Environment-Specific vs. General Knowledge and Their Role in Pro-environmental Behavior [original]

fpsyg-10-00718 March 29, 2019 Time: 18:51 # 1

ORIGINAL RESEARCH

published: 02 April 2019

doi: 10.3389/fpsyg.2019.00718

Edited by:

Giuseppe Carrus,

Università degli Studi Roma Tre, Italy

Reviewed by:

Paola Passafaro,

Sapienza University of Rome, Italy

Stefano Mastandrea,

Università degli Studi Roma Tre, Italy

*Correspondence:

Oliver Wilhelm

oliver[email protected]

Specialty section:

This article was submitted to

Environmental Psychology,

a section of the journal

Frontiers in Psychology

Received: 06 June 2018

Accepted: 15 March 2019

Published: 02 April 2019

Citation:

Geiger SM, Geiger M and

Wilhelm O (2019)

Environment-Specific vs. General

Knowledge and Their Role

in Pro-environmental Behavior.

Front. Psychol. 10:718.

doi: 10.3389/fpsyg.2019.00718

Environment-Specific vs. General

Knowledge and Their Role in

Pro-environmental Behavior

Sonja Maria Geiger1, Mattis Geiger2and Oliver Wilhelm2*

1Institute for Vocational Training and Work Studies, Technische Universität Berlin, Berlin, Germany, 2Psychological Institute,

University of Ulm, Ulm, Germany

Environmental knowledge has been established as a behavior-distal, but necessary

antecedent of pro-environmental behavior. The magnitude of its effect is difficult

to estimate due to methodological deficits and variability of measures proposed in

the literature. This paper addresses these methodological issues with an updated,

comprehensive and objective test of environmental knowledge spanning a broad

variety of current environment related topics. In a multivariate study (n= 214), latent

data modeling was employed to explore the internal factor structure of environmental

knowledge, its relationship with general knowledge and explanatory power on pro-

environmental behavior. We tested competing factor models and uncovered a general

factor of environmental knowledge. The main novel finding of the study concerns

its relationship with general knowledge. Employing an established test of general

knowledge to measure crystallized intelligence revealed a near perfect relationship

between environmental and general knowledge. This general knowledge (including the

environmental domain) accounted for 7% of the variance in environmentally significant

behavior. Age, additionally to acquired education, emerged as a common predictor

for both general knowledge and environmentally significant behavior. We discuss the

consequences of the strong relation between general and environmental knowledge

and provide a possible explanation for the positive age-environmental conservation

relationship reported in the literature.

Keywords: environmental knowledge, general knowledge, domain-specificity, environmentally significant

behavior, structural equation modeling

INTRODUCTION: THE ROLE OF KNOWLEDGE IN

PRO-ENVIRONMENTAL BEHAVIOR

Knowledge about environmental issues is thought to be a precondition for meaningful pro-

environmental behavior and its transmission is considered a key component and criterion for

successful implementation of environmental education programs (UNESCO, 2005;Heimlich and

Ardoin, 2008;Kaiser et al., 2008). A moderate, positive influence of domain-specific environmental

knowledge has been reported in early research on pro-environmental behavior (Hines et al., 1987)

and was replicated in various studies since (Kaiser and Frick, 2002;Frick et al., 2004;Meinhold

and Malkus, 2005;Geiger et al., 2014, 2018). Other studies have found very small cross-sectional

relationships (Roczen et al., 2013;Braun and Dierkes, 2017;Otto and Pensini, 2017) and the limited

use of knowledge interventions on behavior change has been discussed since the seminal paper of

Frontiers in Psychology | www.frontiersin.org 1April 2019 | Volume 10 | Article 718

fpsyg-10-00718 March 29, 2019 Time: 18:51 # 2

Geiger et al. General Knowledge and Environmental Behavior

Kollmuss and Agyeman (2002).Kaiser and Fuhrer (2003)

propose that environmental knowledge, as a distal predictor of

environmental behavior, is systematically underestimated and

therefore not as prominently researched in environmental

behavior studies as normative or attitudinal aspects. If

environmental knowledge is studied, it is often restricted to

subjective self-report measures of ability (e.g., Duerden and Witt,

2010;Milfont, 2012) or conflated with subjective evaluations

of environmental issues as so-called problem awareness (e.g.,

Bamberg and Möser, 2007). Due to these methodological issues,

it is impossible to derive sound conclusions about the influence

of actual environmental knowledge on pro-environmental

behavior. In cases where it is assessed objectively, different

knowledge types have been suggested with insufficient testing of

this claim (e.g., Roczen et al., 2013;Geiger et al., 2014;Liefländer

et al., 2015) and the relationship to an overarching general

knowledge factor has not been tested up to date, although

general knowledge is a widely researched topic in psychological

research as an individual differences with far-reaching practical

implications. The main aim of this paper is to explore the relation

of environmental knowledge to general knowledge and the

predictive power that each has on environmentally significant

behavior. Toward this end we present an updated, objective

environmental knowledge test, and investigate its measurement

properties that challenge the distinction of different knowledge

types in existing models of environmental competence.

Methodological Issues of Knowledge

Assessment

Without a precise definition of knowledge, it is impossible

to meaningfully investigate the structure of environmental

knowledge and its predictive power on corresponding behavior.

Knowledge is the result of a person’s lifelong learning process,

i.e., the voluntarily accessible and organized accumulation of

veridical information (facts, rules, etc.). Veridical means that

the information is unequivocally true or false. To evaluate an

individual’s performance in a knowledge test, her responses

must be compared to the veridical answer of a given item

or question, and consequently be rated as either “correct”

or “wrong” (Cronbach, 1949). In contrast, many studies in

environmental research use confidence or agreement ratings

that assess self-concepts of one’s own knowledge, i.e., “I can

explain what the term ecology means,” (Duerden and Witt,

2010), “How well-informed do you consider yourself to be

on global warming and climate change?” (Milfont, 2012), or

“How would you rate your knowledge/ability/awareness about

composting/recycling/sustainability. . .?” (Redman and Redman,

2014). These tests do not measure actual knowledge, but

a meta-representation of subjective knowledge (Metcalfe and

Shimamura, 1996). Basic research on self-reports of cognitive

abilities shows weak to zero relations with objective ability

measures (Mabe and West, 1982;Jacobs and Roodenburg,

2014), corroborated recently for sustainability related knowledge

(Effeney and Davis, 2013). These low correlations are suggestive

of that the so-called Dunning-Kruger effect also exists in the

environmental domain, where especially novices in a certain

domain grossly overestimate their own expertise (Kruger and

Dunning, 1999;Sanchez and Dunning, 2018).

Another prominent approach in environmental research is the

subjective evaluation of the severity of a certain environmental

issue on again, a Likert-type agreement rating scale, e.g., “Air

pollution from private car use is a threat to plants and animals

in the world” (e.g., Eriksson et al., 2006). This approach is

problematic when it is conflated with objective knowledge

measures, as in the meta-analysis undertaken by Bamberg and

Möser (2007), for example. Their conclusion of an indirect

influence on behavior via normative variables and feelings of guilt

cannot be definitely attributed to objective, verifiable knowledge.

Likewise, the meta-analysis’s results could also reflect effects of

subjective feelings about whether an environmental issue is a

problem. As with confidence ratings, a recent study by Ünal et al.

(2017) could show that problem awareness correlated only weakly

with general knowledge on climate change consequences and not

at all with more specific knowledge on the topic.

These examples show that subjective measures of

environmental “informedness” and problem awareness are

not the same as objective knowledge. Despite recent evidence,

that these two types of knowledge are subject two different

psychological processes (Dunning and Helzer, 2014) and

might contribute to behavior in different ways, there is not

much systematic investigation into these differences in the

environmental domain. One notable exception is a recent

study by Passafaro and Livi (2017) that also evidenced an only

moderate correlation between perceived and actual recycling

skills (r= 0.33). More interestingly, they found that perceived

skills are indicative of if people are likely to act at all (motivation),

and actual skills relate to how well they will perform (accuracy).

To advance systematic research on the role and structure

of environmental knowledge in according behavior, objective

instruments are indispensable, even if their development is

more complex, as they have to be validated for veridicality,

content, uniqueness of correct answers, difficulty levels of

adequate distractors (false answer options) or are based on

highly ecologically valid simulation tasks (Passafaro et al., 2016).

This study contributes to this endeavor with the actualization

and advancement of such an instrument based on existing

instruments (Kaiser and Frick, 2002;Geiger et al., 2014;

Braun and Dierkes, 2017).

Specific Types of Environmental

Knowledge?

The environmental-competence model by Kaiser and colleagues

(Frick et al., 2004;Kaiser et al., 2008;Roczen et al., 2013) is

based on an objective knowledge test that advocates domain-

specific knowledge in environmental education. It specifies

three different knowledge types that have shown to affect

environmental behavior to different degrees. According to

this model, system related knowledge comprises declarative

knowledge about the ecological system and natural laws

(e.g., “What are constituents of biodiversity?”). Action-related,

procedural knowledge about possible actions for environmental

conservation (e.g., “Does buying organic food help to conserve

Frontiers in Psychology | www.frontiersin.org 2April 2019 | Volume 10 | Article 718

fpsyg-10-00718 March 29, 2019 Time: 18:51 # 3

Geiger et al. General Knowledge and Environmental Behavior

biodiversity?”) is expected to exert a strong direct effect on

behavior. Effectiveness knowledge relates to information about

relative effects of different behaviors (e.g., “Is buying organic food

more effective to conserve biodiversity than giving up certain

products?”) and is assumed to have a motivational function on

engaging in relevant behaviors.

The assumed structure of three facets of environmental

knowledge – system, action, effectiveness – has rarely been

empirically tested against a parsimonious one factor model,

where all items are modeled onto a single dimension. When it

has, only one study found marginally better fit parameters for

the three factor model (Frick et al., 2004). One study found

no difference (Kaiser and Frick, 2002) and further studies did

not explicitly test different alternatives (Roczen et al., 2013;

Geiger et al., 2014;Liefländer et al., 2015). Taken this sparse

evidence, it remains unclear whether a multifaceted structure

of environmental knowledge, although theoretically plausible,

is an empirically warranted, appropriate model of people’s

actual knowledge.

General Versus Domain Specific

(Environmental) Knowledge

Unlike research on environmental knowledge, investigating the

structure and influence of general knowledge on behavioral

outcomes has a long tradition in psychological research on

individual differences (Cattell, 1971). In intelligence research,

general knowledge is a well-established construct subsumed

under crystallized intelligence (“gc”). It denominates the corpus

of acquired factual knowledge and a person’s experience over

a lifetime which predicts important outcomes in everyday

life, such as academic achievement (Ones et al., 2005) or

job performance (Hunter, 1986;Lievens and Patterson, 2011).

While most cognitive abilities decline during adulthood, general

knowledge grows with increasing age, up until late adulthood

(Schroeders et al., 2015). In contrast, fluid intelligence (“gf”)

comprises abilities such as reasoning, working memory capacity,

etc., which are independent of a specific learning history (for a

detailed description of both, see Carroll, 1993).

Theoretically, general knowledge might cover any domain,

from knowledge about current events (e.g., celebrities) to detailed

knowledge about decimals of π. The so-called “knowledge-is-

power” hypothesis assumes that independent, domain-specific

factors of knowledge are shaped by individual learning history

in a certain domain, i.e., gaining expertise (Hambrick and Engle,

2002;Hambrick, 2003;Hambrick and Oswald, 2005) and that

this expertise explains domain specific performance. However,

factor analytic research puts the domain specificity into question

(Hambrick and Meinz, 2011). In several large and broadly varying

samples, a carefully developed gc test with multiple domains (e.g.,

history, medicine, literature, and many more distinct disciplines)

always points to a general factor of gc with no evidence of domain

specificity (Schroeders et al., 2013;Schipolowski et al., 2014a,b;

Wilhelm et al., 2014).

Environmental knowledge might be considered another

knowledge domain – a domain of specific interest in research

concerning pro-environmental behavior or reaching sustainable

development goals in a wider scope. Whether environmental

knowledge is embedded in the general accumulated knowledge

base of people has not yet been tested. The relationship

between environmental and general knowledge is important for

practitioners, as it will add empirical evidence to the discussion

on appropriate approaches to environmental education. While

an environmental-specific ability approach is advocated by

Kaiser et al. (2008), recent literature on education for

sustainable development advocates the concept of broad,

content-independent competencies, such as perspective taking,

participation, or handling of complex information (Haan et al.,

2008;Michelsen and Fischer, 2017). If environmental knowledge

shows systematic variation between participants after controlling

for general knowledge, and if this residual variation predicts

environmental behavior, the knowledge-is-power hypothesis will

be supported. On the other hand, if environmental knowledge

is empirically indistinguishable from general knowledge or does

not account for environmental behavior and we will interpret

this as evidence for a general knowledge model in explaining

domain-specific behavior.

The Predictive Power of Knowledge in

Environmental Behavior

It has been convincingly shown that a simple information deficit

model, where the provision of information is deemed sufficient

to change the relevant behavior, is inappropriate (Kollmuss and

Agyeman, 2002;Schultz, 2002). However, this evidence should

not dismiss knowledge as a meaningful predictor of behavior

altogether. Theoretically, knowledge can be considered a limiting

factor for environmental behavior: if people do not know about

environmental effects of their behavior, they cannot intentionally

adjust their behavior toward generating less environmental

impact. Kaiser (1998);Kaiser and Wilson (2000) was one of the

first authors to present a comprehensive measure of ecological

behavior, spanning behaviors from different areas of life as

mobility, waste management, consumerism, energy consumption

and vicarious behavior. Critique on their intention-based

approach, where the intention to seek to minimize environmental

damage solely defines pro-environmental behavior was currently

raised (Steg and Vlek, 2009;Huddart Kennedy et al., 2013;

Geiger et al., 2017). These authors call for inclusion of actual

impact criteria in the investigation of environmental behaviors

in the social sciences, such as e.g., the ecological or carbon

footprint of the services and products consumed by people.

The current paper applies such an impact-based idea, where

environmental behaviors are selected for their actual high impact

on the environment.

With regards to the role of knowledge for environmental

behavior, two extensive meta-analyses on the explanation of

pro-environmental behavior (Hines et al., 1987;Bamberg and

Möser, 2007) have supported an indirect role mediated via

moral norms, feelings of guilt, and/or intentions to act more

environmentally friendly. Various empirical papers show that

when knowledge is measured objectively, direct relationships

can be observed, explaining anywhere between 3 and 24% of

behavioral variance (Kaiser and Frick, 2002;Frick et al., 2004;

Frontiers in Psychology | www.frontiersin.org 3April 2019 | Volume 10 | Article 718

fpsyg-10-00718 March 29, 2019 Time: 18:51 # 4

Geiger et al. General Knowledge and Environmental Behavior

Meinhold and Malkus, 2005;Roczen et al., 2013;Geiger et al.,

2014, 2018;Braun and Dierkes, 2017). The notion of an

indirect, necessary, yet insufficient role of knowledge to bring

about informed behavior is also brought forward by Kaiser

and Fuhrer (2003), along with three arguments on why the

influence of knowledge on behavior has been systematically

underestimated: (a) the existence of overruling more direct

influences as situational restrictions, (b) inadequate assumptions

about the structure of knowledge, and (c) the use of inadequate

statistical procedures that do not account for measurement

errors. In the current paper, we will scrutinize the latter two

obstacles: we investigate the adequate knowledge structure using

latent modeling techniques to see if an influence of knowledge on

behavior is replicated under such circumstances.

RESEARCH GOALS

Responding to the methodological problems outlined in the

introduction, the first goal of this study is to present an

updated and expanded measurement instrument for objective

assessment of environmental knowledge and to test its internal

structure concerning content subdomains and knowledge types

(cf. Kaiser et al., 2008). As an improvement compared to

existing scales, which focus on natural science aspects of

ecology, the new test includes current issues in sustainability

related social, political, and economic topics (see Leach

et al., 2013). As basic intelligence research has called into

question the “knowledge-is-power” hypothesis, the second goal

is to investigate whether environmental knowledge should be

regarded an independent knowledge domain or a subdomain

of general knowledge. Our third goal is to investigate whether

environmental knowledge predicts pro-environmental behaviors

under best of measurement conditions over and beyond the

influence of general knowledge.

MATERIALS AND METHODS

Procedure

An online survey in German language was conducted with the

survey tool Unipark and published online from July to September

2015. The link was advertised to a larger German-speaking online

community sample and additionally advertised on German social

media webpages. The survey included a short sociodemographic

questionnaire followed by three scales in the following order:

general knowledge (Berliner Test zur Erfassung Kristalliner

Intelligenz, BEFKI), the updated environmental knowledge test

(EKT) and a short impact based scale of environmental behavior

(SIBS). The study was conducted according the ethical guidelines

for online studies of the German Society for Online Research

(Deutsche Gesellschaft für Online-Forschung, 2007). Consent of

each participant was requested in digital form on the first page

of the survey and anonymity of participants was guaranteed.

Participation was voluntary and 10 online vouchers worth 20€

were raffled among the participants who completed all measures

and opted in for the raffle. Ethical approval was not required as

per local legislation.

Sample

The website of online study was accessed 443 times. The dataset

was cleaned by excluding participants who dropped out before

the end (n= 228, including instant drop-outs, not accepting

study terms and conditions, study restarts, etc.), because study

aborts must be treated as a reversal of participation consent.

Furthermore, participants that responded to at least one out of

four attention check questions erroneously (n= 1; an example

item reads “This is an Attention Check, please respond with B”),

or showed clear response patterns (n= 0, i.e., with little-to-no

within person variance across response options) were excluded.

Overall, 214 subjects remained and are included in all subsequent

analyses. Mean age is 32.4 years (SD = 11.7 years, ranging from 18

to 92 years of age, German population median 18+: 50 years) and

131 participants are female. With respect to education (German

population in parentheses, Statista, 2014), 5.9% (32.9%) have a

basic school education, 16.8 % (29,4%) have a simple high school

diploma, 39.2 % (29,5%) have passed high school with a university

entrance diploma or comparable, and 38.3% possess a university

degree. Our sample is thus younger and better educated than

the German average.

Measures

Environmental Knowledge Test (EKT)

The environmental knowledge test was constructed to cover

relevant core areas of ecology, climate, resources, environmental

contaminants/health, and consumption behaviors. Based on a

wider understanding of sustainability explicitly endorsing a

socio-economic dimension (Farley and Smith, 2013;Leach et al.,

2013), we additionally included items measuring knowledge

about sustainability-related societal and economic issues. The test

was constructed in a multiple-choice format with one correct

answer and three distractors (see Appendix A) to be comparable

to the validated test of general knowledge in terms of form,

language, and difficulty (see below and Appendix B).

In the first round, items from existing scales were screened

for relevant content and compliance with test construction

standards such as veridicality and comparability of distractors

based on their wording. This step yielded 7 suitable items from

the environmental knowledge scale by Kaiser and Frick (2002):

items 1, 2, 3, 7, 13, 23, 32) 7 items from the EKLA scale

(Geiger et al., 2014: items 6, 8, 15, 17, 18, 20, 21) and 3 items

from Schahn’s scale (Schahn, 1999: items 24, 33, 34). Nineteen

additional items were constructed on the basis of ecology and

environmental study books, course curricula of German schools,

and webpages of official environmental institutions (e.g., German

federal environmental agency1), German federal ministry for

the environmental, nature conservation, building and nuclear

safety (BMU2).

The resulting 36 items can also be classified according to

their knowledge type (cf. Frick et al., 2004), yielding 21 system

1www.umweltbundesamt.de

2www.bmub.bund.de

Frontiers in Psychology | www.frontiersin.org 4April 2019 | Volume 10 | Article 718

fpsyg-10-00718 March 29, 2019 Time: 18:51 # 5

Geiger et al. General Knowledge and Environmental Behavior

knowledge items, 7 action-related knowledge items, and 8

effectiveness knowledge items.

Response format of all items was multiple choice and

constructed or adapted to have one correct answer and three

distractor answers. For a list of all German items included in

the final form (n= 35, 1 item excluded and 3 more items

adapted) with their according thematic clusters, see Appendix A

in Supplementary Material. The English translation has not been

validated and is only displayed for illustrative purposes.

General Knowledge (BEFKI gc)

One extensively validated instrument for the assessment of

general knowledge is the Berlin Test of Fluid and Crystallized

Intelligence (BEFKI gc, Schipolowski et al., 2013;Wilhelm

et al., 2014). The crystallized intelligence section of this test

comprises 64 items that capture general knowledge in a multiple

choice form. The 16 content domains reflect the German

national curriculum and comprise subjects from natural sciences

(chemistry, physics, geography etc.) humanities (music, arts,

history, etc.), and social sciences (finance, economics, religion,

etc.). Confirmatory factor analysis (CFA) has repeatedly revealed

a strong general factor with high reliability (ω= 0.88) and little

domain specificity (Wilhelm et al., 2014). It is available in three

versions for different education levels (until grade 8, grade 8 to

10, grade 11+), as well as in parallel versions for applied and

research purposes. As we aimed for a wide age range (18+), and

as gc increases until late adulthood, we used the version for grade

11+. For three example items of the domain music, history, and

chemistry, see Appendix B in Supplementary Material.

Short Impact Based Pro-environmental Behavior

Scale (SIBS)

The SIBS comprises five of the six behavioral areas mobility,

private energy consumption, waste management, consumption

choices, and social behaviors assessed by Kaiser (1998), plus

nutrition, which was added because of the high environmental

impact of nutrition-related behaviors alongside the consumption

fields of housing and mobility (EEA, 2013;Thøgersen, 2014;

Geiger et al., 2017). The items were selected according to their

environmental impact in terms of greenhouse gas emission or

ecological footprint. Within each domain, so-called “big points”

(Bilharz and Schmitt, 2011) were addressed, as e.g., heating

or solar energy production for private energy consumption, or

airplane and individual motor travels for the mobility sector

and meat consumption and regional food for nutrition (Lorek

and Spangenberg, 2001;Tukker et al., 2010), with the exception

of three general consumption and two social behaviors that

do not have a quantifiable direct environmental impact (items

11–15). Response format was a 5-point Likert scale ranging

from 0 = “never” to 4= “always.” Two dichotomous items

on electricity-provision were combined into one 5-point scale

-item (neither item affirmed = 0, renewable energy provider

affirmed = 2, own solar panel = 3, both items affirmed = 4).

Car use was assessed with three questions on ownership, gasoline

usage and fuel type, that were combined to an overall 5-point

index of car use. After excluding two items on food waste and use

of deposit bottle due to negative loadings, the final form included

18 items (see Appendix C in Supplementary Material). With

only two to four items per subdomain, it is a short scale for the

assessment of environmentally relevant behaviors.

Methodological Framework and

Statistical Analysis

This study uses structural equation modeling as a framework, a

modeling technique that allows to combine factor analytic with

path analytic research questions (for an introduction see Kline,

2011). The factor analytic approach allows to test measurement

models, i.e., if latent constructs (as e.g., environmental knowledge

in our case) are measured appropriately with items that

adequately represent the latent factor. The path or regressional

analytic approach reveals underlying relationships between these

latent factors in a structural model (e.g., how much behavioral

variance is explained by knowledge). Model fit parameters are

used to indicate the overall quality of a specified model including

both aspects of the model, the measurement and structure

model(s). Therefore, the measurement model of a latent variable,

especially when using a new test, has to be tested separately before

including it in an overall structural model, this was done here for

environmental knowledge as well as for behavior.

The statistical analyses to this end were conducted using

R 3.4.1 (R Core Team, 2014), setting the threshold of any

statistical significance testing in this manuscript to α= 0.05.

We calculated McDonald’s (1999) with the package semTools

(version 0.4-13; semTools Contributors, 2016) as an adequate

index of reliability for latent variable modeling (Sijtsma, 2009).

Item analysis was conducted using the package psychometrics

(version 2.2., Fletcher, 2010). To investigate the factor structure

of the EKT, we tested nested confirmatory factor models using

the DIFFTEST option in Mplus (Muthén and Muthén, 2010) due

the dichotomous nature of our data (correct/wrong), requiring

the use of Weighted Least Squares Mean and Variance estimation

(WLSMV, Muthen et al., 1997). Other latent analyses were

conducted using the package lavaan (Rosseel, 2012) in R with

Maximum Likelihood (ML) estimation.

The CFA and structural equation models (SEM) in section

“Relationship of Environmental Knowledge and General

Knowledge” and “The Role of General Knowledge and

Environmental Knowledge in Environmental Behavior” were

conducted with parcels as indicators, as our sample was to small

to estimate these models on an item level, which would have

required 97 indicators (Little et al., 2002). Parcels as proportion of

correct scores were built for all knowledge and behavioral items

belonging to a content domain. For example, all knowledge items

related to financial issues were combined to a finance parcel and

all knowledge items related to climate change were combined to

a climate parcel. We proceeded likewise with behavioral items.

For thematic classification of environmental knowledge items,

see Appendix A and for behavioral items, see Appendix C.

To evaluate the goodness of fit of our models, we use root

mean square error approximation (RMSEA) and comparative fit

index (CFI). RMSEA is an absolute index of model fit reflecting

the discrepancy between observed and postulated model, and

therefore should approximate 0 (“perfect fit”). The CFI, on the

Frontiers in Psychology | www.frontiersin.org 5April 2019 | Volume 10 | Article 718

Loading more pages...