reasons for further studies of A-B relationships, in general and in specific contexts and settings, addressing contemporary expectations and theories for EE. In these instances, researchers may benefit from what has been learned in these meta-analyses as to what matters, and might be reasonably expected as, say, a ‘benchmark’. For example, those who choose to study attitudes and behavior can design and conduct studies that better account for measurement and other methodological factors that appear to influence this relationship, but also reflect on the compatibility and comparability of their findings to different theoretical frameworks and associated studies that include the A–B relationship.

Third, that researchers responsible for these meta-analyses have found that several factors serve as moderators in an A-B relationship (Kraus 1990, 1995; Wallace et al. 2005), while others, such as behavioral intention, serve as mediators (Kim and Hunter 1993b). Thus, EE researchers who further investigate an A-B relationship should be in a better position to design studies that provide stronger or more sophisticated accounts of moderating and mediating factors that may influence this relationship in their research setting (Baron and Kenny 1986).

Fourth, one way to account for possible mediating and moderating factors is to ensure an A–B relationship is located and tested within the context of a larger theoretical framework. For example, Kim and Hunter (1993b) found behavioral intention serves as a mediating factor in A–B relationships. Manstead’s (2001) suggestion was that researchers consider Fishbein and Ajzen’s Theory of Reasoned Action (1975) or Ajzen’s more recent Theory of Planned Behavior (Ajzen 2012; Madden, Ellen and Ajzen 1992). Such suggestions are consistent with Kim and Hunter’s work, because both of these theories can serve as frameworks for testing the influence of behavioral intention as a mediator in the A–B relationship. But our more general point is, that those who further investigate the A–B relationship would do well to select a credible theory and model to serve as a broader framework in which to examine this relationship, and consider how it might relate to or differ from other credible theories and findings (e.g. see reviews of theories and models by Jackson 2005, Darnton 2008, Heimlich and Ardoin 2008, and Gatersleben et al., 2014 and their commentaries, such as on which serve to narrow or broaden the field, and whether they are based on reasonable evidence or argument).

In conclusion, we suspect that interpretations of what matters outlined earlier will be as relevant to assessment and evaluation studies designed to explore EE practices and associated learning outcomes as they are to empirically- and theory-driven research and practice development, particularly if the former seek to make a contribution to the field too (Zint 2012). Simply put, it is highly likely that measurement and other methodological factors, as well as moderating and mediating factors, operate in studies of EE practice, and that rigorous accounting of these is needed if we are to gain better insights from these studies of practice, and arguments to adopt and develop them further as quality uses of evidence (see Rickinson 2001; Rickinson et al. 2017; Zelezny 1999).

As Kollmuss and Agyeman (2002) noted, any stasis, spillover, or interactional effect has to be demonstrated not assumed, while in summarizing the internal (psychological) and external (social) factors that influence any relationship, we do well to remember ‘the question of what shapes pro-environmental behavior is such a complex one that it cannot be visualized through one single framework or diagram’ (p. 248). At stake, they observe, is the hoary conundrum of comprehensiveness vs. value: ‘Such a single diagram with all the factors that shape and influence behavior would be so complicated that it would lose its practicality and probably even its meaning’ (p. 248).

In short, the findings of these meta-analyses and reflections encourage us to move beyond the A–B (or K–A–B) designs of the past. They invite us, as a field, to better report the factors in our study designs and procedures (be they ‘confounding’, ‘moderating’, ‘mediating’ or ‘methodological’). And last but not least, they invite us to reflect on the role of such factors in determining or/and undermining the practices and outcomes of the work of this field, particularly if we are to ensure quality uses of research about A–B relationships and develop critical accounts and practices of environmental education—its uses, limitations and prospects.

Notes on Contributors

Tom Marcinkowski has interest in the organization, review, synthesis, and uses of research in environmental education, with particular attention to assessment, evaluation, and research studies in the areas of environmental literacy, responsible environmental behavior, and environmental quality. He has been involved in the development of assessment tools in these areas and, more generally, in program evaluation studies.

Alan Reid edits the international research journal, Environmental Education Research, and publishes regularly on environmental and sustainability education (ESE) and their research. Alan's interests in research and service focus on growing traditions, capacities and the impact of ESE research. A key vehicle for this is his work with the Global Environmental Education Partnership, and via NAAEE's eePRO Research and Evaluation. Find out more via social media, pages or tags for eerjournal.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 These have included that early environmental education programs were ‘guided primarily by unfounded beliefs and emotionally derived truths in need of documentation’ (Hendee 1972, p. 20), and the plea in the conclusion to that article: ‘Let’s do the job right by establishing a high quality program of research commensurate with the importance of environmental education’ (p. 22). Further examples can be found in the entries to Part 5 of Environmental Education (Reid and Dillon 2016c), on Currents and correctives in environmental education.

2 Hendee (1972, pp. 21-22) also noted the need for environmental educators and researchers to identify and assess possible correlates, such as educational level, achievement, outdoors participation, cultural beliefs, life stability and other factors.