The United Nations’ expert group on biodiversity and ecosystems recently released a bleak forecast of the state of the world’s flora and fauna (Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services 2019). Declining biodiversity exacts high costs on all of humanity, including impacts on energy (e.g., wood fuel), medicines, food and even the air people breathe. As the cause of much of the loss of biodiversity, humans can still mitigate some of this decline and many of its repercussions. Thus, increasing people’s valuing of biodiversity and understanding of the costs of biodiversity loss, as well as their desire to take action represent particularly urgent problems.

However, getting people to notice, care about and ultimately act to prevent declining biodiversity faces an unusual paradox. On the one hand, species of all sorts are highly visible to almost everyone across the entire planet. On the other hand, people struggle to observe species’ rapid disappearance (Pett et al. Reference Pett, Shwartz, Irvine, Dallimer and Davies2016). Similar to people’s perceptions of climate change, the extinction of species represents the kind of distant, non-immediate threat that our brains struggle to perceive and process (Gifford Reference Gifford2014, Clayton et al. Reference Clayton, Devine-Wright, Stern, Whitmarsh, Carrico and Steg2015). Moreover, experts estimate that we have discovered only a small fraction of all the species on the planet (Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services 2019). Consequently, it is even harder to comprehend how valuable yet-to-be-discovered species are and how costly their loss would be.

Expectancy-value theory (Wigfield & Eccles Reference Wigfield and Eccles2000) provides a lens to help understand how people’s values contribute to their motivation. Typically invoked in achievement domains, expectancy-value theory suggests that people will be motivated to pursue certain goals based on the degree to which they might reasonably expect to achieve those goals and how much they value those goals. The third key element of the theory – the costs of pursuing a particular goal – has typically been thought to work in opposition to the valuing of that goal (Flake et al. Reference Flake, Barron, Hulleman, McCoach and Welsh2015). For example, a student who is motivated to get an A on an exam but perceives the costs of missing out on a party as too high might not ultimately be motivated to study. Therefore, this theoretical paradigm views costs as a counterbalance that undermines motivation.

As a theoretical contribution of our study, we explore the possibility that perceived costs could enhance motivation (rather than assuming that values and costs will necessarily function in opposition). Specifically, we test whether making visible the hard-to-see ‘costs’ of biodiversity loss might increase people’s valuing of biodiversity. Scholarship on loss aversion (Kahneman Reference Kahneman2011), which argues that people feel losses more acutely than corresponding gains, also suggests that this approach of using potential costs to increase valuing could be effective.

How might perceived costs lead to greater valuing in the specific context of biodiversity? Expectancy-value theory suggests that seeing photographs and/or captions illustrating the effects of biodiversity loss might help people more clearly focus on an issue that otherwise seems abstract, scientific and unemotional, thereby making it seem more salient, resonant, important and, potentially, more valued. Correspondingly, the strategies to mitigate this loss might seem more useful and necessary.

In addition to photographs being used regularly by conservation groups, scholars have examined the use of images in domains such as animal conservation (e.g., Gunnthorsdottir Reference Gunnthorsdottir2001, Colléony et al. Reference Colléony, Clayton, Couvet, Saint Jalme and Prévot2017). Related research suggests that images rated as being highly relevant to climate change tend to make people feel more emotionally aroused and more negative (Lehman et al. Reference Lehman, Thompson, Davis and Carlson2019). Given the mixed results that informational campaigns have had in persuading people to increase their concerns about climate change (Hart & Nisbet Reference Hart and Nisbet2012, Geiger et al. Reference Geiger, Swim and Fraser2017), photographs or other visual illustrations of what might be lost may better enhance people’s valuing of different species. Presumably, this strategy needs to be enacted with finesse, however; when participants become overwhelmed by dire messages they may feel a sense of helplessness (Salomon et al. Reference Salomon, Preston and Tannenbaum2017) and/or a lack of efficacy to act on the problem (Feinberg & Willer Reference Feinberg and Willer2011, Curnock et al. Reference Curnock, Marshall, Thiault, Heron, Hoey and Williams2019). In other words, photographs that tell a story that is too intense or depressing might inhibit viewers from taking action by making the emotional costs to the self too great. On the other hand, and perhaps most relevant to the present study, Swim and Bloodhart (Reference Swim and Bloodhart2015) use photographs and an emotion-related induction and find that emotions do facilitate pro-environmental behaviour, thus suggesting that finding the right level of emotional arousal to produce behavioural changes is a tractable challenge.

Alternatively, perhaps photographs’ potential effectiveness for increasing people’s valuing of biodiversity stems not from their capacities to elicit negative emotions but instead from their concrete, tangible nature. Seeing that which is hard to otherwise perceive may reduce the psychological distance of otherwise abstract issues such as biodiversity loss (Schuldt et al. Reference Schuldt, Rickard and Yang2018). Furthermore, even if pictures are worth 1000 words, captions offering brief explanations and interpretations might offer an even better way to frame viewers’ understandings of what they are seeing. The multimedia principle establishes that people learn more from visual images and text than from either alone (Mayer Reference Mayer2009). Moreover, strategic framing can affect people’s attitudes about climate issues (Hurlstone et al. Reference Hurlstone, Lewandowsky, Newell and Sewell2014). On the other hand, framing issues of biodiversity is complicated (Elliott Reference Elliott2020); the meaning that is intended by a caption author might not be the meaning that is received by a caption reader. So knowing whether and the extent to which captions enhance the photographic approach will be useful for those working to change people’s attitudes and behaviours regarding biodiversity.

In the current study, we tested whether photographs could bolster participants’ valuing of biodiversity (as assessed by a survey scale) and subsequently impact their behaviours (specifically their willingness to donate to an environmental charity) as compared to a control group that saw no photographs. Because the framing of environmental issues is particularly important (Clayton et al. Reference Clayton, Devine-Wright, Stern, Whitmarsh, Carrico and Steg2015), we included two treatment groups: one that saw photographs only and another that saw the same photograph and a caption that explained the content of the photograph. Because there is no real-world analogue to a caption without a photograph and because our captions explicitly reference the content of our photographs, we did not test a ‘captions-only’ treatment condition. Relatedly, the conflicts in the aforementioned literature suggest that predicting which treatment approach is more effective would be premature; thus, we did not preregister a hypothesis comparing the two treatment groups. We selected the photographs deliberately to signal the value of biodiversity specifically by highlighting the costs of what would be lost if the declines in biodiversity continue (Curnock et al. Reference Curnock, Marshall, Thiault, Heron, Hoey and Williams2019). We tested two preregistered hypotheses (Gehlbach & Robinson Reference Gehlbach and Robinson2018), specifically that (1) participants who view photographs and (2) those who view photographs with captions will report higher valuing of biodiversity than participants who do not see photographs (see https://osf.io/632dk/ for the full preregistration).

Methods

Participants

We first conducted four pilot tests to help calibrate the efficacy of our intervention using different stimuli, the sensitivity of our measures, the magnitude of our effects and an appropriate sample size for a preregistered experiment with three conditions given those effects. From these pilots, we knew that we wanted a final sample size of 250 participants per condition – which would give us 80% power to detect an effect size of 0.19 or greater – and thus we recruited a greater number of participants to allow for attrition (see our preregistration at https://osf.io/632dk/ for more details). After preregistering the study, we used Amazon’s Mechanical Turk to obtain a final sample of 843 participants (47% female, 48% male, with 5% choosing not to answer). Our participants had a mean age of 38.4 years; their median education level was a bachelor’s degree. Politically, 410 participants identified as liberals, 162 as moderates and 262 as conservatives (with 9 participants not reporting). To help compare our sample to a US-based, nationally representative sample on a small but important aspect, we asked our participants the most relevant question of those posed in Yale’s Climate Survey (Leiserowitz et al. Reference Leiserowitz, Maibach, Rosenthal, Kotcher, Bergquist and Ballew2019): ‘How much do you think global warming will harm plant/animal species?’. Our respondents were slightly more concerned (82% reporting ‘a moderate amount’ or ‘a great deal’) than respondents of the most recent Yale Climate Survey (71%) at the time of data collection.

We arrived at our final sample from an original group of 1007 after having excluded participants who failed a screener question (n = 164) that tested whether they were reading the initial instructions and were never assigned to a condition, as well as one participant who failed the screener question but advanced in the survey due to a technical glitch. In line with our preregistration, we removed participants (n = 7) who produced 10 or more identical, sequential responses on the emotion items after our valuing of biodiversity scale. There were also three participants who did not complete the biodiversity scale. Finally, two participants took the survey twice, and we removed the data from the second time they took it.

Measures

Our preregistration and Table 1 list all composite and outcome measures collected in the study; here, we report on the focal measures for the analyses of interest. Four self-report measures and one behavioural item comprised our main measures. First, participants completed a five-item measure assessing how much they value biodiversity (α = 0.82). In the treatment conditions, a relevant image selected by the research team to convey the cost of losing different elements of biodiversity was displayed alongside each item. For example, the item ‘How important do you think it is for humans to take action to make sure that endangered species survive?’ was paired with the photograph of a man feeding a baby elephant that appears in ill health. In the caption condition, participants read how volunteers in Africa were helping to protect elephants from ivory poachers but that the loss of elephants has nearly doubled in the past 20 years. The intervention photographs, including the full set of pairings between each survey item, photograph and caption, are available from the first author upon request.

Table 1. Means, standard deviations and correlations of study variables.

Note: n-values ranged from 839 to 843.

In addition, participants completed two five-item measures of psychological distance. The first focused on psychological distance from climate change in general (α = 0.87), while the second focused on distance from the effects of declining biodiversity in particular (α = 0.74). See Supplementary Table S1 (available online) for the full measures. Finally, we created a composite of five negative emotions (α = 0.93) – sad, guilty, worried, frustrated and scared – by instructing participants to use a slider bar to ‘please indicate how much you’re feeling each different emotion right now’.

In addition to the self-report measures, participants gave a small donation to a non-profit organization of their choice in response to the following invitation: ‘We would like to offer you US$0.50 to donate to charities of your choosing. Donating to a charity will not affect your compensation.’ Participants then used slider bars on the web interface to allocate how much money went to The Nature Conservancy, Save the Children and/or the American Cancer Society and were provided with a brief description of each. The outcome of interest was the amount donated to The Nature Conservancy.

Procedure

After being randomly assigned to a control group (n = 281), a treatment group that saw only photographs (n = 280) or a treatment group that saw photographs and captions (n = 282), all participants responded to the valuing of biodiversity scale. Next, participants recorded their current emotional state (including the five negative emotions of interest), the two scales on psychological distance, the donation item, several additional measures and demographic items. We paid participants US$2.00 upon completion of the survey.

All materials, data and statistical code are available from the lead author upon request.

Results

To guide the testing of our preregistered hypotheses, we first evaluated whether our randomization produced equivalent groups in comparing each treatment group against the control condition. A multinomial logistic regression identified no significant discrepancies between groups based on gender, age, education or political orientation (χ² ₍₈₎ = 14.01, p = 0.082). The remainder of our analyses follow Cumming’s (Reference Cumming2014) recommendations by focusing on confidence intervals (CIs) – 97.5% CIs for our two pre-specified hypotheses and 95% CIs for our exploratory analyses – and effect sizes (rather than p-values). See Table 1 for the descriptive statistics of the variables collected.

Pre-specified hypotheses

We first expected to see significantly greater valuing of biodiversity from the participants who saw (only) photographs relative to our control group. An ordinary least squares regression supported this hypothesis (B = 0.73, SE = 0.066, 97.5% CI = 0.025, 0.320, Cohen’s d = 0.22). Next, we predicted that the photographs-plus-captions treatment group would also value biodiversity to a greater extent than the control group. The support for this hypothesis was weaker and the CI for the difference between these two groups does include 0 (B = 0.123, SE = 0.066, 97.5% CI = –0.024, 0.270, Cohen’s d = 0.16).

Exploratory analyses

Based on a combination of our pilot data and literature suggesting that changes in valuing could result in changes in behaviours (Eccles Reference Eccles and Sonderegger1984, Hulleman et al. Reference Hulleman, Godes, Hendricks and Harackiewicz2010), we thought that increased valuing of biodiversity might affect participants’ behaviour – specifically a greater willingness to donate to a cause that supports biodiversity. We found that participants in the photographs-only condition donated more of their allotted US$0.50 to the environmental charity than their control counterparts (B = US$0.036, SE = 0.015, 95% CI = 0.006, 0.065, Cohen’s d = 0.20; see Fig. 1).

Fig. 1. Donation to environmental charity by condition. Error bars represent 95% confidence intervals. A statistically reliable difference emerged between the control and photographs-only groups but not between the control and photographs-plus-captions groups.

With this evidence that participants in the photographs-only condition came to value biodiversity more and donated more than their control counterparts, we sought to understand why this might have been the case by comparing these two groups further. Based on the prior literature suggesting that psychological distance may play a substantial role in people’s failure to value and act upon the growing evidence of climate change (e.g., Van Lange & Bastian Reference Van Lange and Bastian2019), we examined two measures of psychological distance. One measure assessed perceptions that climate change, in general, remained a distant threat; the other assessed how distant of a threat biodiversity loss seemed to be. Neither measure of psychological distance appeared plausible as a mediator that could explain the effectiveness of the treatment on participants’ valuing of biodiversity. As compared to the control group, the photographs-only treatment did not reduce general psychological distance (B = 0.005, SE = 0.020, 95% CI = –0.036, 0.045) or distance from the effects of declining biodiversity (B = 0.023, SE = 0.017, 95% CI = –0.010, 0.056).

On the other hand, photographs induced significantly more negative emotions in the photographs-only treatment group. Specifically, as compared to control participants, these treatment participants felt stronger negative emotions (B = 6.874, SE = 1.855, 95% CI = 3.232, 10.515, Cohen’s d = 0.35). We also observed that the photographs-plus-captions participants reported even stronger negative emotions compared to the control group (B = 14.657, SE = 1.852, 95% CI = 11.021, 18.292, Cohen’s d = 0.67).

For the photographs-only participants, we then wondered whether these negative emotions could plausibly explain why they valued biodiversity more than the control group. In turn, could participants’ valuing of biodiversity plausibly explain why the photographs-only treatment participants donated more to the environmental charity? We tested this conjecture through a path model (see Fig. 2). This model provided a good fit for our data (χ² _{(2, n = 556)} = 2.84, p = 0.24, comparative fit index = 0.994, root mean square error of approximation = 0.027). The path model suggests that this interpretation – that seeing our photographs induces negative emotions that spark a greater valuing of biodiversity, which, in turn, leads to greater environmental donations – was plausible given our data. Path models of (1) treatment → valuing of biodiversity → negative emotions → donation and (2) that allowed for separate, parallel paths from treatment to donation – one path through negative emotions and one path through valuing of biodiversity – fit the data poorly and thus were not considered plausible. Because we did not find a statistically reliable difference between the photographs-plus-captions group and the control group, our path model did not include the photographs-plus-captions participants.

Fig. 2. Mediation model of the effect of the treatment on donations through negative emotions and valuing of biodiversity. Standardized regression coefficients are given for the associations between photographs-only treatment (as compared to the control group), negative emotions, valuing of biodiversity and donations to The Nature Conservancy. The effect of the treatment on the outcome after accounting for the mediators (in parentheses) is not statistically distinguishable from 0.