When Leaders Are Hard to Find: Reimagining Opinion Leaders in Sparse Communication Networks

Abstract

Diffusion of innovations theory proposes that opinion leaders can enhance the success of science communication campaigns. Such strategies hinge on an assumption that target communities include well-connected and trusted members who need only be identified and recruited. However, sparse or difficult to engage communication networks may not meet this assumption, making opinion leader-based interventions ineffective or impossible. This project reviews two studies involving rural populations where attempts to identify opinion leaders failed. We leverage these examples to highlight barriers to community-based campaigns, theoretical implications for diffusion scholarship, and suggested strategies for communication scholars and practitioners encountering sparse or fragmented networks.

Keywords

opinion leaders superdiffusers social influence diffusion of innovations

Drawing on Rogers’ (2003) diffusion of innovations (DoI) theory, many examples of science communication research and practice have demonstrated the value of identifying and working with opinion leaders—members of a community who can disproportionately influence others’ attitudes and behaviors (Lazarfeld et al., 1948). Exhaustive reviews (Rogers, 2003) suggest that identifying and recruiting opinion leaders for communication interventions can promote a community’s adoption of innovative practices. These innovative practices can help meet diverse prosocial goals, such as promoting daily multivitamin use (Boster et al., 2012), reducing HIV transmission (Kelly et al., 1992), and improving access to safe drinking water (Moser & Mosler, 2008). This legacy of successful campaigns demonstrates that opinion leader-based interventions can be a cost-effective and important strategy in many contexts.

A challenge is that such interventions are only possible if (a) the relevant social network can be identified and all or most members sampled, and (b) there are opinion leaders to be found within that network. A theoretical assumption of these approaches is that any given community will have people who are well connected and trusted sources of advice on a topic of interest; they only need to be identified and recruited. Accordingly, many authors have described the major logistical challenge of opinion leader campaigns to be one of making the recruitment process more feasible and efficient (Boster et al., 2011; Valente & Pumpuang, 2007). Conversely, little to no attention has been paid to the possibility that opinion leaders might not exist within a given network.

The current project was inspired by two studies conducted with rural populations in which attempts to identify opinion leaders were largely unsuccessful. The purpose of the current paper is to explore experiences and findings from these studies to highlight potential barriers to community-based campaigns, consider the implications for DoI theory, and propose solutions or alternatives for when these situations arise.

Theoretical Framework

Diffusion of innovations theory (DoI; Rogers, 2003) describes how novel ideas spread throughout communities. Rogers defines diffusion as “the process by which (1) an innovation (2) is communicated through certain channels (3) over time (4) among the members of a social system” (p. 11, emphasis original). He also specifies that an innovation may include any “idea, practice, or object that is perceived as new by an individual or other unit of adoption” (p. 12). There may be instances when this innovation is not a scientific one. For example, non-scientific information about a new novelist, political candidate, or community tradition could spread among community members. Still, most innovations draw on emerging science in some way, in that they involve novel technologies, behavioral changes supported by new evidence, or timely scientific information that improves understanding of some phenomenon. Thus, like other authors (e.g., Dan et al., 2019; Nisbet & Kotcher, 2009; Silk et al., 2014; Xu et al., 2018), we view science communication as fundamental to the diffusion process. We also embrace Rogers’ (2003) broad definition of innovation, which encompasses the many different contexts (e.g., health care, environmental protection, agriculture) in which scientific advances are made.

Within DoI theory, we focus here specifically on the role of opinion leaders in the diffusion process. Though terms for and definitions of opinion leaders vary, Rogers described them as individuals who can shift others’ attitudes and behaviors in a desired direction, thus increasing the rate and degree of diffusion. Rogers also emphasized the importance of “their unique and influential position . . . at the center of interpersonal communication networks” (p. 27). Early studies, often conducted in rural contexts, focused mainly on working backward to identify opinion leaders after diffusion had already occurred. Scholars began with adopters of a target behavior, such as planting a new strain of corn, then traced the source of adopters’ knowledge back to influential individuals (e.g., Ryan & Gross, 1943). Identifying opinion leaders is also the end goal of several more recent studies that have addressed questions about their key characteristics, motivations, and use of message frames (e.g., Cruz et al., 2025; Winter & Neubaum, 2016; Xu et al., 2018).

Opinion leader-based campaigns, on the other hand, treat identification of opinion leaders as only the first step. The goal of these campaigns is to leverage the influence of identified opinion leaders to accelerate future diffusion of a desirable innovation. The strength of these interventions lies in using knowledge of network structure to change attitudes and behaviors and help those new behaviors encounter reinforcing consequences. In such campaigns, practitioners first identify central actors in the network (recruitment phase), then train those individuals to enact interventions in their communities (training phase). After training, the central actors (i.e., opinion leaders) make a concerted effort to promote the intervention in their communities, and campaign impact can be evaluated based on the level of attitude and behavior change that is ultimately observed. A particularly successful example is Kelly et al.’s (1992) campaign, which recruited and trained opinion leaders to promote safe sex practices among members of LGBTQ+ communities in southern U.S. cities and measurably increased adoption rates within each community.

Challenges With Opinion Leader Campaigns

Although opinion leader-based campaigns can be effective, this may not always be the case. Critiques of these interventions have highlighted the challenges in finding opinion leaders and deciding how, or if, they should be involved. Opinion leaders can be difficult to identify, as evidenced by the variety of techniques that have arisen to find them (Valente & Pumpuang, 2007) and discrepancies in who is found when using different methods in the same populations (Hurtado et al., 2018). Moreover, traditional conceptualizations of opinion leaders as connected people in close geographic proximity (e.g., neighbors) are changing as networks shift to digital spaces, which may overlay or expand the boundaries of person-to-person networks (Nisbet & Kotcher, 2009). New methods, such as improved self-report scales (Boster et al., 2011) and social media analyses (Winter & Neubaum, 2016; Winter et al., 2021), have risen to meet these challenges.

Once opinion leaders have been identified, however, limitations remain. Opinion leaders may be unreceptive to innovative practices and have a disproportionate influence on preserving the status quo (Zhang et al., 2020). Opinion leaders who are too exceptional—for example, because they have greater wealth and social status than most others—may be difficult for the community to approach (de Roo et al., 2023; Feder & Savastano, 2006), and external programs targeting support at these individuals may deepen existing inequities (Matous, 2023). In fact, some studies have challenged the idea that opinion leaders necessarily have a positive, or at least neutral, effect in interventions (Bhandari et al., 2003). In short, the operationalization of the opinion leader construct in real-world applications comes with acknowledged variability and nuance.

Yet even these critiques perpetuate an underlying assumption of DoI theory: Opinion leaders exist within networks, and researchers and practitioners simply need better ways to find and engage them. Further, the theorization of opinion leaders as actors in dense, interconnected networks may not always be realistic, particularly in networks that lack connectivity and cohesion. The potential for sparse networks to exist without effective opinion leaders seems relatively unbroached save for a few studies in medical contexts (Curran et al., 2005; Doumit et al., 2011; Young et al., 2003). For example, Doumit et al. (2011) found no opinion leaders among colorectal cancer pathologists using self-report surveys and social network mapping. Similarly, Curran et al. (2005) found that physicians did not always agree on who was an opinion leader, and some sites seemed to have none. Moreover, some sites had highly fragmented networks, such that even if there was an opinion leader, they had few people in their sphere of influence.

This absence can also occur even when communities are open to opinion leaders. Young et al. (2003), for instance, found that most members of the Royal Australasian College of Surgeons (88%) agreed that “There are colleagues who influence me in such a way that I think of changing my practice [and sometimes do],” (p. 789), but they could not name any local colleagues who fulfilled this role. If opinion leaders could not be found in these contexts, it is certainly feasible that such leaders may not be present or easily identified in rural settings where networks are physically dispersed. This justifies closer examination of possible boundary conditions for opinion leader influence and their implications for science communication applications.

In the interest of contributing to this goal, we present two studies conducted in rural contexts in which we attempted to, but could not, clearly identify opinion leaders to recruit for subsequent communication-based interventions. The first study was designed to identify and leverage opinion leaders to encourage the diffusion and adoption of best occupational safety practices among vineyard managers and owners in Oregon (USA). Similarly, the second sought to use an opinion leader-based campaign to increase adoption of best stream management practices among riparian landowners in Pennsylvania (USA).

Why Study Rural Communities?

In addition to offering an opportunity to reflect on what may be a boundary condition for opinion leader-based campaigns, there are several theoretical and practical reasons why understanding influence in communication networks among rural community members is worthwhile. Rural areas are defined as those with small, sparsely concentrated populations, often with limited access to social services (Hussain et al., 2023). From a practical perspective, rural communities are a priority for communication about environmental innovations, given that the food, agriculture, and land use sector is one of the largest contributors of greenhouse gas emissions (Project Drawdown, 2025). Meanwhile, rural communities face unique challenges that can make campaigns difficult, in that limited services and reduced mobility can make it more challenging for residents to cope with difficulties such as poverty (Klärner & Knabe, 2019), loneliness (Hussain et al., 2023), and access to health care (Amoah et al., 2018). Thus, the same strategies that work in urban communities may fail or be less successful in rural ones.

From a theoretical perspective, there is also a robust body of literature on the nature and importance of social networks in rural communities. Rural communities have traditionally been conceptualized as having strong ties among community members, complemented by weak ties (see Granovetter, 1973) to neighboring communities (Koziol et al., 2015; see, for example, Montes et al., 2018). Compared to urban networks, rural networks tend to also feature more numerous ties among family members and close acquaintances but fewer ties that offer social support or connect network members to institutions (Klärner & Knabe, 2019). In other words, although these networks may often be dense from the perspective of their social and familial connections, they simultaneously have sparse connections to resources and outside information.

This structure can also have important implications for a community’s ability to assimilate and act on scientific information. For example, Tompkins and Adger (2004) argue that strong social networks increase the resilience of both social and ecological systems, in part because accepting and adopting strategies to cope with environmental threats such as climate change is difficult to accomplish otherwise (see also Lombardi et al., 2020). However, Newman and Dale (2005) point out that strong networks alone are insufficient to enable communities to implement novel innovations; rather, communities need bridging ties (i.e., weak ties; Granovetter, 1973) that connect different networks together and provide access to outside resources and information. Networks that have only sparse bridging ties can be limited in their ability to cope with environmental change in several ways (Bodin et al., 2006). For example, sparse networks have reduced redundancy, meaning that if one person is lost, there may not be others available who can adequately fill the same role. If rural communities have only sparse connections to other communities or organizations, a loss of just one bridging tie could leave the whole community cut off from important resources and information. Altogether, this suggests that opinion leaders could be uniquely important in rural networks because they often serve in these bridging roles (Boster et al., 2011; Burt, 1999), which makes it particularly important to understand how health and environmental innovations can be shared effectively when that role goes unfilled.

Finally, it is possible that understanding barriers to the spread of innovations in rural networks could contribute to an understanding of this process in other networks with similar characteristics. For example, networks of rural communities—exemplified by many links within the community and few links to other communities (Montes et al., 2018)—share certain structural features with politically polarized online networks (e.g., Tokita et al., 2021). Some of the work on these online networks has also explicitly noted the importance of key figures who control the flow of information or try to bridge between political groups (Garimella et al., 2018), both roles that opinion leaders might typically fill. Understanding limits on opinion leader approaches in rural contexts may thus offer insights into why disseminating scientific information in other highly clustered networks may prove difficult.

Study 1: Vineyards

In Study 1, our objective was to develop a campaign to identify opinion leaders among vineyard managers and owners in the Eola-Amity Hills American Viticultural Area (AVA) of Oregon’s Willamette Valley. An AVA is a distinct grape-growing region in the U.S. that is recognized for having features that make its viticulture unique, including the climate, geology, soils, physical features, and elevation (Alcohol and Tobacco Tax and Trade Bureau, U. S. Department of the Treasury, 2025). Vineyard owners and managers were chosen as a focus in part because of surveillance data related to mobile machinery-related fatalities. Further, choosing a specific AVA had several potential strengths. First, an AVA has a specific geographical boundary within which vineyard owners and managers could be feasibly identified. Second, members of a specific AVA would likely have a natural affinity with each other due to the shared grape varietals grown and the defined geographical and climatic features of their region. And finally, researchers had initial strong contacts with key people in that region. These features, as well as members’ physical proximity to each other, all seemed aligned with the traditional focus of DoI theory on innovations diffusing through neighbors’ conversations or communications. This stood in contrast to farmers of other crop types that had more dispersed professional associations across the state.

In the study and outreach design, the plan was to target identified opinion leaders with occupational safety and health interventions. Equipment safety, especially related to mobile machinery, was a focus of the project. Agriculture is among the most hazardous industries, and it can be difficult to effectively communicate and promote safety best practices across dispersed farms. Tractors, and particularly tractor rollovers, are a leading cause of fatal work injuries on farms (Rondelli et al., 2018). Similarly, all-terrain vehicles (ATVs) are used in a wide variety of agricultural operations and also present risks for injury and death. The risk of tractor and ATV rollovers is elevated on steep and uneven terrain, as is present in many vineyards (Vigoroso et al., 2019; Figure 1A). Many of these injuries and fatalities can be prevented by equipment safety best practices, such as following manufacturers’ instructions for load stability and use (Fatality Assessment & Control Evaluation, 2021), and the installation of tractor rollover protective structures (Great Plains Center for Agricultural Health, 2014) which are often missing from older tractors that are still in use (Occupational Safety & Health Administration [OSHA], 2019). Accordingly, the goal was to promote agricultural equipment safety and fatality prevention by encouraging vineyard personnel to adopt best safety practices.

Figure 1.

Context of Case Studies With Corresponding Target Best Practices. Study 1 occurred in the Eola-Amity Hills American Viticultural Area of Oregon’s Willamette Valley and aimed to promote adoption of equipment safety best practices, such as wearing personal protective equipment and operating tractors with rollover protective structures (A). Study 2 occurred in four Pennsylvanian counties (B, in green) within the Chesapeake Bay Watershed (B, in blue). The campaign aimed to promote the adoption of riparian buffers, which are vegetation plantings along streams (C). Photo A is courtesy of Dr. Kent Anger, and images were arranged with BioRender.

Study Context

The Eola-Amity Hills region of Oregon’s Willamette Valley is a dedicated AVA. The Willamette Valley was first defined as an AVA in 1984, and in 2006, the Eola-Amity Hills was recognized as a distinct AVA within the valley (Eola-Amity Hills AVA, n.d.). The region is primarily known for pinot noir, but several other grape varieties are also grown there. At the time of the study, the AVA included 2,850 planted acres, around 80 vineyards, and 30 commercial wineries. There are varied relationships among the vineyard owners, managers, and wineries. In the simplest case, a local winery also owns and independently manages one or more vineyards. In other cases, ownership, vineyard management, and winemaking may be handled by three different individuals and/or organizations. In many instances, the management of the vineyards may be contracted out by an owner to a vineyard management service (e.g., Vitis Terra, n.d.), particularly when the owners do not live in the local area. Identifying a named vineyard is also not sufficient for identifying the name and contact information of an owner or manager.

Data Collection

To identify opinion leaders who might help promote equipment safety across vineyards in the AVA, surveys were conducted with vineyard managers and owners in the Eola-Amity Hills area. Researchers invested substantial time through multiple outreach channels to identify and reach each vineyard owner in the AVA and relevant vineyard managers (and/or management companies). This included speaking with vineyard owners and workers at Oregon State University Agricultural Safety Extension classes, phone calls with members of the Eola-Amity Hills Winegrowers Association, and phone calls with vineyard management organization employees. In early 2019, research assistants attempted to contact each vineyard via mail or in person to collect survey responses from owners and/or managers in charge of equipment safety decisions who agreed to participate. The total number of vineyards with an identified possible or confirmed contact was 76. The sample included 23 respondents who were involved with 26 different vineyards. Note that although this sample size may seem small, it represents a substantial proportion of the complete network (i.e., population) of the AVA. Of these participants, 12 (54.5%) were vineyard owners, six (27.3%) were vineyard managers, and two (4.5%) were winemakers. Most were older (M = 60.30 years old, SD = 15.38, range = 38–77), male (90.9%), and had a long job tenure (M = 14.18 years, SD = 9.59).

In addition to details about themselves and their vineyard (e.g., number of acres, varietals grown), participants were asked to identify up to three people and three organizations they would consider their most valued sources of advice about equipment safety. For each source listed, they were also asked how long they had sought advice from that source (in years) and how much they trusted that source’s advice (1 = least trusted, 5 = most trusted).

Results

Based on DoI theory, an assumption going into this study was that vineyard managers and owners would communicate about the industry and business, including safety management and equipment practices, forming a network of influence including opinion leaders with high centrality. The available data, however, were inconsistent with that assumption. Respondents expressed that many people in this industry felt alone when it came to making decisions (vineyard owner, personal communication). Consistent with this comment, two respondents listed themselves among their most trusted sources, and two listed no sources at all. Overall, individual members of the network were surprisingly hard to identify, and among those who responded, the network appeared to be sparse, with many isolates.

People listed by name as sources of information (N = 30) were mostly idiosyncratic to a particular respondent; only seven nominees (23.3%) were listed by multiple people as a trusted source. Ten nominees were associated with a vineyard management company, seven were associated with a particular vineyard, seven were associated with an equipment supplier, and six were from other types of organizations (e.g., a winegrowers’ association). Most sources were located near the respondent’s vineyard (M = 32.80 miles away, SD = 103.72), and only one respondent listed an individual based more than 50 miles away. Respondents reported that they had well-established relationships with these individuals, between 6 and 9 years on average. Unsurprisingly (as the survey focused on trusted sources), the average level of trust was quite high (M = 4.67, SD = 0.56). Members of vineyard management companies tended to be highly trusted (M = 4.80, SD = 0.42, n = 10), as were people from other types of organizations (M = 4.90, SD = 0.22). Representatives from equipment suppliers were slightly less highly trusted (M = 4.31, SD = 0.82, n = 7), even though they were again the people who had the longest relationships with respondents (around 9–12 years, on average).

Several types of organizations were also identified either as general sources of information or as employers of a trusted individual. Of the 29 organizations listed across all participants, seven were vineyard management companies, six were vineyards, five were equipment suppliers, three were professional associations, and eight were another type of organization (e.g., a law firm). Equipment suppliers also tended to be mentioned most often (a total of 19 times), with one particular supplier named 13 times across the sample (compared to a median of 1). Most organizations listed were near the respondent’s vineyard (M = 37.52 miles away, SD = 104.86); only two respondents listed an organization more than 50 miles away, and only one listed an organization that was out-of-state (CA, USA). Respondents reported that they had well-established relationships with many organizations, between 6 and 9 years on average. Again, the average level of trust in these organizations was quite high (M = 4.62, SD = 0.55), particularly for vineyard management companies (M = 4.79, SD = 0.39) and professional associations (M = 4.82, SD = 0.17). Equipment companies had a slightly lower level of trust (M = 4.17, SD = 0.95), though they were the most common sources of equipment safety information and often had the longest-lasting relationships reported (more than 9–12 years, on average).

Overall, these findings suggested that first, a regionally defined network of individuals proved difficult to reach. Second, and contrary to what would be expected based on DoI theory, there was not a clear influence network that emerged in this community relative to safety and equipment decisions. Vineyard owners appeared to consult one another only sporadically (at least on this particular topic), and no particular vineyard stood out as a widely trusted source. Moreover, though respondents most commonly relied on equipment suppliers—and one supplier in particular—for this information, these sources had somewhat lower levels of trust. The applied outcome of the study was that researchers partnered with the most frequently nominated supplier to conduct seasonal safety and fatality prevention outreach, rather than conducting an opinion leader campaign.

Study 2: Riparian Buffers

To increase the chances of identifying opinion leaders, our second study combined the network-oriented approach of Study 1 with a self-report scale (Boster et al., 2011) to capture both community-identified and self-identified opinion leaders. Specifically, our objective was to develop an opinion leader campaign to promote riparian buffer adoption among land managers and owners in four Pennsylvanian counties in the Chesapeake Bay watershed (Figure 1B). Riparian buffers are plantings of permanent vegetation along streams, which are considered a best management practice for improved water quality (Figure 1C). In addition, choices in the vegetation can lead to co-benefits for ecosystems and for landowners; for example, wide forested buffers can serve as habitat for vulnerable species, and planting fruit- or nut-bearing trees produce goods that can be sold as supplemental income (Herbstritt et al., 2019). Though the ecological benefits of riparian buffers are well established (Cole et al., 2020), the adoption rates of this practice have consistently fallen short of annual goals in all six states of the Chesapeake Bay watershed (Chesapeake Progress, 2022). One key need for accelerating adoption is new outreach efforts to private landowners (Herbstritt et al., 2019), particularly efforts that target ever-limited conservation resources into more cost-effective communication campaigns. As such, this work aimed to identify opinion leaders among riparian landowners who could help engage in targeted communication about the riparian buffers’ value and encourage adoption.

Study Context

As the largest estuary in the USA, Chesapeake Bay is a vital resource both for its ecological functions and for economic livelihoods. However, these services have been increasingly threatened by worsening water quality from nitrogen, phosphorus, and sediment pollution (U.S. Environmental Protection Agency, 2010). A key strategy for mitigating such pollution is the planting of riparian buffers (Chesapeake Bay Program, 2015). This has led to ambitious state-level goals, such as Pennsylvania’s intent to plant 110,000 acres of buffers by 2025 (Pennsylvania Department of Conservation and Natural Resources [DCNR], 2016). Despite significant cross-stakeholder investments, Pennsylvania (like all Bay states) has consistently fallen short of its annual planting goals. Since 2010, state-wide planting has averaged 147 miles out of its 900-mile annual goal—an 84% gap between target and outcome. Pennsylvania’s shortfalls are alarming because the state contains most (75%) of the Bay’s largest tributary (Pennsylvania Department of Environmental Protection [DEP], 2021). As such, riparian buffer adoption in Pennsylvania offered a high-priority scenario to test an opinion leader campaign.

Data Collection

Unlike in the first study, there was no clear boundary defining the communication network among riparian landowners. As such, we employed a two-step approach to identifying opinion leaders: We identified and recruited an initial wave of riparian landowners for an online survey, then we recruited trusted sources generated from the first wave for a follow-up survey. First, riparian landowners were identified from four Pennsylvania counties—Bedford, Centre, Lancaster, and York—prioritized to reduce 50% of the state’s nitrogen pollution to the Chesapeake Bay (DEP, 2021). Riparian landowners were identified through a geospatial analysis of 1–m² enhanced water flow path data (Conservancy Innovation Center, 2023) joined with parcel landownership data. After quality checks (e.g., to remove duplicate landowners, public land, etc.), this process identified a total of 35,458 eligible landowners, including 7,131 in Bedford, 5,777 in Centre, 10,912 in Lancaster, and 11,638 in York. From this list, 625 mailing addresses per county were randomly selected, for a total target sample of 2,500. Selected landowners received two postcards: an invitation to complete a survey and a reminder two weeks later (a best practice for increasing response rates; Sakshaug et al., 2019). The invitation included both a hyperlink and a QR code to a survey hosted on Qualtrics. To incentivize participation, subjects could enter a random drawing for a chance to win a $10 Amazon gift card.

The survey attempted to identify opinion leaders through two approaches: a network analysis and a self-assessment. As in Study 1, participants were asked to identify up to three people whom they would consider their most valued sources of advice about stream management. For each source listed, they were asked how much they trusted that source’s advice (1 = least trusted, 5 = most trusted). Names generated were used for a second wave of sampling. To capture advice seeking beyond individuals, surveys included an open-ended question on other trusted sources. In addition, we attempted to capture self-identified opinion leaders through the 15-item superdiffuser scale (Boster et al., 2011) tailored to environmental mavenness (1 = strongly disagree, 5 = strongly agree). The superdiffuser scale proposes that highly influential people have three characteristics: They are highly socially connected (connectors), highly knowledgeable in the topic over which they exert influence (mavens; here, regarding the environment), and highly persuasive when sharing knowledge with their contacts (persuaders).¹

Of the 2,500 participants invited to take the survey, 52 (2.1%) submitted valid responses. This response rate was lower than previous studies of riparian landowners in this region (e.g., Metcalf et al., 2019). To increase participation, the survey was then distributed via mailing lists of county conservation district offices.² This extended sampling led to 64 valid responses (a 23.1% increase in sample size) who named 16 contactable trusted sources. Among these trusted sources, seven (43.8%) responded to the second-wave survey. Across both waves, respondents were older (M = 56.8 years, SD = 14.5, range = 24–86 years), mostly male (59.7%), owned and managed their own property (93.7%), and tended to have long land tenure (58.3% residing there for 15+ years). Most respondents had a riparian buffer on their stream (74.6%), and a few respondents intended to plant one (3.2%). However, most respondents with a riparian buffer were not the ones who made the decision to plant it (68.1%), and 21.7% of respondents had not implemented any practices to improve water quality on their property.

Results

Based on the DoI theory, an assumption going into this study was that land managers and owners would seek advice on management practices from each other or highly influential individuals within their communities. However, this assumption was not supported by reports of network connections nor by self-assessments of opinion leadership.

Half (50.0%) of landowner respondents said that there was no one at all whom they would go to for advice on stream management. Among the landowners who did name an individual as a source of information, sources tended to be sparse; for example, respondents were asked to name up to three trusted sources, but most (57.1%) could only name one source. In addition, sources tended to be idiosyncratic to a particular respondent; a total of 48 people were listed by name as sources of information, yet only five (10.4%) were listed by more than one person. In fact, seeking advice directly from people was a less common source of information (20.0%) relative to seeking advice from organizations (50.0%) or the internet (44.0%). However, the reliance on organizations was not necessarily by choice: 16.7% of respondents said that they would only seek advice from people, but they had no one to go to for advice. Overall, this suggests that landowners are not averse to seeking advice on stream management from individuals, but no one may currently be fulfilling this role, perhaps due to sparse or idiosyncratic networks.

These findings on advice seeking were supported by self-assessments of opinion leadership using the superdiffuser scale (Boster et al., 2011). Few participants were self-identified opinion leaders, and this lack of opinion leaders likely stems from low social connectivity (Table S2). Connectivity scores in this population were lower than the midpoint of the scale (M = 2.89, SD = 1.19 on a 5-point Likert-type scale) and lower relative to both persuasiveness (M = 3.70, SD = 0.78) and environmental mavenness (M = 3.83, SD = 0.83). These connectivity scores were also lower than, though within range of, average American adult connectivity scores identified in a previous large-scale study (M = 3.13, SD = 0.89; Boster et al., 2012). In short, neither a self-assessment approach nor a network-oriented approach identified a strong opinion leader presence among riparian landowners, and this result seems to be due to low social connectivity rather than lack of persuasiveness or environmental mavenness.

Moreover, the few self-identified opinion leaders that were found seem to have limited influence in their communication networks. None of the first-wave respondents who self-identified as opinion leaders were named as a trusted source for stream management advice by other respondents. These respondents were also no more likely to list trusted sources of advice than were non-opinion leaders: combined, they only listed one name as a trusted source. Further, none of the second-wave respondents (who were named as trusted sources of information by others in the community) met the threshold to be self-identified opinion leaders. These results, mirrored by results from the network data, suggest sparse or idiosyncratic communication networks in which opinion leader campaigns would be ineffective or impossible.

Discussion

Rogers (2003), in his original formulation of DoI theory, proposed that highly influential community members were important elements of the diffusion process. Opinion leader-based campaigns, which attempt to recruit these influential members in the service of promoting the adoption of novel innovations, have since emerged as an important science communication tool. Drawing on previous examples of successful campaigns (e.g., Boster et al., 2012; Kelly et al., 1992; Moser & Mosler, 2008), the original purpose of both Study 1 and 2 was to identify opinion leaders among rural land managers and owners, then recruit these leaders as partners in communication-based interventions.

However, neither a network-oriented approach (Studies 1 and 2) nor a self-assessment approach (Study 2) found evidence to support a core assumption underlying these opinion leader-based campaigns: that community members sought advice from each other or from highly influential individuals within their communities. Instead, communication networks were difficult to capture, and the fragments that were captured appeared sparse or idiosyncratic, such that opinion leaders seemed absent or—for the few individuals who self-identified as an opinion leader—had limited influence. Similar findings in other research contexts (Curran et al., 2005; Doumit et al., 2011; Young et al., 2003) suggest that this is not an isolated phenomenon; network density may be a boundary condition for opinion leader-based campaigns, such that their application in sparse networks is challenging. As such, it is worth discussing how future researchers and practitioners might navigate contexts in which traditional methods for and applications of opinion leader campaigns may be impossible or ineffective.

To be clear, the points that follow are intended to be speculative, presenting ideas for what we see as plausible explanations for our findings and paths forward. Future research will be vital for exploring these possibilities and continuing to advance the understanding of influence in sparse networks. We also want to emphasize that opinion leadership is only one component of DoI theory, so this discussion does not speak to other components of the theory not explored in our studies.

Do We Need New Methods to Capture Opinion Leaders in Sparse Networks?

A first consideration is that opinion leaders may exist in the study communities, but our methods failed to identify them. Although network analyses and self-assessments have successfully identified opinion leaders in past interventions, much of the theoretical work is conducted in relatively bounded and well-connected networks, such as a college campus (an acknowledged limitation in Carpenter et al., 2015), or on broader populations that can be captured through national crowd-sourcing platforms (e.g., Boster et al., 2011). As such, one possibility is that we need new or modified methods to capture sparse communication networks in targeted populations, as well as the opinion leaders within them.

In applied contexts, the first hurdle is to find and survey members of the relevant social network. Collecting representative rural data is challenging because populations are small and too specific for public or crowd-sourced data to accurately reflect (Scally et al., 2020). Identifying whom to survey among rural communities can be complicated by discrepancies between who owns the land, manages the land, and makes management decisions, as well as whether these individuals are contactable across time. For instance, Study 1 encountered the challenge that vineyards can evolve quickly as different parcels change ownership, which is exemplified by the AVA’s shift from 89 to 76 vineyards over just 5 years (Eola-Amity Hills AVA, n.d.). Similarly, in Study 2, contact depended upon mailing addresses, which required disentangling land ownership versus tenure, because a manager could be interested in adopting a practice that a landowner does not permit, and an owner could invest in a practice that a renter does not support. In addition, once members are contacted, surveys are often not returned, even when leveraging strategies to increase response rates. For example, one strategy is to partner with “intermediary organizations” that have access to the target population (e.g., in our cases, extension and county conservation organizations); yet even “insider” contact lists may not be comprehensive or up-to-date (e.g., Farley et al., 2014). Other methods that increase responses tend to increase money, time, and labor, such as five-step contact protocols with multiple monetary incentives (Martinez et al., 2020) or having survey packets home-delivered by someone the participants personally know (Edelman et al., 2013). In short, researchers still need innovative methods to capture representative data in sparse networks.

Once sparse networks are captured, the next hurdle is to maximize the potential for capturing these networks’ opinion leaders. One option for doing so might be to seek generalist opinion leaders. In fragmented communication networks, surveying localities that lack bridges between one another may require asking a broader referent question (J. Dearing, personal communication). For example, Study 2 asked participants from whom they seek advice regarding the narrow topic of “stream management,” whereas broader terms such as “environmental issues” or “land management” may have encouraged respondents to name more generalists that they trust for advice. This approach would likely have generated more names of trusted sources; however, there are caveats. First, it is unclear whether a generalist opinion leader would be as effective for diffusing specialized knowledge. Opinion leadership is traditionally conceptualized as monomorphic (influential over a specific topic or domain) rather than polymorphic (influential over broader topics or a range of domains, i.e., “generalist” opinion leaders; Rogers & Cartano, 1962). As an illustration, Carpenter et al. (2024) found that opinion leaders on a broader topic (health) were less effective at promoting a specific target intervention (vaccination) than leaders on a narrower topic (vaccines). Thus, generalist opinion leaders may pose a tradeoff of quantity over quality, because it is unlikely that an individual will know enough about all facets of a broad topic to provide reliable advice about a narrower topic (Carpenter et al., 2024). Moreover, in Study 2, the superdiffuser scale was tailored to find general environmental mavens, rather than narrower stream management mavens. As such, the question remains as to why self-assessment with a general term did not capture more opinion leaders or those with greater influence.

Another possible method is identifying opinion leaders after a campaign. Britt et al. (2022) proposed that in communities where opinion leaders cannot be identified a priori, researchers should consider launching the campaign and recruiting opinion leaders as they emerge. For example, Study 1 ultimately led to focused outreach at a particular equipment supplier. Perhaps with further intention, we could have leveraged those outreach events to uncover additional opinion leaders in the region via conversations with customers or store employees. Such emergent methods may be effective for topics that are new to a community. However, in contexts where participants are already highly knowledgeable about a topic, it is less clear if new opinion leaders would emerge and, if so, why they had not already emerged. For example, most participants in Study 2 appeared knowledgeable about the target innovation: Most already had (74.6%) or planned to adopt (3.2%), and several who intended not to adopt (7.9%) cited barriers that suggest knowledge of the innovation, such as cost and time, rather than a need to learn more. This implies that environmental information has been dispersing into these communities, and opinion leaders on this topic should already have emerged if they were going to be a viable avenue of communication. Furthermore, this “emergence” approach spurs a more fundamental question: Were these existing opinion leaders hidden within the network and unveiled by the communication campaign, or did the campaign serve as a catalyst to create them? This is an important theoretical question that likely deserves attention in future DoI research.

Do We Need to Increase Connectivity in Sparse Networks First?

An alternative explanation for our inability to identify opinion leaders is that none currently exist in our study contexts. Other researchers attempting to enact opinion leader campaigns have reached similar conclusions in sparse or fragmented networks (Doumit et al., 2011; Young et al., 2003). In such contexts, an opinion leader campaign may be ineffective at the time of the study, yet this is not to discount the potential for future opinion leader campaigns. As Curran et al. (2005) speculated, “Can opinion leaders be created?” (p. 703). Considering this question, particularly in fragmented networks, may be an important tool for improving the feasibility of opinion leader campaigns in communities like the ones studied here.

Effective opinion leader campaigns require connectivity for knowledge to flow through informal networks. Though researchers acknowledge that “connectedness is an essential part of opinion leadership” (Carpenter et al., 2015, p. 121), they less often discuss what happens when this “essential part” is missing. Social connections are assumed to exist and emerge organically, but such an assumption may not be met in fragmented networks. A spontaneous increase in connectivity between fragmented groups is rare (Krebs & Holley, 2004). Instead, fragmented networks need intentional bridging, particularly in contexts where cultural, social, and political divides may lead to deeply fragmented networks (Laird-Benner & Ingram, 2010).

Identifying where and why network fragments exist is possible with methods already used in opinion leader research: social network analysis. Social network analysis methods examine the network structure to find highly central actors as an indicator of opinion leadership (Valente & Pumpuang, 2007). Common approaches include snowball sampling from random or representative samples, or, for bounded networks, surveying the entire population (Valente & Pumpuang, 2007). All of these approaches collect data to reconstruct a representative network, such that even if researchers cannot identify opinion leaders, they can identify fragmented groups or isolates instead (Cross et al., 2002).

Finding fragmented groups is a vital first step to discerning the reasons for fragmentation and implementing interventions to overcome them. Networks may fragment around shared interests, goals, or skillsets, but they may also fragment around boundaries (Cross et al., 2002). Boundaries may be institutional—such as individuals working on a common issue but in isolated sectors (Vance-Borland & Holley, 2011)—or hierarchical. Boundaries may be geographical (Laird-Benner & Ingram, 2010), yet physical boundaries can form even in seemingly near proximity (e.g., due to building layouts inconducive to informal contact; Curran et al., 2005). Similar geographical boundaries may characterize our study contexts, because rural communities have physical distances between neighbors which may make interactions rarer without natural places to gather and connect. Physical proximity is further complicated by differences between rural land ownership and management. For example, many vineyards in Study 1 had outsourced management to a vineyard management company; this created a sub-group of vineyards in which the location of the land was not the location of the owner or manager. Regardless of how boundaries arise, identifying boundaries can help to target support that increases connectivity across them.

In such cases, targeted network interventions focused on functionally located people or organizations may be especially important, because fragmented networks are improved through strategic connectivity (Krebs & Holley, 2004). Forming connections at random or maximizing connections by quantity over quality may not be effective. Rather, strategic interventions intentionally revive existing weak links (Tian et al., 2010) or, more often, create new links between fragmented groups. New links can be formed by building relationships (Hussain et al., 2023) and facilitating opportunities for collaboration on a shared goal (Krebs & Holley, 2004; Tompkins & Adger, 2004). A potential example applied to Study 1 would be partnering with vineyard management organizations for safety and health outreach, given their functional position and potential influence in the fragmented network. However, the role of external entities in this process varies. In some contexts, prescribed approaches have been successful, such as jointly appointing members from isolated clusters on collaborative projects (Cross et al., 2002) or inviting peripheral network members (Cross et al., 2006). In other contexts, external entities have opted to create structures for participant-led collaborations to emerge organically, such as opening lines of communication between isolated clusters (Cross et al., 2002; Tompkins & Adger, 2004) or creating awareness of expertise distributed across the network (Cross et al., 2006). Whether changes are externally or internally driven, targeted network interventions have a legacy of successfully increasing strategic connectivity.

After enacting targeted interventions, a further goal is to assess structural changes through longitudinal social network analyses. While it is challenging to have repeatedly high participation in network assessments, such an approach can offer insight into two key outcomes. First, communication researchers and practitioners can assess metrics of increased connectivity over time and ensure structures are in place for opinion leader campaigns. For example, longitudinal studies have found that, when placed in new networks, people who perceive themselves as highly connected do eventually emerge as central (Carpenter et al., 2019). This suggests that networks can evolve into structures more conducive to effective opinion leader campaigns and can be monitored for this change. Second, the network can be assessed for vulnerabilities that cannot be solved by opinion leader campaigns alone. Networks in isolated communities may develop from scattered fragments to structures reliant on a key individual to connect flows of information (Krebs & Holley, 2004). While this individual would be a prime candidate for recruiting as an opinion leader, the overall network structure is unsustainable—the individual is often overburdened, and the overall network is vulnerable because it depends on one actor (Krebs & Holley, 2004). This raises the concern that placing the burden of interventions on such highly connected individuals may not be appropriate (e.g., “individual bias,” where individuals promoting innovations bear the burdens of structural inadequacies; Rogers, 2003) and may call for alternatives that place less stress on key people.

Do We Need Alternatives to Opinion Leader Campaigns in Sparse Networks?

A final consideration is that science communication researchers and practitioners may need productive alternatives to traditional opinion leader campaigns in sparse networks; in other words, network structure may be a boundary condition for this approach that cannot be easily overcome. Though key individuals are a preferred contact, both studies highlight the importance of organizations in rural networks. Organizations—whether private companies, professional associations, non-governmental organizations, county government agencies, or other “intermediaries” (Cunningham et al., 2021)—were not necessarily landowners’ most highly trusted nor preferred sources of information, but they were often the most common and long-lasting relationships. This longevity may be important for sustained engagement because organizations often have access to more consistent resources, such as long-term grant funding, relative to individuals in a community (e.g., Jacobs et al., 2010). For example, while Study 1 largely failed to identify individuals as opinion leaders in the vineyard network, the discovery that equipment suppliers were common sources of safety information led to subsequent seasonal outreach at the storefront for the most popular supplier.³ As such, there is an opportunity to leverage these relationships to support individuals with the potential to become influential, to promote innovations directly, or to generate connectivity in fragmented networks.

For example, organizations might focus on supporting less widely influential community members, who may still hold promising aptitudes for promoting behavioral changes if provided the opportunity. Some successful campaigns have recruited individuals who were not widely influential but spanned diverse pools of knowledge, known as network “brokers” (Zhang et al., 2020). Likewise, another campaign found that randomly selected, non-central network members were surprisingly more effective at promoting innovation adoptions than opinion leaders identified by social network analyses and community interviews (Matous, 2023). Matous (2023) concludes that communities may be biased in who they perceive as influential, and over-reliance on these individuals could miss effective campaigns delivered by persuasive but less-established community members. Overall, organizations that have committed to opinion leader-based campaigns but struggle to find them might consider the potential for overlooked candidates—often women or younger, less-affluent community members—to become more influential if extended external support.

In addition to promoting innovations through individuals, organizations can promote changes in attitudes and behaviors directly. Organizations are also better suited for sharing information and promoting innovations over long periods of time (multiple generations), independent of an individual opinion leader who may leave a network. Many communication campaigns already use this organizational model with varying success (e.g., Public Service Announcements from the U.S. Centers for Disease Control and Prevention and the U.S. National Weather Service).

Finally, organizations could shift their focus from convincing people to care about a particular issue (changing attitudes) to connecting people who already care (changing the network). For example, supplemental data from Study 2 found that a lack of opinion leaders among riparian landowners was not explained by a lack of concern—rather, landowners’ scores on environmental concern scales (Schultz, 2001) were much higher than has been reported for the general U.S. population (see Table S3 vs. Cruz & Manata, 2020). As such, organizations could function as the bridge to connect individuals who are highly concerned but lack the connections to act on these concerns. Many organizations have explored ideas to fulfill this role, such as hosting peer-to-peer learning circles (which may appeal particularly to those who have already adopted and are looking to connect with others; Read & Wainger, 2023), gathering neighbors to demonstrate successful practices (e.g., Herbstritt et al., 2019), or using language in campaigns that emphasizes individuals’ concerns are already a social norm (Metcalf et al., 2019).

Strengths and Limitations

The findings of these studies should be interpreted with their strengths and limitations in mind. The major strength of the investigation is its external validity. The two studies focused on socially important priorities in rural communities. They each involved different interventions in different locations at different points in time, but identified similar overall patterns of difficult-to-identify, sparse networks and isolated community members. The fact that the findings were similar in two very different contexts suggests that similar problems are likely to emerge in other contexts and for other communities, particularly those in rural areas. Thus, we believe this is a problem of broad applicability to science communication scholars and practitioners. Our consideration of these cases points to the need for more theoretical and applied work that attempts to address the challenges of, and alternatives to, opinion leader approaches in sparse networks.

The results observed in both studies were also likely limited by relatively small samples. Particularly for Study 2, where the response rate was very low, it is possible that survey respondents happened to be people who were less connected or less likely to be opinion leaders than other community members. There is no clear evidence to suggest this is the case, however, especially since the Study 2 sample did include several people who self-identified as high in mavenness and persuasiveness. In other words, these people would otherwise have been good candidates for recruitment for an opinion leader campaign had connectivity not been so limited. Furthermore, even if we had been able to identify some opinion leaders in other segments of these communities, that would not have addressed the broader problem that many members of our samples could identify no trusted sources of advice. These communities are clearly fragmented to some extent, even if more developed clusters do exist among individuals not sampled.

A final limitation is that although we have endeavored to speculate about a range of tactics that could enhance, supplement, or substitute for traditional opinion leader campaigns, the present investigation did not examine directly whether these suggested tactics would be effective. Future research is needed to identify how frequently different communities evidence the types of sparse networks investigated here and how effective various strategies may be at overcoming this barrier. There are also many other factors that could be relevant to consider beyond those we discussed here. For example, we focused on a single-item measure of trust in different sources, but this may not effectively capture the breadth of trust (e.g., see Besley et al., 2018). Alternatively, it could have been useful to consider measures other than trust, such as shared values or knowledge of a particular topic. Either way, it is possible that a different measure could have prompted participants to think about different kinds of valued relationships they might have, and this could be an alternative way to attempt to capture network structure in the future. Future research that explores both the possibilities we raise and other considerations would provide important insights into the boundary conditions of opinion-leader campaigns and contribute to further theoretical development.

Conclusion

Motivated by two studies in which efforts to identify opinion leaders proved wholly or partially ineffective, we have endeavored to identify productive alternatives for science communication researchers and practitioners who encounter inaccessible and fragmented networks. For networks with high fragmentation and/or low connectivity, we suggest three explanations and potential solutions: (a) we may need modified methods to identify opinion leaders in sparse networks, such as finding generalist opinion leaders or finding specialists after launching a campaign rather than before; (b) we may need to increase network connectivity before opinion leaders can emerge; and (c) we may need to consider productive alternatives that emphasize the role of organizations with great potential to connect and influence the target population. Alone or in combination, these tactics may enhance community-based interventions when traditional opinion leader campaigns are difficult or infeasible. Exploring whether these findings also apply to other networks that share structural features with rural networks (e.g., polarized online networks; Garimella et al., 2018; Tokita et al., 2021) may also be useful. Additional research along these lines would have substantial practical value in addition to advancing DoI theory and our understanding of the boundary conditions for opinion-leader campaigns in diverse contexts.

Supplemental Material

sj-docx-1-scx-10.1177_10755470261435541 – Supplemental material for When Leaders Are Hard to Find: Reimagining Opinion Leaders in Sparse Communication Networks

Supplemental material, sj-docx-1-scx-10.1177_10755470261435541 for When Leaders Are Hard to Find: Reimagining Opinion Leaders in Sparse Communication Networks by Marissa W. Kopp, Shannon M. Cruz and Ryan Olson in Science Communication

Footnotes

Acknowledgements

For Study 1, we would like to acknowledge Illa Gilbert-Jones and Layla Mansfield for their contributions to project design and management; Haley Strenke, Melina Rodriguez, and Sydney Running for their assistance with project planning and field work; the Oregon State University Agricultural Extension program, the Eola-Amity Vineyard Association, regional vineyard management companies, and Kerry Boenisch for assistance identifying and reaching out to vineyard owners/managers; Jim Dearing and Ginger Hanson for their consultation related to social network analysis; and participants who shared their insights.

ORCID iDs

Marissa W. Kopp

Shannon M. Cruz

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The work was funded by the Pennsylvania State University College of Agricultural Sciences under the 2022–2023 Graduate Student Competitive Grants Program; the National Institute for Occupational Safety and Health through the Oregon Fatality Assessment and Control Evaluation (OR-FACE) program (grant number U60 OH008472); and partially supported by the Oregon Institute of Occupational Health Sciences at Oregon Health & Science University via funds from the Division of Consumer and Business Services of the State of Oregon (ORS 656.630).

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

Data are available upon request to the corresponding author.

Supplemental Material

Supplemental material for this article is available online at .

Notes

Author Biographies

Marissa W. Kopp (PhD, Pennsylvania State University) is a Forestry Technical Officer at ACR (formerly the American Carbon Registry). This research was completed while earning her PhD in Ecology at the Pennsylvania State University. Her research interests span forest ecosystem ecology and science communication.

Shannon M. Cruz (PhD, Michigan State University) is an assistant professor in the Department of Communication Arts and Sciences and a co-fund of the Huck Institutes of the Life Sciences at Pennsylvania State University. Her research interests include social influence and environmental and risk communication.

Ryan Olson (PhD, Western Michigan University) is Professor of Occupational Health Psychology at the Rocky Mountain Center for Occupational and Environmental Health at the University of Utah. His applied research program has focused on interventions to improve the safety, health, and well-being of isolated or lone workers who perform hazardous work. Dr. Olson is a charter member of the Society for Occupational Health Psychology and Directed the Oregon Fatality Assessment and Control Evaluation program for over a decade.

References

Alcohol and Tobacco Tax and Trade Bureau, U. S. Department of the Treasury. (2025). AVA map explorer. https://www.ttb.gov/regulated-commodities/beverage-alcohol/wine/ava-map-explorer

Amoah

P. A.

Edusei

Amuzu

(2018). Social networks and health: Understanding the nuances of healthcare access between urban and rural populations. International Journal of Environmental Research and Public Health, 15(5), 973. https://doi.org/10.3390/ijerph15050973

Besley

J. C.

Dudo

Yuan

(2018). Scientists’ views about communication objectives. Public Understanding of Science, 27(6), 708–730. https://doi.org/10.1177/0963662517728478

Bhandari

Devereaux

P. J.

Swiontkowski

M. F.

Schemitsch

E. H.

Shankardass

Sprague

Guyatt

G. H.

(2003). A randomized trial of opinion leader endorsement in a survey of orthopaedic surgeons: Effect on primary response rates. International Journal of Epidemiology, 32(4), 634–636. https://doi.org/10.1093/ije/dyg112

Bodin

Ö.

Crona

Ernstson

. (2006). Social networks in natural resource management: What is there to learn from a structural perspective? Ecology and Society, 11(2), r2. https://doi.org/10.5751/es-01808-1102r02

Boster

F. J.

Carpenter

C. J.

Andrews

K. R.

Mongeau

P. A.

(2012). Employing interpersonal influence to promote multivitamin use. Health Communication, 27(4), 399–407. https://doi.org/10.1080/10410236.2011.595771

Boster

F. J.

Kotowski

M. R.

Andrews

K. R.

Serota

(2011). Identifying influence: Development and validation of the connectivity, persuasiveness, and maven scales. Journal of Communication, 61(1), 178–196. https://doi.org/10.1111/j.1460-2466.2010.01531.x

Britt

B. C.

Britt

R. K.

Anderson

Fahrenwald

Harming

(2022). Finding the invisible leader: When a priori opinion leader identification is impossible. Communication Reports, 35(1), 38–52. https://doi.org/10.1080/08934215.2021.2015796

Burt

R. S.

(1999). The social capital of opinion leaders. The Annals of the American Academy of Political and Social Science, 566(1), 37–54. https://doi.org/10.1177/000271629956600104

10.

Carpenter

C. J.

Boster

F. J.

Kotowski

Day

J. P.

(2015). Evidence for the validity of a social connectedness scale: Connectors amass bridging social capital online and offline. Communication Quarterly, 63(2), 119–134. https://doi.org/10.1080/01463373.2015.1012217

11.

Carpenter

C. J.

Rahman

Kotowski

M. R.

(2024). Vaccine mavens or health mavens? Southern Communication Journal, 89(1), 79–89. https://doi.org/10.1080/1041794X.2023.2277432

12.

Carpenter

C. J.

Zhu

Smith

R. A.

(2019). Do people who identify as popular become popular in a new network? A 9-month longitudinal network analysis. Journal of Social Structure, 20(1), 1–24. https://doi.org/10.21307/joss-2019-001

13.

Chesapeake Bay Program. (2015). Riparian forest buffer outcome management strategy 2015–2025, v. 1. https://d18lev1ok5leia.cloudfront.net/chesapeakebay/documents/3a_forest_buffer_final.pdf

14.

Chesapeake Progress. (2022). Forest buffers. https://www.chesapeakeprogress.com/abundant-life/forest-buffers

15.

Cole

L. J.

Stockan

Helliwell

(2020). Managing riparian buffer strips to optimise ecosystem services: A review. Agriculture, Ecosystems & Environment, 296, 106891. https://doi.org/10.1016/j.agee.2020.106891

16.

Conservancy Innovation Center. (2023). PA data downloader. Chesapeake Conservancy. https://www.chesapeakeconservancy.org/conservation-innovation-center/high-resolution-data/enhanced-flow-paths/pa-data-downloader/

17.

Cross

Borgatti

S. P.

Parker

(2002). Making invisible work visible: Using social network analysis to support strategic collaboration. California Management Review, 44(2), 25–46. https://doi.org/10.2307/41166121

18.

Cross

Laseter

Parker

Velasquez

(2006). Using social network analysis to improve communities of practice. California Management Review, 49(1), 32–60. https://doi.org/10.2307/41166370

19.

Cruz

S. M.

Manata

(2020). Measurement of environmental concern: A review and analysis. Frontiers in Psychology, 11, Article 363. https://doi.org/10.3389/fpsyg.2020.00363

20.

Cruz

S. M.

Manata

High

A. C.

Worley

T. R.

(2025). On the nature of influence: Identifying and characterizing superdiffusers in seven countries. Human Communication Research, 51(1), 29–39. https://doi.org/10.1093/hcr/hqae017

21.

Cunningham

Jacklitsch

Richards

(2021). Intermediary perspectives on total worker health in small businesses. International Journal of Environmental Research and Public Health, 18(19), 10398. https://doi.org/10.3390/ijerph181910398

22.

Curran

Thrush

C. R.

Smith

J. L.

Owen

R. R.

Ritchie

Chadwick

(2005). Implementing research findings into practice using clinical opinion leaders: Barriers and lessons learned. Joint Commission Journal on Quality and Patient Safety, 31(12), 700–707. https://doi.org/10.1016/S1553-7250(05)31091-9

23.

Dan

Osterheider

Raupp

(2019). The diffusion of innovations in agricultural circles: An explorative study on alternative antimicrobial agents. Science Communication, 41(1), 3–37. https://doi.org/10.1177/1075547018819159

24.

de Roo

Amede

Elias

Almekinders

Leeuwis

. (2023). Diffusion of agricultural knowledge in Southern Ethiopia: Finding the real opinion leaders through network analysis. The Journal of Agricultural Education and Extension, 29(1), 99–115. https://doi.org/10.1080/1389224X.2021.1987282

25.

Doumit

Wright

F. C.

Graham

I. D.

Smith

Grimshaw

(2011). Opinion leaders and changes over time: A survey. Implementation Science, 6(1), 117. https://doi.org/10.1186/1748-5908-6-117

26.

Edelman

L. S.

Yang

Guymon

Olson

L. M.

(2013). Survey methods and response rates among rural community dwelling older adults. Nursing Research, 62(4), 286–291. https://doi.org/10.1097/NNR.0b013e3182987b32

27.

Eola-Amity Hills AVA. (n.d.). Our history. https://eolaamityhills.com/our-history/

28.

Farley

Hanbury

Thompson

(2014). Gathering opinion leader data for a tailored implementation intervention in secondary healthcare: A randomised trial. BMC Medical Research Methodology, 14, Article 38. https://doi.org/10.1186/1471-2288-14-38

29.

Fatality Assessment & Control Evaluation. (2021). Farm worker crushed in ATV rollover during herbicide application on a hillside – Oregon (Report 2020OR02). https://www.ohsu.edu/sites/default/files/2021-09/OR-FACE-REPORT%202020OR02%20final%2001Sept2021.pdf

30.

Feder

Savastano

(2006). The role of opinion leaders in the diffusion of new knowledge: The case of integrated pest management. World Development, 34(7), 1287–1300. https://doi.org/10.1016/j.worlddev.2005.12.004

31.

Garimella

De Francisci Morales

Gionis

Mathioudakis

(2018). Political discourse on social media: Echo chambers, gatekeepers, and the price of bipartisanship. In WWW ’18: Proceedings of the 2018 World Wide Web Conference (pp. 913–922). International World Wide Web Conferences Steering Committee. https://doi.org/10.1145/3178876.3186139

32.

Gerbing

D. W.

(2023). Package ‘lessR’. https://cran.project.org/web/packages/lessR/lessR.pdf

33.

Granovetter

M. S.

(1973). The strength of weak ties. American Journal of Sociology, 78(6), 1360–1380. https://doi.org/10.1086/225469

34.

Great Plains Center for Agricultural Health. (2014). Tractor overturns. https://gpcah.public-health.uiowa.edu/resources/tractor-overturns.html

35.

Hair

J. F.

Babin

B. J.

Anderson

R. E.

Black

W. C.

(2018). Multivariate data analysis (8th ed.). Cengage.

36.

Herbstritt

Vazhnik

Fowler

Richard

T. L.

Harvey

Nardone

Kleinman

Fanok

Ernst

Duncan

Hinrichs

Circle

Nicholas

Coulter

Stark

(2019). Establishment of multifunctional riparian buffers (STAC Publication Number 19-008). Chesapeake Bay Scientific and Technical Advisory Committee. https://www.chesapeake.org/stac/wp-content/uploads/2019/12/FINAL_STAC-Report_Multifunctional-Buffers_12.20.2019.pdf

37.

Hunter

J. E.

Gerbing

D. W.

(1982). Unidimensional measurement, second order factor analysis, and causal models. Research in Organizational Behavior, 4, 267–320.

38.

Hurtado

D. A.

Dumet

L. M.

Greenspan

S. A.

Rodriguez

Y. I.

(2018). Social network analysis of peer-specific safety support and ergonomic behaviors: An application to safe patient handling. Applied Ergonomics, 68, 132–137. https://doi.org/10.1016/j.apergo.2017.11.009

39.

Hussain

Mirza

Baines

Burns

Stevens

Asthana

Chatterjee

(2023). Loneliness and social networks of older adults in rural communities: A narrative synthesis systematic review. Frontiers in Public Health, 11, Article 1113864. https://doi.org/10.3389/fpubh.2023.1113864

40.

Jacobs

Lebel

Addams

Matson

McCullough

Garden

Saliba

Finan

(2010). Linking knowledge with action in the pursuit of sustainable water-resources management. Proceedings of the National Academy of Sciences, 113(7), 4591–4596. https://doi.org/10.1073/pnas.0813125107

41.

Kelly

J. A.

St. Lawrence

J. S.

Stevenson

L. Y.

Hauth

A. C.

Kalichman

S. C.

Diaz

Y. E.

Brasfield

T. L.

Koob

J. J.

Morgan

M. G.

(1992). Community AIDS/HIV risk reduction: The effects of endorsements by popular people in three cities. American Journal of Public Health, 82(11), 1483–1489. https://doi.org/10.2105/AJPH.82.11.1483

42.

Klärner

Knabe

(2019). Social networks and coping with poverty in rural areas. Sociologia Ruralis, 59(3), 447–473. https://doi.org/10.1111/soru.12250

43.

Koziol

N. A.

Arthur

A. M.

Hawley

L. R.

Bovaird

J. A.

Bash

K. L.

McCormick

Welch

G. W.

(2015). Identifying, analyzing, and communicating rural: A quantitative perspective. Journal of Research in Rural Education, 30(4), 1–14. https://eric.ed.gov/?id=EJ1059979

44.

Krebs

Holley

(2004, March 21). Building sustainable communities through social network development. Nonprofit Quarterly. https://nonprofitquarterly.org/building-sustainable-communities-through-social-network-development/

45.

Laird-Benner

Ingram

(2010). Sonoran desert network weavers: Surprising environmental successes on the U.S./Mexico border. Environment: Science and Policy for Sustainable Development, 53(1), 6–17. https://doi.org/10.1080/00139157.2011.539943

46.

Lazarfeld

P. F.

Berelson

Gaudet

(1948). The people’s choice: How the voter makes up his mind in a presidential campaign. Columbia University Press.

47.

Lombardi

Lopolito

Andriano

A. M.

Prosperi

Stasi

Iannuzzi

(2020). Network impact of social innovation initiatives in marginalised rural communities. Social Networks, 63, 11–20. https://doi.org/10.1016/j.socnet.2020.04.001

48.

Martinez

D. C.

Way

Moon

Killian

Garrison

(2020). Does a web-first tailored design method work with rural populations? Journal of Rural Social Sciences, 35(2), 5. https://egrove.olemiss.edu/jrss/vol35/iss2/5

49.

Matous

(2023). Male and stale? Questioning the role of “opinion leaders” in agricultural programs. Agriculture and Human Values, 40, 1205–1220. https://doi.org/10.1007/s10460-023-10415-9

50.

Metcalf

A. L.

Angle

J. W.

Phelan

C. N.

Muth

B. A.

Finley

J. C.

(2019). More “bank” for the buck: Microtargeting and normative appeals to increase social marketing efficiency. Social Marketing Quarterly, 25(1), 26–39. https://doi.org/10.1177/1524500418818063

51.

Montes

Jimenez

R. C.

Onnela

J. P.

(2018). Connected but segregated: Social networks in rural villages. Journal of Complex Networks, 6(5), 693–705. https://doi.org/10.1093/comnet/cnx054

52.

Moser

Mosler

H. J.

(2008). Differences in influence patterns between groups predicting the adoption of a solar disinfection technology for drinking water in Bolivia. Social Science & Medicine, 67(4), 497–504. https://doi.org/10.1016/j.socscimed.2008.04.002

53.

Newman

Dale

(2005). Network structure, diversity, and proactive resilience building: A response to Tompkins and Adger. Ecology and Society, 10(1), R2. https://doi.org/10.5751/es-01396-1001r02

54.

Nisbet

M. C.

Kotcher

J. E.

(2009). A two-step flow of influence? Opinion-leader campaigns on climate change. Science Communication, 30(3), 328–354. https://doi.org/10.1177/1075547008328797

55.

Occupational Safety & Health Administration. (2019). Rollover protection and operating tractor safety. www.osha.gov/sites/default/files/publications/OSHA3965.pdf

56.

Pennsylvania Department of Conservation and Natural Resources. (2016). Forest buffers along waterways. https://www.dcnr.pa.gov/Conservation/Water/RiparianBuffers/Pages/default.aspx

57.

Pennsylvania Department of Environmental Protection. (2021). Pennsylvania phase 3 Chesapeake Bay Watershed implementation plan. https://files.dep.state.pa.us/Water/ChesapeakeBayOffice/WIPIII/FinalPlan/PA_Phase_3_WIP_Final.pdf

58.

Project Drawdown. (2025). Drawdown food. https://drawdown.org/programs/drawdown-science/food

59.

Read

D. J.

Wainger

(2023). Assessing intervention effectiveness at promoting voluntary conservation practice adoption in agrienvironments. Conservation Biology, 37(1), Article e14009. https://doi.org/10.1111/cobi.14009

60.

Rogers

E. M.

(2003). Diffusion of innovations (5th ed.). Free Press.

61.

Rogers

E. M.

Cartano

D. G.

(1962). Methods of measuring opinion leadership. The Public Opinion Quarterly, 26(3), 435–441. https://doi.org/10.1086/267118

62.

Rondelli

Casazza

Martelli

(2018). Tractor rollover fatalities, analyzing accident scenario. Journal of Safety Research, 67, 99–106. https://doi.org/10.1016/j.jsr.2018.09.015

63.

Rosseel

Jorgensen

T. D.

Rockwood

Oberski

Byrnes

Vanbrabant

Savalei

Markle

Hallquist

Rhemtulla

Katsikatsou

Barendse

Scharf

(2023). Package ‘lavaan’. https://cran.r-project.org/web/packages/lavaan/lavaan.pdf

64.

Ryan

Gross

N. C.

(1943). The diffusion of hybrid seed corn in two Iowa communities. Rural Sociology, 8(1), 15–24.

65.

Sakshaug

J. W.

Vicari

Couper

M. P.

(2019). Paper, e-mail, or both? Effects of contact mode on participation in a web survey of establishments. Social Science Computer Review, 37(6), 750–765. https://doi.org/10.1177/0894439318805160

66.

Scally

C. P.

Burnstein

Gerken

Immonen

(2020). In search of ‘good’ rural data. Urban Institute. https://www.urban.org/sites/default/files/publication/102134/in-search-of-good-rural-data.pdf

67.

Schultz

W. P.

(2001). The structure of environmental concern: Concern for self, other people, and the biosphere. Journal of Environmental Psychology, 21(4), 327–339. https://doi.org/10.1006/jevp.2001.0227

68.

Silk

K. J.

Hurley

Pace

Maloney

E. K.

Lapinski

(2014). A diffusion of innovations approach to understand stakeholder perceptions of renewable energy initiatives. Science Communication, 36(5), 646–669. https://doi.org/10.1177/1075547014549891

69.

Tian

Zhao

Liu

Lee

(2010). Boosting social network connectivity with link revival. In CIKM ’10: Proceedings of The19th ACM International Conference on Information and Knowledge Management (pp. 589–598). Association for Computing Machinery. https://doi.org/10.1145/1871437.1871514

70.

Tokita

C. K.

Guess

A. M.

Tarnita

C. E.

(2021). Polarized information ecosystems can reorganize social networks via information cascades. Proceedings of the National Academy of Sciences, 118(50), Article e2102147118. https://doi.org/10.1073/pnas.2102147118

71.

Tompkins

E. L.

Adger

W. N.

(2004). Does adaptive management of natural resources enhance resilience to climate change? Ecology and Society, 9(2), 10. https://doi.org/10.5751/es-00667-090210

72.

U.S. Environmental Protection Agency. (2010). Chesapeake Bay total maximum daily load for nitrogen, phosphorus and sediment. https://www.epa.gov/sites/default/files/2014-12/documents/cbay_final_tmdl_exec_sum_section_1_through_3_final_0.pdf

73.

Valente

T. W.

Pumpuang

(2007). Identifying opinion leaders to promote behavior change. Health Education & Behavior, 34(6), 881–896. https://doi.org/10.1177/1090198106297855

74.

Vance-Borland

Holley

(2011). Conservation stakeholder network mapping, analysis, and weaving. Conservation Letters, 4(4), 278–288. https://doi.org/10.1111/j.1755-263X.2011.00176.x

75.

Vigoroso

Caffaro

Cavallo

(2019). Warning against critical slopes in agriculture: Comprehension of targeted safety signs in a group of machinery operators in Italy. International Journal of Environmental Research and Public Health, 16(4), 611. https://doi.org/10.3390/ijerph16040611

76.

Vitis Terra. (n.d.). https://www.facebook.com/VitisTerra/

77.

Winter

Neubaum

(2016). Examining characteristics of opinion leaders in social media: A motivational approach. Social Media+ Society, 2(3), 2056305116665858.

78.

Winter

Neubaum

Stieglitz

Ross

(2021). #Opinionleaders: A comparison of self-reported and observable influence of Twitter users. Information, Communication & Society, 24(11), 1533–1550. https://doi.org/10.1080/1369118X.2019.1705374

79.

Song

(2018). User engagement in public discourse on genetically modified organisms: The role of opinion leaders on social media. Science Communication, 40(6), 691–717. https://doi.org/10.1177/1075547018806526

80.

Young

J. M.

Hollands

M. J.

Ward

(2003). Role for opinion leaders in promoting evidence-based surgery. Archives of Surgery, 138(7), 785. https://doi.org/10.1001/archsurg.138.7.785

81.

Zhang

A. J.

Matous

Tan

D. K.

(2020). Forget opinion leaders: The role of social network brokers in the adoption of innovative farming practices in North-western Cambodia. International Journal of Agricultural Sustainability, 18(4), 266–284. https://doi.org/10.1080/14735903.2020.1769808