Can Nature Exposure Impact Strategic Decision-Making?

The study uses two interventions, combined to form four room conditions. The first is the presence of live indoor plants: in sessions with this intervention, a potted plant sits on each participant's desk next to the computer for the whole session. The second is a window view of a natural landscape: in sessions with this intervention, the laboratory's windows are left uncovered, so participants can see an outdoor natural landscape. In sessions without it, the windows are covered.

Crossing the two interventions gives four conditions:

Control: no plants and the window view covered.
Plants only: plants on the desk and the window view covered.
View only: the window view open and no plants.
Plants and View: plants on the desk and the window view open.

Conditions are assigned by session. Everyone in the same session experiences the same room setting from arrival to the end of the session. Apart from the room setting, all participants follow the same procedures: a short timed task, a set of computerized decision-making games with other participants, and brief questionnaires.

2026-09-22

2026-10-16

Decisions in five one-shot economic games:
1. Cooperation in the Stag Hunt (choosing Stag, 0/1).
2. Cooperation in the Prisoner's Dilemma (choosing Cooperate, 0/1).
3. Trust: amount sent by the Sender in the Trust Game ($0–10).
4. Trustworthiness: share of the received amount returned by the Receiver in the Trust Game (0–1).
5. Wage offered by the Employer in the Gift Exchange game ($0–20).
6. Reciprocity: Employee choosing high effort in the Gift Exchange game (0/1).
7. Offer made by the Proposer in the Ultimatum Game ($0–20).
8. Acceptance by the Responder in the Ultimatum Game (0/1).
9. A summary index of prosocial and cooperative behavior built from outcomes 1–8.

Outcomes 1–8 are recorded directly in oTree.
* Trustworthiness (4) is the amount returned divided by the doubled amount received. It is undefined when the Sender sends $0, so those observations are excluded from that outcome.
* Reciprocity (6) and Ultimatum acceptance (8) are analyzed conditional on the first mover's decision (the wage or the offer), entered as a control.
The summary index (9) follows Anderson (2008):
1. Each outcome is standardized using the mean and standard deviation of the Control condition (no plants, view covered).
2. Outcomes are signed so that higher values mean more cooperative or prosocial behavior.
3. The standardized outcomes are averaged across the games each participant played in each role. Each participant plays one role per game, so each person's index averages the outcomes available for that person.

1. hysiological arousal: galvanic skin response (skin conductance level and the frequency of skin conductance responses), measured during the stress task and during the games.
2. Blink rate.
3. Facial expression indicators of emotion.
4. Performance on the mental arithmetic task: number of correct answers and response time.
5. Willingness to accept (WTA) for the plant.
6. Decision response times in the games.

* GSR, blink rate and facial-expression measures are processed in iMotions. They are summarized for each participant over predefined task windows: baseline/acclimation, stress task, and games. They are expressed as changes relative to that participant's baseline window.
* These measures are also used to test whether stress mediates the effect of nature exposure on game decisions.
* WTA is the stated minimum dollar amount accepted instead of taking the plant home. All participants complete this task.
* Participants who fail eye-tracking calibration are excluded only from the eye-tracking outcomes.

This is a laboratory experiment with a 2×2 between-subjects design. It crosses the presence of live indoor plants with a window view of a natural outdoor landscape, giving four room conditions: neither, plants only, view only, or both. Conditions are assigned by experimental session, so all participants in a session share the same room setting. Adult participants complete a short timed task, a series of incentivized computerized decision-making games with other anonymous participants, and short questionnaires. Physiological and attention measures are recorded throughout the session.

Not available

Experimental sessions were assigned to conditions by a computer-generated random schedule prepared before data collection.

Experimental session. All participants in a session receive the same condition, and there is no individual-level randomization of treatment. Roles within the games and the order of the games are randomized at the individual level by oTree, but neither is a treatment.

Yes

60 Sessions. Each session is a 45 minutes time slot the participants sign up.

300 students.

around 75 participants per treatment arm.

Model. All tests are based on the regression Y_is = α + β_P·P_s + β_V·V_s + β_PV·(P_s × V_s) + X_is'γ + ε_is, where Y_is is the outcome of participant i in session s. P_s indicates live plants on the desks and V_s an open window view of a natural landscape. Both are assigned at the session level and coded −1/2 (absent) and +1/2 (present). X_is is a vector of pre-specified controls. Standard errors are clustered by session, and p-values are computed with the wild cluster bootstrap. With this coding, the coefficients have a direct meaning in the units of the outcome: α is the average of the four condition means. β_P is the average difference in the outcome between sessions with and without plants, averaged over the two view conditions. β_V is the average difference between sessions with the view open and covered, averaged over the two plant conditions. β_PV is the difference-in-differences: how much the effect of plants changes when the view is open. Hypotheses tested. H1 (Plants and game behavior, two-tailed). Plants change cooperative and prosocial behavior in five economic games (Stag Hunt, Prisoner's Dilemma, Trust, Gift Exchange, Ultimatum) and in a summary index. Theory supports both a positive effect through stress reduction and a negative effect through more deliberative decision-making, so the test is two-tailed. H0: β_P = 0 versus H1: β_P ≠ 0. H2 (View and game behavior, two-tailed). Same outcomes and reasoning as H1. H0: β_V = 0 versus H1: β_V ≠ 0. H3 (Physiological stress, one-tailed). Plants and the view each lower physiological arousal, measured by skin conductance level and the frequency of skin conductance responses during the stress task and the games. H0: β_P ≥ 0 versus H1: β_P < 0; H0: β_V ≥ 0 versus H1: β_V < 0. H5 (Plant exposure and valuation, one-tailed). Participants who had a plant on their desk state a higher willingness to accept (WTA) for the plant. H0: β_P ≤ 0 versus H1: β_P > 0. H6 (View as a substitute, one-tailed). Participants who had the open view state a lower WTA for the plant. H0: β_V ≥ 0 versus H1: β_V < 0. Interaction (exploratory, two-tailed). H0: β_PV = 0 versus H1: β_PV ≠ 0. The mediation of the H1 and H2 effects by stress (H4) is estimated as supporting evidence and is not part of this power analysis. Minimum detectable effect. For each test, the MDE is the smallest true |β| that rejects H0 with 80% power at α = 0.05, given 300 participants and the clustering assumptions below. MDEs are reported in standard deviations of the outcome (the SD in the Control condition) and converted to natural units where a reference SD is available. For one-tailed tests, the MDE applies only in the stated direction. The calculation assumes α = 0.05 and 80% power, an intraclass correlation of 0.05 within sessions, and about [m] participants per session (design effect ≈ 1.2 for m ≈ 5). Effects of plants and view on game behavior (H1, H2) are tested two-tailed. Effects on physiological stress (H3) and plant valuation (H5, H6) are tested one-tailed, since the literature supports a direction for these. MDEs are in standard deviation units of each outcome. Game behavior (H1, H2; two-tailed): Outcomes every participant contributes (Stag Hunt, Prisoner's Dilemma, summary index; 150 vs 150): MDE = 0.32 SD without clustering and 0.35 SD with clustering. For a binary cooperation rate near 50%, this is about 18 percentage points. Role-specific outcomes (Trust sent and returned, wage, effort, Ultimatum offer and acceptance; 75 vs 75): MDE = 0.46 SD without clustering and 0.50 SD with clustering. For example, with an SD of about $3 for amounts sent in a $10 Trust Game, this is about $1.50. Physiological stress (H3; one-tailed): skin conductance level and skin conductance response frequency, relative to baseline; 150 vs 150. MDE = 0.29 SD without clustering and 0.31 SD with clustering. This assumes complete GSR data. Lost recordings will raise the MDE in proportion. Plant valuation (H5, H6; one-tailed): WTA is elicited from all 300 participants, 150 vs 150 for each main effect. MDE = 0.29 SD without clustering and 0.31 SD with clustering. Plants × View interaction (two-tailed): MDE ≈ 0.71 SD. The study is not powered to detect interactions of plausible size, so interaction estimates are exploratory. The mediation analysis (H4) is not separately powered and is reported as supporting evidence on mechanism.