I Do Not Think It Means What You Think It Means: Explorations of Mental Models of Soil Health

Final report for GNC18-272

Project Type: Graduate Student
Funds awarded in 2018: $11,810.00
Projected End Date: 11/30/2020
Grant Recipient: The Ohio State University-Wooster Campus
Region: North Central
State: Ohio
Graduate Student:
Faculty Advisor:
Dr. Steven Culman
Ohio State University
Expand All

Project Information



The paradigm of soil health has become widely embraced in recent years by a variety of stakeholders. Increased discussion of soil health among these entities and in the farm press illustrates that soil heath knowledge is transmitted through complex networks of information exchange, rather than originating from a single source. Thus, there is a diverse set of stakeholders that are both consuming and/or disseminating soil health information. This is emblematic of a transition from the traditional model—where knowledge is thought to be created by scientists and consumed by farmers— to a more interactive, network-based model where farmers and researchers stand on more equal footing. As these networks broaden, they include increasingly diverse sets of stakeholders, who may conceptualize soil health distinctly from one another.

This project—“ I do not think it means what you think it means: Explorations of mental models of soil health”—seeks to describe the mental models of soil health within and across groups in order to facilitate communication and cooperation across these stakeholders. This will result in knowledge creation that is simultaneously valued by all stakeholders and actionable by farmers, ultimately increasing farmer adoption of practices that improve soil health. To further this goal, we will specifically target the mental models of three distinct groups of agricultural stakeholders: farmers, Natural Resource Conservation Service (NRCS) employees, and agricultural researchers involved in the fields of agronomy and/or soil science. The three primary objectives are to: 1) evaluate the perceived importance of soil health to farmers by each group, 2) compare mental models of soil health, and 3) determine the self-reported use and value of current soil testing methods.


To describe these mental models, we used a two-stage survey development. First, we performed 30 to 60 minute semi-structured phone interviews with 5 to 10 stakeholders from each group. From these interviews, we identified broad areas of conceptual agreement across groups, ultimately determining 6 emergent concepts surrounding soil health: soil fertility and biological functioning, soil physical functioning, crop productivity, profitability, environmental harm, and agrochemical inputs. From these emergent concepts, we used a mix of paper mail surveys (for farmers) and email (for NRCS employees and agricultural researchers) to understand perceived relationships between these concepts and soil health. We used multiple choice questions to ask what effect an increase in soil health has on each emergent concept, with the provided responses as “increases” (+1), “decreases” (-1), or “they are not related” (0). In order to differentiate between bidirectional (i.e. reciprocal relationships) and unidirectional relationships, we then asked this question in reverse: what effect an increase in each emergent concept has on soil health, with the same answer choices as before. For example, “what effect does an increase in soil health have on crop productivity?” would correspond to “what effect does an increase in crop productivity have on soil health?”. Thus, our respondents indicated which concepts are affected by soil health—which we refer to as unidirectional relationships (i.e. soil health → emergent concept) and which concepts also affect soil health—which we refer to as bidirectional relationships (i.e. soil health ↔ emergent concept).

In addition to understanding each group’s conceptualizations of soil health, we were also interested in perceived soil health importance to farmers and the use and efficacy of various types of soil testing. First, we asked farmer respondents to self-identify the priority of soil health on their farm from 0 (not a priority) to 10 (highest priority). To understand if this prioritization is perceived by non-farmer groups, we also asked NRCS and researcher respondents to evaluate the degree to which they believe farmers prioritize soil health. Lastly, to determine the efficacy of soil tests, we asked all groups if they had used various types of soil tests to evaluate soils (either on their farm or for their work). For respondents who used a given test, we then asked them to rate the quality of information from that test, ranging from 1 (not at all helpful) to 5 (extremely helpful).

Using our findings from the survey, we conducted focus groups with stakeholder groups of interest to verify our interpretation of the survey, communicate about each group’s mental models, and to gather qualitative data on soil health conceptualizations. Focus groups were also used to evaluate changes in understanding of both in-group and out-group mental models of soil health.

Findings and Conclusions
Obj 1 &2: Soil health prioritization and conceptualization
uWe found that soil health is a potentially useful communication and framing tool. In particular, farmers prioritize soil health to a greater extent than other groups realize (Figure 1), although future work is needed to better understand the cause of this underestimation. While there are areas of shared conceptualizations of relationships between soil health and other production components, there are certain areas that differ systematically across groups (i.e. environmental harm and agrochemical inputs) (Figure 2). Interestingly, the conceptualizations of researchers were generally more closely aligned with those of farmers (5 out of 6 emergent theme) than were the conceptualizations of NRCS employees (4 out of 6 themes). Points of divergence in soil health conceptualizations represent potentially unexplored and complex areas for researchers and NRCS personnel to communicate with farmers. Therefore, when communicating about soil health across groups, framing and communication strategies should be tailored to match the target group’s conceptualization. Our results also highlight the value of soil health as a framework within which to discuss agronomic, environmental, and potentially economic considerations of on-farm decision making.

As an example of how to implement these findings, if a researcher wanted to communicate with farmers about the benefits of soil health, discussing the effect of soil health on agrochemical inputs would likely be impactful. Both researchers and farmers are more likely to see this relationship as unidirectional and therefore more straightforward. However, if a researcher used the framing of soil health to communicate with NRCS personnel about agrochemical inputs, our results suggest that the NRCS personnel are more likely to see environmental harm as both influencing and being influenced by soil health. This reciprocal relationship could stymy communication due to individualized conceptualizations of this relationship and the high potential for unintended (yet inferred) feedback loops. That is, the audience could be inferring an additional relationship due to their differing conceptualization. For concepts with broad agreement (e.g. crop productivity and profitability), communication across groups is more likely to be successful and straightforward. Therefore, careful consideration of message framing is need to match communication goals and the specific audience.

Obj 3: Soil test use and valuation
Overall, standard agronomic soil tests (e.g., pH, organic matter, extractable nutrients) were the most commonly used type of soil test and provided the most useful information for both farmers and researchers (Figure 3). This was followed by in-field soil tests (e.g., shovel, by feel, etc.) which were generally less common and less useful to these groups. However, these rankings were opposite for NRCS personnel, who were most likely to use in-field tests and found those the most useful. Although there were slight differences between groups, both in-field and agronomic tests were considered quite useful by all groups, often scoring at least 3.75 out of 5. Soil health tests were the least used by all groups. NRCS personnel and researchers ranked them as the least useful of the three tests, whereas farmers ranked them as more useful than in-field tests and less useful than agronomic tests.

Greater frequency of use of soil tests was generally correlated with higher reported usefulness (r = 0.96, p < 0.001; Figure 3), with the exception of farmer responses regarding soil health tests. Farmer reports of the usefulness of soil health tests were anomalously high (3.92 out of 5) given their low usage (14.8%). The close relationship between use and valuation is noteworthy, but it is unclear how or if this is causal. We cannot determine if respondents were more likely to use a soil test they deemed valuable or if tests were deemed more valuable because respondents used it more frequently. However, our results imply that low farmer use of soil health tests is not necessarily due to a lack of perceived utility. This suggests other barriers to farmer use of soil health tests, such as availability or cost of these tests.

Our results also have important implications for the development and refining of soil health tests. Conceptually, farmers see a linkage between soil health and agrochemical inputs and crop productivity. However, only a handful of studies have explored these linkages using common soil health tests, despite farmers demonstrating a greater willingness to implement information derived from a soil test than from non-test sources. Therefore, although the overall usage of soil health tests was low amongst farmers, the information farmers glean from those tests is both valuable and more likely to be implemented. This indicates that soil health testing could be effectively and broadly utilized if barriers to adoption were better understood and mitigated. An understanding of barriers to adoption will help ensure that current large-scale projects measuring and vetting soil health tests will be widely implemented by farmers.

Adoption and Learning Outcomes
The first learning outcomes of this project was to validate the survey results using qualitative data from focus groups. Focus groups showed that while the underlying explanation for conceptual linkages may vary slightly from individual to individual, our survey largely provided an accurate representation of the general mental models of each group. This is an important step towards ensuring that we have appropriately interpreted our survey results.

A second learning outcome was to increase the knowledge of stakeholders about soil health conceptualizations within their group (in-group) as well as in other groups (out-group). Although we were unable to conduct an NRCS focus group within the project timeline due to a combination of geographic and institutional obstacles, 100% of participants in both the farmer and researcher focus groups reported an increasing in learning about in-group soil health conceptualizations. Similarly, 100% of farmers and 50% of researcher participants reported increased understanding of out-group conceptualizations of soil health. Using the benchmarks from the project’s inception, we consider this learning outcome a success. Farmers generally agreed with one another and therefore none of them (0%) reported changing their own conceptualization soil health. However, the differing conceptualizations among researchers led to 50% of them reporting changes in their own conceptualizations of soil health. Therefore, while the learning objective was achieved for both groups, only one reported a change in attitudes.

Collectively, this implies that within-group heterogeneity of soil health conceptualization is not evenly distributed across groups. That is, some groups are more similar in their soil health conceptualizations than others. Specifically, researchers differ from one another more than farmers differ from one another. This may contribute to the agreement and ensuing lack of attitude change reported in our farmer focus group.

Project Objectives:

This project has four explicit learning outcomes. These outcomes are: 1) to describe mental models for stakeholder groups surrounding the term “soil health”, 2) determine similarities and differences in these mental models of “soil health” amongst stakeholders, 3) validate the information gleaned from these models through qualitative methods, and 4) increase knowledge of stakeholders of in-group and out-group mental models surrounding soil health. Learning Outcomes 1) and 2) were primarily accomplished by project leaders through a survey that contains closed- and open-response questions. As discussed previously (Summary > Adoption and Learning Outcomes), Learning Outcome 3 and part of Learning Outcome 4 was achieved; our initial dataset was validated for use and focus group participants reported increased knowledge of soil health conceptualizations.

Learning Outcome 4 is (as yet) incomplete. Specifically, we have not completed the desired number of public-facing outputs that we intended. This is attributable to several overlapping phenomena. First, the peer-reviewed journal article has not been officially published yet. The article has received reviewer and editor comments in the Journal of Soil & Water Conservation, which (if published) will target both researchers and NRCS personnel. There have been numerous delays in peer review with the pandemic and we expect this to be completed within the next 6 months. Next, publication in farmer trade magazines and extension-facing publications has not been finalized yet. We have been in touch with the magazines, but until the peer-reviewed article has been published, we will likely get little traction. In the meantime, we have made an effort to reach the farmer and extension-facing audiences by (1) publishing a blog post on the Soil Health Nexus blog, which has a soil health-minded audience within the north-central region and (2) sending postcard summaries of our findings to each of our survey participants. We will continue these efforts at disseminating information upon the completion of publication of our peer-reviewed article.


Materials and methods:

We developed surveys in two stages. First, we conducted semi-structured phone interviews of 30 to 60 minutes with 5 to 10 members from each group of interest. Using content analysis of these interviews, we identified 19 themes, each of which were discussed by at least two interviewees. To minimize redundancy in the survey, we consolidated similar themes, ultimately establishing six emergent themes related to soil health: soil fertility and biological functioning, soil physical functioning, crop productivity, profitability, environmental harm, and agrochemical inputs.

To describe the mental models surrounding soil health, we developed a survey to ask participants how these emergent themes related to one another and to soil health. We then sent out surveys to 650 farmers, 1,427 NRCS employees, and 381 agricultural researchers from ten states in the Midwest: Ohio, Michigan, Indiana, Illinois, Iowa, Wisconsin, Minnesota, Nebraska, North Dakota, and South Dakota. We purchased the farmer survey sample from Farm Market ID (farmmarketid.com), which was comprised of corn, soybean, and wheat farmers equally represented by farm size: 50-250 acres, 500-1000 acres, and 1000+ acres. We compiled NRCS personnel contacts from publicly available databases, including all employees with “soil” or “agronomy” in their position title. Agricultural researcher contacts were compiled from departmental and extension websites of 1862 Land Grant institutions. Inclusion was based on stated research expertise where available, as well as the topic of publications from 2013 to 2018. Farmer surveys were sent by mail, while NRCS and researcher surveys were sent via email. Farmers could complete the survey on paper, but all participants had the option to complete the study online. We received complete, usable responses from 89 farmers, 134 NRCS employees, and 42 researchers, giving response rates of 13.7%, 9.4%, and 11.0%, respectively.

First, we asked farmer respondents to rank the priority of soil health on their farm from 0 (not a priority) to 10 (highest priority). Similarly, we asked NRCS and researcher respondents to evaluate the degree to which farmers prioritize soil health. Second, to better understand the mental models surrounding soil health within each group, we wanted to explore perceived causal linkages among emergent themes, including soil health. For each pair of themes (X,Y), multiple choice survey questions asked what effect an increase in X causes on Y, where the provided responses were “increases” (+1), “decreases” (-1), or “they are not related” (0). We used each respondent’s answers to these answers to construct a network comprised of causal linkages among the themes (i.e. depict the individual’s mental model). Third, to determine the frequency of use of soil tests, we asked all groups if they had used standard agronomic soil tests, soil health tests, or in-field tests to evaluate their soils in the last three years. For respondents who had used a given test, we then asked them to rate the quality of information from that test, ranging from 1 (not at all helpful) to 5 (extremely helpful).

Statistical analyses

Mental models were represented as directed networks. Links between nodes in that directed network represent a causal path that connects these themes or concepts. In network analyses, the importance of a given node is often evaluated using network centrality measures, which can be quantified in several ways. The betweenness of a node is a measure of its presence in the shortest path or set of linkages between other nodes, i.e. its importance in connecting themes that would otherwise be linked less directly. Betweenness is therefore interpreted as the tendency of a node to exclusively mediate causal paths. Information centrality is a measure of the number of links originating with a given node and is interpreted as the tendency of a node to initiate causal paths within the network. Prestige is a measure of the incoming links to a given node and is interpreted as the importance of that node.

Research results and discussion:

There are several notable outcomes from our survey:

  • Soil health is more important to farmers than others realize: Our findings show that soil health is a high priority to farmers in their on-farm operations, but NRCS personnel and researchers undervalue this importance to farmers. Farmers rated soil health an 8.5 out of 10 priority on their farms (lowest reported value was a 5), while both NRCS personnel and researchers had statistically lower ratings for farmers—4.9 and 5.7, respectively. The attitudes informing this systematic undervaluation were captured in a comment from one NRCS respondent who said, “I think the ‘how valuable is soil health to Farmers’ question could be split into two questions. How valuable is it to them and how valuable should it be. I answered 3 in your survey which might be high. How valuable should it be….12!”. These results illustrate that farmers prioritize improving soil health to a degree not recognized by educators or researchers.
  • Soil health is equally prominent in mental models across groups: Our analysis showed limited variation between members of the three groups in the centrality of soil health in their mental models. On the whole, soil health was similarly prominent in each group’s mental models. However, a large degree of within-group heterogeneity may obscure potentially meaningful differences. Previous work has shown that farmer attitudes and beliefs are highly heterogeneous, but there is a lack of data on NRCS and agricultural researchers as distinct groups, our results indicate similar within-group heterogeneity in their conceptualizations of soil health.
      Information Centrality Prestige Betweenness
    Farmers 0.144 (0.001) 0.317 (0.031) 0.242 (0.028)
    NRCS 0.145 (0.001) 0.202 (0.025) 0.296 (0.024)
    Researchers 0.146 (0.001) 0.252 (0.045) 0.216 (0.041)
    Effect size (F-ratio) 2.23 (p = 0.109) 0.73 (p = 0.483) 1.93 (p = 0.148)
  • Group-level mental models differ in the connections: Consensus graphs show linkages that have 80% agreement within a group. Consensus graphs showed that the overall structure of NRCS and researchers’ mental models of soil health diverged from farmer mental models of soil health (Figure 1). Most notably, farmer mental models showed soil health as the only connection to environmental harm or agrochemical inputs—both of which were negative causal relationships. Conversely, NRCS and researchers perceived many connections to these concepts, indicating multiple potential connections to be leveraged when communicating on these topics. This position of soil health in causal paths to environmental harm and agrochemical inputs contributes to a higher betweenness values of soil health in farmer consensus mental models than in researcher or NRCS consensus mental models. Moreover, soil health had a higher information centrality in farmer mental modes. Collectively, this suggests that soil health is more likely to be the initiator and intermediary in farmer mental models than for NRCS personnel or agricultural researchers. However, soil health was only of median importance to farmers in terms of receiving linkages (i.e. prestige), relative to NRCS personnel and researchers. Given the qualitative nature of these findings, we caution against overgeneralization and encourage future work to identify potential drivers of between- and within-group heterogeneity in mental models.
    Consensus graphs of mental models for (a) farmers, (b) Natural Resource Conservation Service (NRCS) personnel, and (c) agricultural researchers. Each linkage represents  80% agreement on the presence, directionality, and nature (i.e. positive or negative value) of a tie. Solid blue lines indicate a positive relationships (+1) and dashed red lines indicate a negative relationship (-1). Lighter ties indicate bidirectional relationships. The threshold of 80% agreement was chosen to appropriately balance within-group agreement and between-group differences. Note that node placement and distance between nodes is based on ease of visualization and should not be interpreted as conceptual closeness or centrality within the network.
  • Conclusions: Our ten-state survey of Midwestern farmers, NRCS personnel, and agricultural researchers showed that soil health is equally central in mental models of all three groups. Despite the consistent centrality of soil health, mental models were differently structured between groups. Notably, soil health was the sole link to several themes with important environmental and agronomic implications. Paired with the consistent underestimation of the importance of soil health to farmers by NRCS personnel and agricultural researchers, this work suggests that using the language of soil health may be an undervalued way to communicate with farmers on these topics. These findings are essential for informing extension programming and communication around soil health. Future work is needed to understand if and how this communication can lead to changes in decision-making or adoption of preferred management techniques.
Participation Summary
100 Farmers participating in research

Educational & Outreach Activities

1 Journal articles
1 Published press articles, newsletters
4 Presentations and posters

Participation Summary

5 Farmers
6 Ag professionals participated
Education/outreach description:

Thus far, education and outreach activities have taken several forms. 

Focus Groups

  • One farmer focus group: 5 attendees; >90 minutes. Overall discussion was lively and enthusiastic. Farmers expressed their overall dedication to improving soil health, as well as the benefits of doing so. They also exchanged ideas about challenges and pitfalls they experienced and strategies to overcome obstacles.
  • One researchers focus group: 6 attendees; 60 minutes. Discussion was constructive and informative. Researchers expressed uncertainty about how to define soil health, as well as skepticism about generalities surrounding it. Lots of discussion about contexts where specific soil health practices would/wouldn’t work.


  • Soil Science Society of America meeting (video recording): ~100 attendees; 15 minutes with additional 15 minutes of questions/discussion; many people expressed interest in the results and were surprised to learn that farmers place a high priority on soil health. Discussion on the implications of research findings.
  • Soil Research Symposium (Ohio State University): ~80 attendees to 2-hour poster session; Several of the graduate students seemed excited about the work, and a researcher at the Wooster campus was asking about potential overlap and coordination efforts with their project on organic farmers. The faculty seemed intrigued as well, although it was apparent that many of them had never considered the importance of the definition of soil health before.
  • Hayes Graduate Forum (Ohio State University): poster session; received 1st place; invited abstract for submission into Hayes Knowledge Bank; judges and other graduate students commented that the project was very clear and accessible but seemed very important to improve communication efforts. (abstract and summary)
  • Society for Risk Analysis Meeting: ~30 people stopped to read the poster; people were impressed and intrigued by the methodology used. Researchers from the University of Washington and the University of Albany wanted to use the methodology to inform other studies.
  • Department of Crop Sciences seminar (University of Illinois, Urbana-Champaign): ~20 attendees; 30 minutes; attendees expressed skepticism of results due to sample size, extensive discussion on personal definitions and operationalizing of “soil health”

Other Outreach

  • Survey participants: We have sent our ~100 postcard summaries of our findings to survey participants and other interested parties.
  • Blog post: we have published a summary of our findings on the Soil Health Nexus blog 

Ongoing Outreach

  • Peer-reviewed journal article: one open access, peer reviewed journal article is in review with Journal of Soil & Water Conservation. At least one other article is planned.
  • Trade journals: pending completion of peer-review, we will convert the contents of the journal article into a format that is amenable to trade journals or online publications. These will largely mirror the Soil Health Nexus blog. Targeted outlets include the Ohio C.O.R.N Newsletter, Progressive Farmer, Corn & Soybean Digest, and AgClips Daily.

Project Outcomes

5 Farmers reporting change in knowledge, attitudes, skills and/or awareness
3 New working collaborations
Project outcomes:

Our research shows that the paradigm of soil health unites disparate populations. Thus, soil health can be a unifying approach to sustainable agroecosystem management. Thus far, this insight seems to be novel and thrilling to stakeholders and audiences. Moreover, our results highlight areas of overlap and differences in terms of how the soil health paradigm is operationalized. These results can be easily and seamlessly integrated into communication strategies and extension programming. Specifically, it suggests that using the language of soil health can be used to discuss difficult topics (e.g. environmental harm) with farmers. By leveraging a shared language, we could improve working relationships and social connection between farmers and other agricultural stakeholders, ultimately striving towards the shared long-term goal of agricultural sustainability.

Knowledge Gained:

Survey results speak directly to changes in attitude and knowledge in our collective knowledge of soil health. Since initiating this project, all of our findings have been nearly the inverse of our initial hypotheses. We expected soil health to be much more prominent in farmer mental models than either NRCS personnel or researchers, but our findings showed that there was nearly-equal prominence. Additionally, we expected farmers and NRCS personnel to have a more accurate perception of farmer prioritization of soil health, but found that these groups systematically undervalue how high of a priority soil health is to farmers.

Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the U.S. Department of Agriculture or SARE.