Volume 5, Issue 3 e12902
Open Access

Mapping student understanding of bees: Implications for pollinator conservation

Shannon M. Cruz

Corresponding Author

Shannon M. Cruz

Department of Communication Arts & Sciences, Pennsylvania State University, University Park, Pennsylvania, USA


Shannon M. Cruz, Department of Communication Arts & Sciences, Pennsylvania State University, 219 Sparks Building, University Park, PA 16802, USA.

Email: [email protected]

Search for more papers by this author
Christina M. Grozinger

Christina M. Grozinger

Department of Entomology, Center for Pollinator Research, Huck Institutes of the Life Sciences, Pennsylvania State University, University Park, Pennsylvania, USA

Search for more papers by this author
First published: 11 February 2023

Funding information: Communication, Science, and Society Initiative (CSSI), Pennsylvania State University, Grant/Award Number: N/A


Global declines in populations of several bee species have highlighted the importance of efforts to conserve bees and other pollinators. Because research on the social dimensions of pollinator conservation is limited, however, developing clear strategies to promote conservation behaviors remains a challenge. In an effort to contribute to understanding of these social dimensions of conservation, we used semantic network analysis and content coding to investigate knowledge and understanding of bees among U.S. college students with either a low (n = 233) or high (n = 93) interest in this topic. Results revealed that both groups' understanding of bees was organized around their utilitarian value for humans, particularly honey production. Furthermore, although student knowledge of bees was fairly accurate, it was not very sophisticated. Knowledge about honey bees was also more accurate than knowledge about bees in general. Implications for future conservation and education efforts are discussed.


There is rising concern from members of the public, growers, and policymakers about reported dramatic declines in populations of wild and managed bees worldwide (Bruckner et al., 2020; IPBES, 2016; Pollinator Health Task Force, 2015). In the United States, 30% of honeybee colonies die each winter despite the management efforts of beekeepers (Bruckner et al., 2020), and half of studied wild bumble bee species are showing significant declines (Cameron et al., 2011). Globally, analysis of bee occurrence data suggests that a quarter of all bee species are showing declines (Zattara & Aizen, 2021). These declines are especially concerning given bees' importance for plant reproduction, and thus for agricultural crop production and sustaining terrestrial food webs (Rodger et al., 2021). Approximately 80% of flowering plant species, corresponding to three-quarters of major global food crops—including the fruits, vegetables, and nut crops that provide critical micronutrients for human diets—benefit from the pollination services of animals, with bees being the most important pollinators (Eilers et al., 2011; Klein et al., 2006; Rodger et al., 2021). These pollination services contribute to nearly $34 billion in agricultural value in the United States (Jordan et al., 2021) and $361 billion worldwide (Lautenbach et al., 2012). Beyond their importance for food production and supporting biodiversity, bees contribute broadly to human health and well-being, and are key features in cultures across the world (Potts et al., 2016).

Addressing bee species declines will require public engagement in conservation behavior (Schultz, 2011). However, there is limited research on how best to promote such behavior (Hall & Martins, 2020; Knapp et al., 2020). Many existing studies (e.g., Bhattacharyya et al., 2017; Kalaman et al., 2020; Schönfelder & Bogner, 2018) and initiatives (Marselle et al., 2021) have focused on education as the solution to this problem, but education alone often fails to produce attitude and behavior change (Schultz, 2011; Simis et al., 2016; Sturgis & Allum, 2004). There are many reasons why someone may support or oppose a particular policy or course of action, regardless of their level of knowledge (e.g., see Knapp et al., 2020; Turo & Gardiner, 2019). Meaningful attitude and behavior change, accordingly, is likely to require not just the provision of information about how to protect pollinators, but the development of communication campaigns that persuade people they should protect them (see Dörge et al., 2022; Priest, 2019).

A focus on persuasion, however, does not imply that investigating public knowledge of bees is unimportant. Persuasive messages are likely to be most effective if they focus on what people already understand to be the most important and central issues about a given topic (Russell & Reimer, 2018), so studies of knowledge are a vital precursor to message design.

1.1 Existing studies of bee knowledge

There are over 20,000 bee species in the world, with over 4000 in the United States alone (Danforth et al., 2019). These bees vary dramatically in their appearance, behavior, and life history traits. Honey bees form social colonies of up to 50,000 individuals, and have been managed by humans for centuries for honey production, and more recently for pollination of agricultural crops (Delaplane & Mayer, 2000; Kritsky, 2017; Tautz, 2008). The vast majority of bee species, on the other hand, are solitary, with 70% living in underground nests which are cryptic and not amenable to research or management (Danforth et al., 2019).

Previous studies suggest that the public has a very limited awareness of this breadth or diversity of bee species (Wilson et al., 2017). Studies have investigated knowledge in a number of groups, including farmers and growers (Hanes et al., 2018; Tarakini et al., 2020), master gardeners (Kalaman et al., 2020), consumers (Khachatryan & Rihn, 2018), beekeepers (Maderson & Wynne-Jones, 2016; Penn et al., 2019), and students (Penn et al., 2020; Schlegel et al., 2015), but find that knowledge and awareness are consistently low (e.g., Elisante et al., 2019; Penn et al., 2020; Sawe et al., 2020; Wilson et al., 2017). Furthermore, people tend to have better knowledge of honey bees than of other bee species (Hall & Martins, 2020; Kasina et al., 2009; Wilson et al., 2017), even though they are not representative of most species or of native bees in many parts of the world (Colla & MacIvor, 2017).

Still, these previous investigations may tell us more about what people do not know than what they do know. Common methods of assessing knowledge include factual questions and species identification tasks (e.g., Bhattacharyya et al., 2017; Mata et al., 2019; Schönfelder & Bogner, 2017), with many fewer studies (e.g., Golick et al., 2018) examining understanding in more depth. In order to develop persuasive messages that are effective even among people with low knowledge, more such in-depth investigations are needed.

1.2 A novel approach to examining knowledge

One approach that may be particularly useful for building on previous studies of bee knowledge is semantic network analysis. Semantic network analysis involves mapping the relationships among different words or concepts (Danowski, 1993; Doerfel, 1998). Specifically, different concepts (e.g., honey) are treated as the nodes in the network and the connections between concepts (e.g., makehoney) are treated as the edges. Then, once the network is constructed, network analytic techniques can be used to gain insights about the structure of the semantic network. This approach can be used on any form of text, such as social media posts (Kang et al., 2017) or even the topics of conference papers (Doerfel & Barnett, 1999).

Asking people to express in words what they know about a topic—such as bees—makes it possible to explore their knowledge and understanding of that topic using semantic network analysis. This approach has important advantages over other approaches to knowledge assessment. For one, semantic network analysis can reveal not only the content of the public's knowledge—what concepts they know—but its structure—the way that they connect those concepts to one another. This structure is important because it provides an indication of how the concepts are organized cognitively (Danowski, 1993), which can reveal which are most important or salient in the public's thinking. For instance, concepts may be important because they lead to many other concepts, follow from many other concepts, or help connect many different concepts to one another (Newman, 2018). This information gives a deeper picture of knowledge than can be obtained from many other approaches.

Furthermore, analyzing knowledge using semantic network analysis has direct applications for persuasion. Recent work in communication science, drawing on an approach called probabilistic persuasion theory (PPT; Reimer et al., 2012), suggests that arguments that incorporate more central concepts in a semantic network will be stronger and more persuasive than arguments that incorporate more peripheral concepts (Russell & Reimer, 2019, 2020). The implication is that if we use semantic network analysis to understand which concepts are most central to public knowledge and understanding of bees, we can also gain insights into which concepts will be most effective to focus on when developing persuasive messages to promote their conservation. To illustrate, an argument promoting bee conservation could focus on many properties of bees that people might find worthwhile, including their importance to food security, importance to ecosystem function, intrinsic value, accessibility, beauty, and complex social lives. Without any information about how people organize their thinking about these concepts, however, it is difficult to know which would be most useful to focus on. With semantic network analysis, we can identify which concepts are most central, and thereby select those with the greatest potential for generating attitude and behavior change.

1.3 The present study

With the preceding points in mind, our goal in this investigation was to build on previous studies of public knowledge and understanding of bees. Specifically, we employed semantic network analysis to examine the content and structure of self-reported student knowledge of bees. In particular, we sought to identify the most central concepts in this network, in order to identify those that might be the most promising foundation for persuasive messages on this topic. We also examined whether central concepts differed in two audiences: (1) a high-interest group enrolled in a class specifically on honey bees and (2) a low-interest group enrolled in psychology classes. High-interest groups, such as people who choose to be involved in citizen science projects, have been the subject of several studies in the past (Deguines et al., 2018; Domroese & Johnson, 2017; MacPhail et al., 2020; Toomey & Domroese, 2013) and represent an important target for messages promoting conservation. However, broad scale conservation initiatives are unlikely to be successful if they cannot reach members of the general public with less prior interest in bees and other pollinators. Designing messages for such low-interest audiences is thus equally, if not more, important.

As a complement to this analysis, we also coded the accuracy of students' statements about bees and the extent to which the statements focused on honey bees. These connect to common themes in previous studies of bee knowledge (Hall & Martins, 2020) and gave us additional insights into the students' understanding of bees. The formal research questions for this investigation were:

RQ1.(a) What concepts are most important (central) to students' understanding of bees, and (b) does this differ based on level of interest (enrollment in a course on honey bees vs. courses in psychology)?

RQ2.(b) How accurate is student knowledge of bees, and (b) does this differ based on level of interest (enrollment in a course on honey bees vs. courses in psychology)?


2.1 Sample and procedure

During the first week of the 2020–2021 academic year, college students were recruited to participate anonymously in the study with the incentive that they could receive a small amount of course credit for their participation. The high-interest group of students (n = 93) was recruited from a course entitled Honey Bees and Humans. This is a lower-level course offered by the Entomology Department which can count toward students' general education requirements, and it attracts a broad range of majors from across the university. The low-interest group of students (n = 233) was recruited from several lower-level psychology courses. For demographic information about each group, see Table 1. Note that the students were recruited to the survey and participated before they participated in any of the lectures, and thus the students in the Honey Bees and Humans course did not have any course-based knowledge of bees.

TABLE 1. Demographic information.
High-interest group (entomology, n = 93) Low-interest group (psychology, n = 233)
Male 29.0% 21.0%
Female 69.9% 78.5%
Year in school
Freshman (first year) 30.1% 15.9%
Sophomore (second year) 31.2% 34.3%
Junior (third year) 12.9% 32.2%
Senior (fourth year) 21.5% 15.0%
5th year or above 4.3% 2.6%
Political affiliation
Democrat 44.1% 45.5%
Republican 19.4% 21.0%
Independent 20.4% 25.8%
Other 15.0% 7.7%
Freq. religious service attendance
Never 49.5% 42.9%
About once a year 22.6% 27.5%
About once a month 12.9% 19.7%
About once a week 14.0% 8.6%
Several times a week 1.1% 1.3%
Age (years) M = 19.74, SD = 3.32 M = 20.74, SD = 4.75
Political ideology M = 4.68, SD = 1.65 M = 4.62, SD = 1.64
  • Note: Political ideology was measured on a 7-point scale (1 = strongly conservative, 7 = strongly liberal).

Data were collected online using Qualtrics. The link to the survey instrument was sent to instructors, who then shared it with their students. Students who clicked on the link and gave their informed consent to participate were given the prompt: “What do you know about bees? Please write down anything that comes to mind.” Students could write as much or as little as they wanted to in response. After responding to the prompt, students filled out other survey measures that were part of a separate project, then finished by responding to the demographic questions. All procedures were approved by the university's Institutional Review Board (study #00015857) prior to the start of data collection.

2.2 Semantic network analysis

To create semantic networks for the high- and low-interest group, responses to the prompt were first corrected for spelling errors and edited to replace different ways of describing the same concept (e.g., pollinate, pollinating, pollination, pollinator) with a single, common term (e.g., pollination). These steps helped ensure that different words genuinely represented different concepts, rather than simply variations in spelling, grammar, or word choice. A drop list of words to exclude from the analysis was also developed in advance. The drop list included words that were not relevant for understanding important concepts, such as articles (e.g., the, an, a), pronouns (e.g., I, me, her, he), and “to be” verbs (e.g., am, is, are).

Next, the edited responses were analyzed using the WORDij 3.0 software program (Danowski, 2013). The program works by reading the text of the responses word by word and keeping track of how often each word appears (word counts) and how often two words co-occur (word pair counts), then generating matrices of the word co-occurrences. As it does, the program also preserves word order, and thus identifies directional (rather than nondirectional) ties between words in the semantic network. For example, the word pair “love bees” would be counted separately from the word pair “bees love,” reflecting the fact that these two pairs have very different meanings. For the purposes of this analysis, a three-word window was used to identify word pairs, meaning the program identified two words as occurring together any time they fell within three words of one another in the text. Uncommon words and pairs (words appearing fewer than three times and word pairs appearing fewer than two times) were also dropped from the matrices. For ease of interpretation, the matrices generated by WORDij were converted into binary matrices before network analyses were performed.

Finally, in order to identify the most central concepts for each group and visualize the semantic networks, the directional matrices were analyzed using UCINET 6.721 statistical software (Borgatti et al., 2002). The centrality of words in the semantic networks was evaluated in three ways: indegree centrality, outdegree centrality, and betweenness centrality. Indegree centrality evaluates the importance of a concept in a network by estimating how often it appears after other concepts (Freeman, 1978), and is measured by adding up the number of words in the semantic network with directional ties flowing into a focal word. Outdegree centrality evaluates the importance of a concept by estimating how many other concepts it leads to (Freeman, 1978), and is measured by adding up the number of words in the semantic network with directional ties flowing out from a focal word. Betweenness centrality evaluates the importance of a concept by estimating how often it connects two concepts that would not otherwise be connected (Freeman, 1978), and is measured by adding up the number of shortest paths between two words in the semantic network that run through a focal word. All measures of centrality were normalized to facilitate comparisons between words and groups (Borgatti et al., 2018).

2.3 Knowledge coding

Responses to the prompt were coded manually to evaluate the accuracy of students' knowledge. First, responses were unitized by one of the authors. This process separated each response into individual statements (i.e., units), so that the statements could be evaluated separately for accuracy. For example, a response reading “they're black and yellow, they pollinate” would be broken down into two units: (1) “they're black and yellow” and (2) “they pollinate.” Next, a research assistant coded each statement based on its focus (0 = statement about bees in general, 1 = statement about honey bees only) and accuracy (from 1 = completely false to 5 = completely true). It is important to note that this survey was completed prior to the students completing any coursework or attending any lectures, so neither group of students was expected to have a deeply sophisticated or nuanced understanding of bee biology and ecology. Thus, statements were scored on based on the level of understanding appropriate for the general public. For example, though 20% of plant species and 25% of human food crops do not depend on pollinators, the statement “we need them on Earth to survive” was coded as completely true, since it demonstrates the knowledge that, without pollinators, there would be dramatic changes to our natural and agricultural systems. The codes were checked by another author, and any disagreements (n = 10, 0.76% of coded statements) were resolved through discussion. For instance, there was initial disagreement about whether the statement “There are hundreds of thousands of bee species” was close enough to the true estimate (~20,000 species) to be considered at all accurate. Following discussion, however, both agreed that it was far enough off to be considered completely false (1).

After coding, the accuracy of the coded statements (N = 1312) was explored with a 2 (statement type: general vs. honey bee-specific) x 2 (class: low- vs. high-interest) between-subjects ANOVA.


3.1 Preliminary analyses

Participants in the high-interest group made a total of 391 statements (M = 4.20 statements per person, SD = 2.85), with an average of 8.51 words per statement, unedited (SD = 5.91, range = 1–47). Participants in the low-interest group made a total of 1015 statements (M = 4.36 statements per person, SD = 2.55), with an average of 7.28 words per statement, unedited (SD = 5.07, range = 1–33). Independent-samples t-tests revealed that the average number of statements per person did not differ between the two groups, t (324) = 0.47, p = .64, MD = 0.15 statements, 95% CI [−0.48, 0.79], nor did overall response length, t (324) = 1.41, p = .08, MD = 4.78 words, 95% CI [−1.90, 11.47]. Across the two groups, the WORDij analysis revealed that statements included a total of 223 unique words and 1397 unique word pairs. The most common words and word pairs for each group, as well as for the whole sample, are listed in Table 2. A visualization of the network for each group can be found in Figures 1 and 2.

TABLE 2. Most common words and word pairs.
Overall (n = 326) High-interest group (entomology, n = 93) Low-interest group (psychology, n = 233)
Word counts Bees 452 Bees 158 Bees 299
Pollination 221 Pollination 62 Pollination 159
Produce 203 Queen_bee 61 Produce 152
Honey 199 Know 57 Honey 145
Queen_bee 179 Honey 54 Sting 130
People 158 Produce 51 Queen_bee 118
Sting 151 Hive 50 People 114
Know 146 People 44 Flowers 93
Hive 136 Live 36 Know 89
Important 122 Important 35 Important 87
Word pair counts Produce honey 152 Produce honey 34 Produce honey 118
Know bees 73 Know bees 29 Sting people 59
Sting people 70 Bees pollination 19 Pollination flowers 50
Bees pollination 67 Important pollination 15 Bees pollination 49
Pollination flowers 61 Types bees 14 Know bees 44
Important nature 55 Pollination plants 13 Important nature 42
Bees sting 52 Important nature 13 Bees sting 40
Pollination plants 46 Different types 13 Pollination plants 33
Different types 46 Bees people 12 Bees important 33
Bees important 45 Bees sting 12 Different types 33
  • Note: For the high-interest group, there were four other word pairs with a word pair count = 12: bees produce, bees important, different bees, and queen_bee hive. Words and pairs could be used multiple times by each participant; counts reflect how often a word or pair appeared across all participants.
Details are in the caption following the image
Semantic network: high-interest group. Node size is scaled by degree centrality, such that more central nodes are larger. For ease of interpretation, only words with more than 20 connections to other words are labeled.
Details are in the caption following the image
Semantic network: low-interest group. Node size is scaled by degree centrality, such that more central nodes are larger. For ease of interpretation, only words with more than 20 connections to other words are labeled.

3.2 Network results

The first research question focused on identifying (a) the most central concepts in the semantic networks of student knowledge about bees and (b) any differences between students with a high versus low level of interest in the topic (based on their enrollment in an entomology vs. psychology class). As displayed in Table 3, the concepts most central to students' understanding of bees tended to be specific to or closely associated with honey bees (produce, honey, hive, queen_bee). The words people and pollination were also central, which suggests that participants understood bees mainly from a functional and utilitarian point of view, focusing mainly on the value they provide to humans through honey production and pollination of crops and flowers.

TABLE 3. Semantic network analysis results.
Overall (n = 326) High-interest group (entomology, n = 93) Low-interest group (psychology, n = 233)
In-degree In-degree In-degree
Bees 0.34 Bees 0.33 Bees 0.35
Pollination 0.24 Honey 0.22 Pollination 0.27
Produce 0.23 Queen_bee 0.19 Produce 0.23
Queen_bee 0.20 Pollination 0.18 People 0.21
Hive 0.20 People 0.18 Sting 0.20
Honey 0.20 Produce 0.18 Queen_bee 0.19
Out-degree Out-degree Out-degree
Bees 0.48 Bees 0.55 Bees 0.46
Queen_bee 0.26 Know 0.26 Pollination 0.29
Pollination 0.24 Queen_bee 0.20 People 0.29
People 0.21 Pollination 0.17 Honey 0.26
Know 0.20 Hive 0.16 Sting 0.23
Honey 0.19 Honey 0.15 Produce 0.21
Betweenness Betweenness Betweenness
Bees 20.98 Bees 33.57 Bees 18.24
Pollination 6.59 Queen_bee 9.85 Pollination 8.11
Queen_bee 6.20 Hive 7.49 People 7.66
Hive 5.85 Honey 6.87 Queen_bee 7.19
People 4.97 Produce 5.58 Produce 5.92
Produce 4.55 People 4.59 Sting 3.50
  • Note: All centrality measures are normalized.

These patterns also held when looking at the low- and high-interest students separately; honey, queen_bee, produce, hive, pollination, and people were all central concepts in both groups. The main difference between the two was that sting was among the most central concepts for the low-interest students (normalized indegree = 0.20, outdegree = 0.23, betweenness = 3.50), but not for the high-interest students (normalized indegree = 0.09, outdegree = 0.08, betweenness = 1.54). These findings suggest that students in general have a honey bee-centered, utilitarian understanding of bees, and that the threat or fear of being stung is more salient among those with low interest in bees than among those with high interest.

3.3 Knowledge coding results

The second research question focused on evaluating (a) the accuracy of student knowledge about bees and (b) any differences in accuracy between students with a high versus low level of interest in the topic. Descriptive results revealed that the statements made about bees were fairly accurate, with relatively few statements that were mostly (n = 86, 6.6%) or completely false (n = 27, 2.1%) (see Table 4). However, this may have been partly because most statements were very broad. For example, statements like “I know they are essential” and “We need them on Earth to survive” were coded as completely true (5), but they clearly do not indicate a detailed or sophisticated knowledge of bees: while bees are critical pollinators in both natural and agricultural areas, many plants do not require pollination services of bees to reproduce, and rather rely on other pollinators (such as flies) or are wind pollinated.

TABLE 4. Statement accuracy by topic.
Honey bees only All bees
Completely false 0 (0.00%) 27 (2.10%)
Mostly false 2 (5.60%) 84 (6.60%)
Partly true/partly false 1 (2.80%) 499 (39.10%)
Mostly true 3 (8.30%) 138 (10.80%)
Completely true 30 (83.3%) 528 (41.40%)
M (SD) 4.69 (0.79) 3.83 (1.11)
  • Note: The average accuracy for the statements about honey bees was significantly higher than for the statements made about all bees, F (1, 1308) = 23.45, p < .001.

As noted above, accuracy was explored in more detail by conducting a 2 (statement type: general vs. honey bee-specific) x 2 (class: low- vs. high-interest) between-subjects ANOVA on the coded statements (N = 1312).1 The results revealed that statements about honey bees (M = 4.77, 95% CI [4.39, 5.00]) tended to be more accurate than statements about bees in general (M = 3.81, 95% CI [3.75, 3.88]), F (1, 1308) = 23.45, p < .001. On the other hand, there was no significant effect of class (low-interest M = 4.19, 95% CI [3.97, 4.42], vs. high-interest M = 4.39, 95% CI [4.07, 4.71]), F (1, 1308) = 1.00, p = .32. Furthermore, there was no significant interaction effect between class and statement type, F (1, 1308) = 1.74, p = .19. Overall, these results indicate that students in both groups showed the same tendency to have fairly accurate, albeit shallow, knowledge of bees, and more accurate knowledge about honey bees than about bees in general. Examining the effect of statement type also revealed that the difference was likely due to the fact that many students in both groups thought that facts about honey bees were true of all bees. For example, it was common for students to make statements like “They make honey” and “They live in hives,” even though the question prompt referred to bees in general, not to honey bees.


The purpose of this study was to investigate student knowledge of bees, focusing on the concepts that were most central to their understanding of bees and the overall accuracy of their knowledge. We also explored whether there were differences in knowledge between students with a high or low level of interest in the topic, as suggested by their enrollment in either a course focused on honey bees or courses on psychology, respectively.

Semantic network analyses revealed that student understanding of bees is dominated by concepts that are either specific to or closely associated with honey bees and that connect mainly to bees' basic functions and utilitarian value for humans. This is consistent with results from van Vierssen Trip et al. (2020), who found in a recent survey of the Canadian public that half of participants listed honey bees (Apis mellifera) as a native bee of Canada—though this bee was imported to North America from Europe in the 1600s (Carpenter & Harpur, 2021)—and a majority of respondents listed honey and pollination as the most important reason to conserve bees. Moreover, coding of student statements about bees revealed that their understanding tended to be fairly accurate, but not very sophisticated. Knowledge about honey bees was also much more accurate than knowledge about bees in general, mainly because students frequently assumed that facts about honey bees were true of all bees. These findings echo results reported by Vlasák-Drücker et al. (2022), who found that honey, pollination, and stings often come to mind even when people are prompted to think about insects more broadly. Notably, these results were largely the same across the low- and high-interest groups of students, with the only difference being that bee stings were much more salient for the low-interest participants than for the high-interest participants.

4.1 Practical implications

There are several important implications of these findings for bee conservation. First, although some authors have suggested that focusing conservation efforts on honey bees can be counterproductive (Colla & MacIvor, 2017; Ford et al., 2021; Geldmann & González-Varo, 2018; Iwasaki & Hogendoorn, 2021; Wood et al., 2020), the findings here suggest that they may still be among the best ambassadors for bee and pollinator conservation. Students, regardless of their interest level, knew much more about honey bees than about other species. Furthermore, honey was one of the most central concepts in students' knowledge networks. It followed from, led to, and connected many other concepts, giving it an extremely important role in knowledge organization. According to PPT (Reimer et al., 2012; Russell & Reimer, 2018, 2019, 2020), focusing pro-conservation persuasive messages on honey and honey bees may thus be more effective in producing attitude and behavior change than focusing on other species. For similar reasons, the findings suggest that focusing on bees' utilitarian value to humans, such as for honey production and crop pollination, may be more effective than focusing on their environmental or intrinsic value. Pollination, honey, and people were all highly central to student's knowledge networks, which suggests that messages focusing on these features are also likely to be persuasive (Russell & Reimer, 2018, 2019, 2020). In other words, messages that help frame or define bee conservation by highlighting connections to agriculture may resonate more than messages with more environment-oriented frames (e.g., see Kusmanoff et al., 2020).

To be clear, this is not to suggest that messages focusing only on honey bees will be effective in promoting the conservation of other species. Instead, honey bees can serve as a familiar place to start, before transitioning into a discussion of other bees and pollinators. By drawing connections between other bee species and honey bees, practitioners can help tap into existing knowledge structures, making behaviors like native bee conservation more familiar and appealing than they might be otherwise. For instance, a message might say something like:

There are many reasons to love honey bees and take steps to protect them. They make delicious honey and pollinate important crops like almonds, among others. But did you know that Pennsylvania's native bee species need our help too? Native bees like bumble bees also help pollinate many of Pennsylvania's crops. Some bumble bee species that used to be common in Pennsylvania have now all but disappeared from our landscapes. It is more important than ever to take action to protect these vital species. Here's what you can do…

This message targets native bee conservation, but in a way that takes advantage of highly central, honey bee-oriented concepts. In this way, it may be more successful that a message that does not connect to such topics.

Second, and similar to previous findings (e.g., Hall & Martins, 2020; Turo & Gardiner, 2019), the results suggest that the main barrier to bee conservation is likely to be fear of bee stings. Stings were a highly central concept, especially among students who had less interest in bees to begin with. This fear was also common even among students who knew bees were important, suggesting that many of them may be ambivalent about the idea of increasing bee populations. Persuasive messages that work to reassure the public about this highly central fear, or programs that work to decrease this fear (e.g., Cho & Lee, 2018), may thus be another effective option for promoting attitude and behavior change. For example, messages might focus on cultivating positive emotions instead, by portraying bees as cute, highlighting their complex behaviors like maternal care, or emphasizing the many species that do not sting. Interestingly, a survey of people's knowledge and feelings around bees and wasps demonstrated that individuals generally associated bees with their products (honey), pollination services, and stinging, while wasps were associated primarily with stinging, pain, and creating feelings of annoyance (Sumner et al., 2018). Bees also evoked much more positive emotions than wasps. Thus, emphasizing the value to ecosystems and humans may help counteract the fear of stings.

Notably, several authors have suggested that the media focus on honey bees is a primary driver for the lack of public knowledge of other bee or insect species (Smith & Saunders, 2016; Sumner et al., 2018). While increasing stories about other bees available to the public may be another way to increase awareness and interest in other species, it would likely be less efficient than using direct messaging (see Petty et al., 2009). Moreover, stories in the media that seek to dissuade members of the public from being concerned about honey bee declines—though well-intentioned—may be undermining interest in conservation. Messages that target one attitude or behavior can have lateral or spillover effects on other behaviors (Cruz, 2019; Glaser et al., 2015; Smith et al., 2022). Thus, a message persuading people not to worry about honey bee conservation may have the effect of persuading them not to worry about any bee or pollinator conservation. Or, it may lead them to the incorrect impression that “solving” the problem of honey bee conservation has solved problems with bee conservation more broadly—similar to how people often assume that antibiotics may treat many more conditions (e.g., viral infections) than they actually do (Smith et al., 2018). Indeed, the publication of such stories often results in one of the authors being contacted by members of the public congratulating us on the success of our efforts to halt or reverse bee declines.

4.2 Limitations and future directions

This study provides baseline information, which is limited but provides a foundation for future studies. One of the limitations is that we focused on a college student sample, which is not representative of the general population. College students are an important demographic to focus on, in that the conservation decisions they make while in school may have lifelong impacts on behavior, but they also differ from the general public in key ways. In particular, they have higher levels of formal education than average, which may impact their views of bee conservation. For instance, Sumner et al. (2018) found that higher levels of education were associated with higher self-reported interest in nature, which was in turn associated with more positive assessments of the ecosystem services and value of both wasps and bees. This effect was much stronger for wasps, as even people with low interest in nature reported high ecosystem value for bees, but these effects of education may still be important to keep in mind. Future studies on samples of the general public would reveal whether the same concepts are central—and thus likely to be persuasive—with a more diverse audience. Follow-up studies with other specific populations might also yield important insights. For example, it may be useful to understand the knowledge structure among beekeepers or among farmers and growers. Management decisions among these groups can have a meaningful impact on pollinator abundance and pollination services (Breeze et al., 2019; Elisante et al., 2019), so laying the foundation for effective communication campaigns with these groups may be especially useful for conservation efforts.

Another limitation of this study is that we focused on describing students' knowledge structures rather than using that information to test predictions about persuasion. Although there is good reason to expect that the central concepts identified here could be used to develop strong persuasive messages (Reimer et al., 2012; Russell & Reimer, 2018, 2019, 2020), we cannot say for certain that is the case. Future research that explicitly tests the effects of messages based on the predictions of PPT would be an important next step in promoting conservation behavior. Experimental designs, in particular, would help provide evidence for the effectiveness of such messages in producing pro-conservation attitudes and behavior change.

4.3 Conclusion

The goal of this study was to build on previous studies of public understanding of bees by examining the structure of students' self-reported knowledge using semantic network analysis. The results revealed that the concepts most central to student understanding of bees included honey, bees' utilitarian value for humans, and the threat of bee stings. Work in communication science (Russell & Reimer, 2018, 2019, 2020) suggests that focusing on these concepts may be most effective when designing persuasive messages to promote bee conservation. Moreover, given that a recent study (Vlasák-Drücker et al., 2022) demonstrated that members of the public most associated insects with the terms bees, useful, nature, and pollination, leveraging the public's knowledge and interest in bees may be an effective strategy to address factors leading to insect biodiversity declines (Wagner et al., 2021) more broadly.


Shannon Cruz: Conceptualization; methodology; investigation; formal analysis; writing – original draft preparation. Christina Grozinger: Conceptualization; writing – original draft preparation.


We would like to thank Allyson Ray for her assistance in coding the accuracy of the student statements, and Harland Patch and Cathy Hunt for allowing us to survey the students in their courses. Funding for this study was provided by the Communication, Science, and Society Initiative (CSSI), a partnership between Penn State's Department of Communication Arts and Sciences and the Huck Institutes for the Life Sciences.


    The authors have no conflict of interest to declare.


  1. 1 Beforehand, we also checked for possible dependence in the data due to the fact that statements were nested within subjects. The intraclass correlation coefficient for subject was very small (ICC(1) = .04), suggesting dependence was not a concern for this analysis.

    Data files for this project are accessible at: https://osf.io/5927r/?view_only=6ffc58109e204a86b35fd9e205246ce5.