Measures

The next subsections present the main variables of interest in all studies.

Additionally, the selection of covariates/confounders varied between the studies.

Covariates are listed in Table 3 and the general rationale for their selection is provided in Section 1.3.5. More details on their exact wordings and operationalisations are provided in the original articles.

2.2.1 Well-being, mood and attention – outcomes

Emotional well-being (Studies 1 and 2) in the past four weeks was measured on the Emotional Well-Being subscale in the RAND 36-item health survey 1.0 (Hays, Sherbourne, & Mazel, 1993; Finnish validation by Aalto, Aro, & Teperi, 1999),

44

comprising five questions (for example, ‘Have you been a happy person?’), measured on a 6-point ordinal scale ranging from ‘All of the time’ (1) to ‘Not at all’ (6). Three items were reverse-coded so that greater values indicated better well-being. This scale was operationalised as a confirmatory latent factor in the analyses.

In Study 3, one of the outcomes was also labelled as emotional well-being, but instead of an existing, validated scale, it was an exploratory latent factor comprising three items on positive emotions frequently experienced in everyday life (happiness, calmness, and joy; Zelenski & Larsen, 2000), one item on vitality (‘I felt alive and vital’; Ryan & Frederick, 1997) and one item on life satisfaction (‘I was fairly satisfied with my life’).

General health (Study 1) was elicited by a single item (widely used practice, for instance Idler & Benyamini, 1997; Miilunpalo, Vuori, Oja, Pasanen, & Urponen, 1997), ‘In general, would you say your health is…’, with the options ‘good’ (1), ‘fairly good’ (2), ‘average’ (3), ‘fairly poor’ (4), and ‘poor’ (5). This scale was reverse-coded so that greater values indicated better health.

Sleep quality (Study 1) was elicited by a single item ‘How often in the past four weeks have you experienced sleep problems or poor sleep quality’, options ranging from ‘All of the time’ (1) to ‘Not at all’ (6). Asking about sleep problems by a single item has been found to correlate strongly with more comprehensive measures of sleep quality (Carpenter & Andrykowski, 1998). For the analyses, this scale was reverse-coded so that greater values indicated better sleep quality.

Creativity (Study 3) after the most recent visit to a natural setting consisted of four items (such as ‘I came up with many new ideas’), evaluated on an ordinal scale ranging from ‘Describes my experience completely’ (1) to ‘Not at all’ (7). This scale was developed on the research project (Tyrväinen et al., 2014). These items formed an exploratory latent factor.

Valence and activation (Study 4) were measured with the two-dimensional affect grid in which the participants are asked to evaluate their mood by marking a single cross on a 9 × 9 grid, with the axes representing these two core affects (Russell, Weiss, & Mendelsohn et al., 1989; Västfjäll & Gärling, 2007).

Sustained attention (Study 4) was measured with the Sustained Attention to Respond Task (SART; Robertson, Manly, Andrade, Baddeley, & Yiend, 1997). In SART participants are required to respond to a rapidly-paced stimulus, digits 1-9 on screen, but to withhold their response if the digit is 3. They are asked to pay equal attention to speed and response accuracy. To create the task, we used Stothart’s (2015) source code in the open source software PsychoPy (Peirce, 2009). Several measures from this task can be calculated: commission errors that measure response

accuracy, controlled attention, and response inhibition (Johnson et al., 2007; Manly et al., 2003); mean response time (RT); and standard deviation of response time (SDRT) reflecting the stability of responses (Robertson et al., 1997; Manly et al., 2003; Smilek, Carriere & Cheyne, 2010). SDRT was further analysed by differentiating fast and slow frequency attentional lapses (for the methodology, see Johnson et al., 2007), indicating moment-to-moment variability (FFAUS) and slower, gradual changes (SFAUS) in response times. The results for FFAUS and SFAUS were similar to those for SDRT and thus, for the sake of simplicity, they are not discussed in detail in this summary but further information regarding them appears in the original research article.

2.2.2 Self-reported restoration – outcome and mediator

Restoration was measured on the Restoration Outcome Scale (ROS; Korpela et al., 2008). The scale has two versions, a shorter 6-item (ROS6), and a longer 9-item version (ROS9, in which the respondents are asked to what extent different statements describe their experiences. The items may be phrased in the past or the present tense, depending on the design of the study. Drawing from restoration literature (Hartig, Lindblom, & Ovefelt, 1998; Staats, Kieviet, & Hartig, 2003), in the original 6-item version three items measure relaxation and calmness (for example, ‘I calmed down’), one attention restoration (‘My concentration and alertness increased’), and two clearing one's thoughts (for example, ‘I forgot everyday worries’). In the longer version, two of the additional items measure vitality (for example, ‘My vitality and energy increased’) and one self-confidence (‘I became more self-confident’; Korpela & Ylén, 2009; Korpela, 1992; Ryan et al., 2010). All items are evaluated on a 7-point ordinal scale, with greater values indicating greater restoration.

ROS is one of the few measures on self-reported restoration outcomes to include aspects of both ART and SRT (Han, 2018). Like most other restoration measures, its psychometric properties, especially those of the longer version, require more research in different contexts (Han, 2018). In this thesis, ROS is modelled as an exploratory latent factor in Study 2 (mediator, 9-item scale) and Study 3 (outcome, 6-item scale), and as a summary score in Study 4 (outcome, 6-item scale).

46

2.2.3 Attentional focus during the most recent nature visit – mediator To examine attentional foci (Study 3) during the most recent visit to a typical place or route in nature, the respondents were asked to what extent they had focused on

‘the activity, that is, outdoor recreation’ (focus on the activity), ‘your own thoughts and emotions’ (focus on oneself), ‘other people around you’ (focus on others), and

‘the environment, that is, the natural or urban surroundings’ (focus on the environment), on a 7-point ordinal scale ranging from ‘completely’ to ‘not at all’

(McIntyre & Roggenbuck, 1998). In the analysis the items were reverse-coded so that higher values indicate greater focus.

2.2.4 The frequency and settings for PA – independent variables

The respondents were asked about the settings in which they had taken their a) most recent bout of PA (Study 2), and b) PA in general (Studies 1 and 2). For the analysis we grouped these into indoor (home or indoor sports facilities such as gyms), built outdoor (such as streets, cycle lanes, sport fields), and natural outdoor settings (for example, nature close to home or to second home, usually a rural summer cottage).

Outdoors in front or back yard was excluded because we had no information if these were perceived as built or natural.

Weekly frequencies of PA in indoor, built outdoor, and natural outdoor settings were obtained by multiplying the percentage of PA taking place in each of these settings by weekly frequency of PA overall.

2.2.5 Motives for the most recent nature visit – independent variables The respondents were asked how important a variety of factors were in their decisions to go outdoors, on a 4-point scale ranging from (1) very important to (4) not important at all (reverse-coded for the analysis; Study 3). These were grouped into motives to be physically active (“maintaining physical fitness”), to socialise (‘I can be with friends’, ‘I can be with family’), to be alone (‘I get to be alone’), to reduce stress (‘I can relax,’ ‘I can withdraw from daily routines,’ and ‘I can reduce stress’), and to experience nature (‘I can enjoy nature’, ‘I can learn from nature’). This categorisation was based on the mechanisms that link visits to nature and well-being (Section 1.2) and Knopf’s (1987) categorisation of motives for visiting natural settings.

In the analyses the relationship between these motives and attentional foci (Section 2.2.3; mediators) was specified in two ways based on different theoretical reasoning: Model 1 assumed that motives and attentional focus should theoretically match (‘restricted model’), and Model 2 allowed all motives to be connected to all types of attentional foci (‘unrestricted model’).

2.2.6 Covariates

All studies controlled for different sets of covariates, which were selected a priori in Studies 1 and 2, and in Study 4 post-hoc, based on their association with the outcomes due to smaller sample size.

The analyses in Study 1, assessing everyday PA patterns, controlled for gender, age, household income, household size, exceptional situation in life, constraints for outdoor recreation, season, and general activity level. Studies 2 and 3 did not include any covariates but a sensitivity model reported alongside Study 2 controlled for gender, age, and household income. In Study 4, a number of plausible covariates based on the literature were tested, and those that showed a statistically significant association with an outcome in either study were selected for the analyses. These included stress in the past four weeks, age, start time of the experiment, ease of wayfinding, and method of navigation (mobile application / paper map; Study 4B only). Excluded potential outcomes not related to the outcomes included gender, walk duration, temperature, weather, sleep, and unusual events or fear during the walk (Study 4B only).