1

B. S. Jorgensen and R. C. Stedman

Measuring Sense of Place:

Lakeshore Owners' Attitudes Toward their Properties

by

Bradley S. Jorgensen

Department of Rural Sociology

University of Wisconsin-Madison

Richard C. Stedman

Department of Rural Sociology

University of Wisconsin-Madison

April, 1999

All correspondence should be sent to Bradley S. Jorgensen,

Department of Rural Sociology, University of Wisconsin,

420 Agriculture Hall, 1450 Linden Drive,

Madison, Wisconsin 53706.

[e-mail]

Fax: [phone]

Ph: [phone]

Measuring Sense of Place

2

Abstract

Despite the theory-laden character of Sense of Place (SOP) research,

empirical investigations have been surprisingly few in number. Some

researchers have argued that a lack of adequate measuring instruments has

impeded empirical investigations of human relationships with physical

surroundings. Existing attempts to measure SOP are open to a number of

different interpretations, some of which are well established in attitude research.

Attitude theory can provide a basis for conceiving of SOP as cognitive,

affective and behavioral reactions to a spatially demarcated object. Sense of

Place was defined as a multidimensional construct comprising (1) beliefs about

the relationship between self and place; (2) feelings toward the place; and (3) the

behavioral exclusivity of the place in relation to alternatives. A 12-item SOP

scale, consistent with a multidimensional theoretical prescription, was developed

and subsequently tested in the field with a sample of lakeshore property owners

in northern Wisconsin.

A number of measurement models were posed as potential explanations

of the scale's construct validity. Results suggested that the SOP scale measured

a general Sense of Place dimension that gained expression in property owners'

thoughts, feelings and behavioral commitments for their lakeshore properties.

There was also clear support for the existence of three univariate dimensions

having interpretations consistent with Place Identity, Place Attachment, and

Place Dependence. The degree of covariation between these primary constructs

was indicative of shared variability with the general SOP variable that was most

3

B. S. Jorgensen and R. C. Stedman

synonymous with feelings of Attachment and least comparable to beliefs about

Identity. However, this general evaluative dimension was more explanatory of

observed responses than were the component constructs. The dominance of the

SOP factor over the narrower dimensions was prevalent in three different

measurement models that posited both general and specific factors.

The SOP scale can provide researchers with an easily administered and

reliable means by which to measure attitude toward spatially demarcated objects.

Directions for future research are discussed.

Measuring Sense of Place

4

Measuring Sense of Place:

Lakeshore Owners' Attitudes Toward their Properties

Introduction

There are a plethora of concepts describing the relationship between

people and spatial settings, but Sense of Place is perhaps the most widely used.

Sense of Place (SOP) has been referred to as an overarching concept which

subsumes other concepts describing relationships between human beings and

spatial settings (Shamai, 1991). In a general sense, SOP is the meaning

attached to a spatial setting by a person or group. Tuan (1979) has provided the

most oft-cited definition, declaring that a place is a center of meaning or field of

care that emphasizes human emotions and relationships. Ryden (1993) added

that "a place…is much more than a point in space…but takes in the meanings

which people assign to that landscape through the process of living in it" (pp. 37-

38). Accordingly, SOP is not imbued in the physical setting itself, but resides in

human interpretations of the setting.

Despite this common beginning, different researchers have emphasized

different constructs, but three appear in the literature with some regularity: Place

Identity, Place Attachment and Place Dependence. Place Identity involves

"those dimensions of self that define the individual's personal identity in relation

to the physical environment by means of a complex pattern of conscious and

unconscious ideas, beliefs, preferences, feelings, values, goals and behavioral

5

B. S. Jorgensen and R. C. Stedman

tendencies and skills relevant to this environment" (Proshansky, 1978, p. 155).

As a cognitive structure, Identity is a substructure of a more global self-

identification in the same way that one might consider gender identity and role-

identity (Proshansky, Fabian, & Kaminoff, 1983).

Place Dependence (Stokols & Shumaker, 1981) is based on Thibaut and

Kelly’s (1959) “comparison level / comparison level for alternatives” model, which

is a person's rational analysis of the probability of remaining in a relationship.

This process involves a comparison of the current outcomes to those that would

be obtained by selecting an alternative course of action. Thus, Dependence

concerns how well the setting serves goal achievement ("how does this setting

compare to others for what I like to do?").

The third construct - Place Attachment - is widely cited and subject to a

variety of definitions. Some scholars consider it to have a predominantly

emotional content. For example, Riley (1992) emphasized Attachment as the

"affective relationship between people and the landscape that goes beyond

cognition, preference, or judgement" (p. 13). Others have suggested that

Attachment "involves an interplay of affect and emotions, knowledge and beliefs,

and behaviors and actions in reference to a place" (Altman & Low, 1992, p. 5).

Research has not focused on the interplay between these variables,

leading some to suggest "construct overload" and note vague relationships

between place variables (Hammitt & Stewart, 1996). This may in part be to the

diversity of approaches utilized in understanding SOP. Lalli (1992) divides SOP

theory and research into phenomenological and positivistic approaches.

Measuring Sense of Place

6

Positivistic research on SOP is characterized by researcher-defined variables,

quantitative methods, and traditional hypothesis testing. In contrast,

phenomenological approaches to understanding SOP address the intentional

interaction between person and environment: "the world of things, persons, and

events as experienced by the individual" (pp. 286).

Most of the major place theorists, (e.g., Relph, Tuan, Sack) either explicitly

identify place research as a phenomenological endeavor or otherwise do not use

quantitative methods to “test hypotheses” in any formal sense. Rather, these

scholars make strong statements about the general nature of SOP. For

example, Relph (1976) asserts that attachment to place grows through time and

is based strongly on relationships with people in the setting rather than the

physical environment. Places to which we are most attached are those where

we have had a wide variety of experiences; “the identity of a place…varies with

the individual, group, or consensus image of that place (p.56). Many of these

statements, despite their phenomenological underpinnings, suggest testable

hypotheses about the nature of place. Many theorists suggest, however, that

there are perils in improperly dissecting a multidimensional concept (Altman and

Low 1992, Hummon 1992). They argue that the researcher cannot separate

components of SOP without losing the meaning of the overall concept.

Approaches to Measuring Sense of Place

Perhaps owing in part to the phenomenological emphasis, empirical

investigations of SOP utilizing quantitative methods have been few in number,

7

B. S. Jorgensen and R. C. Stedman

and have generally lagged behind theory (Shamai 1991; Lalli, 1992) despite calls

for measurement (Krupat 1983). Five approaches to empirical measurement are

described in the following section.

Shamai's (1991) Sense of Place Scale

Shamai (1991) recognized the need for an easily administered means to

measure SOP, and addressed the shortfall by providing a scale based on Relph's

(1976) seven ways of sensing a place. This scale represented four ordered

categories ranging from an absence of a SOP to a profound commitment toward

a place. Respondents were required to identify a position on the scale that best

reflected their relationship to the place in question.

Shamai (1991) stated that the scale measures variability in the intensity of

feelings and behavior of people residing in the same place at a particular point in

time. For example, respondents lacking a sense of place would not be expected

to be prepared to sacrifice their interests for it. In other words, respondents who

were prepared to endure some form behavioral cost would also hold developed

beliefs and strong feelings about the place. The scale is unidimensional, which

contrasts somewhat with suggestions made by other theorists that emphasize

the multidimensional nature of SOP.

Measuring Sense of Place

8

The Cuba and Hummon (1993) Place Identity Measure

Cuba and Hummon (1993) focused on the concept of Place Identity, which

they defined as an expression of "at homeness". Three elements of Identity were

identified and measured: (1) its existence ("Do you feel at home here?"); (2) its

affiliations ("Why do you feel at home here?"); and, (3) its locus ("Do you

associate feeling at home with living in this particular house or apartment, with

living in this community, or with living on the Cape [Cod] in general?"). This

approach to measuring Identity accounts for an absence of the construct for

certain individuals as well as any relationship between the reasons for feeling at

home and the specific place loci. Unlike other approaches, however,

respondents' intensity of feelings toward a place are not addressed.

McAndrew's (1998) Rootedness Scale

McAndrew (1998) sought to measure the concept of "rootedness" which

he equated with an affective interpretation of place attachment. Rootedness was

assumed to have positive and negative components that defined the ends of a

general bipolar concept. The positive dimension - labeled "Desire for Change" -

was measured with six belief statements with Likert response scales (e.g.,

"Moving from place to place is exciting and fun"). The negative dimension was

measured with four belief statements and labeled "Home/Family Satisfaction"

and included items such as "I love to reminisce about the places I played when I

9

B. S. Jorgensen and R. C. Stedman

was a child". Due to McAndrew's (1998) use of an orthogonal rotation in his

Principal Components analysis, the two dimensions of rootedness were not

correlated. Cronbach's alpha for the positive and negative subscales were 0.79

and 0.70, respectively.

The two-component structure of the rootedness scale was replicated on

an independent sample of undergraduate students. However, the item loadings

were only moderate (particularly for Home/Family Satisfaction) despite the

Principal Components technique in which item communalities are set to unity.

The reliability coefficients also decreased to 0.56 (Desire for Change) and 0.51

(Home/Family). Further, there was a significant negative correlation between the

two subscales, despite the assumption of independence manifest in the

orthogonal rotation method. McAndrew (1998) noted that that this association

was "consistent with the conceptualization of Home/Family Satisfaction and

Desire for Change as opposite ends of the same dimension" (p.415).

Lalli's (1992) Urban Identity Scale

Consistent with theoretical approaches that emphasize

multidimensionality, Lalli (1992) developed the Urban Identity Scale and

validated it in a sample of Heidelberg residents. This instrument comprised five

subscales of urban-related identity: Evaluation, Familiarity, Attachment,

Continuity, and Commitment. The first subscale referred to evaluative

comparisons regarding the uniqueness of the town relative to other towns (e.g.,

Measuring Sense of Place

10

“There are many things here which are envied by other towns”). Familiarity was

theorized as a well-developed cognitive orientation grounded in everyday

experience, and was measured with items such as "When I amble through

Heidelberg, I feel very strongly that I belong here". The Attachment dimension of

urban identity corresponded with general sense of place definitions in the

literature. Lalli defined Attachment in terms of feelings of belonging, and

measured it with items like "I feel really at home at Heidelberg." Continuity was

defined by Lalli as the extent to which the respondent perceives his or her past

experiences as being synonymous with the history of the town. One item

representative of those included in the Continuity subscale was "Lots of things in

the town remind me of my own past." Finally, the Commitment component of

urban identity referred to the perceived significance of the town in one's future

and was measured by items that reflected a commitment to a future relationship

with the town (e.g., "I would like to stay in Heidelberg indefinitely").

All five subscales showed strong correlations with one another that ranged

from 0.38 to 0.82. The highest average inter-scale correlation occurred for the

General Attachment dimension (average r = 0.71). These summary statistics

suggest that the five dimensions of the Urban Identity Scale are indicative of a

more simple factorial structure best represented by Lalli's concept of Attachment.

This interpretation of the data is supported when the subscale correlations and

reliabilities were subjected to confirmatory factor analysis. The scale loadings on

a general factor ranged from 0.32 (t = 7.69, p < .001) for Evaluation to 0.87 (t =

15.40, p < .001) for General Attachment.

11

B. S. Jorgensen and R. C. Stedman

The Williams et al. (1992) Place Attachment Scale

Williams et al. (1992) also adopted a multidimensional approach. These

authors equated Place Attachment with SOP and define it as an emotional bond

between and individual and a particular spatial setting. They identified two

dimensions of Attachment: Identity and Dependence. Dependence was

conceived in terms of how the setting compares with other alternatives

supporting similar behavioral goals. Identity, on the other hand, referred to a

form of attachment resulting from the symbolic importance of the physical

environment to self-definition:

Thus, in addition to being a resource for satisfying explicitly felt behavioral

or experiential goals, a place may be viewed as an essential part of one's

self, resulting in strong emotional attachment to place (Williams et al.,

1992, p. 32).

In measuring Attachment, Williams et al. (1992) employed a 13-item scale

requiring respondents to rate their agreement with statements such as "This

place means a lot to me" and "I wouldn't substitute any other area for doing the

type of things I did here." Responses to the items were aggregated to form a

Place Attachment scale. No information was provided by the authors as to the

extent of the correlation between the Dependence and Identity subscales, but

presumably it was reasonably large given the high overall scale reliability

(Cronbach's

α

= 0.93) reported in the study.

Measuring Sense of Place

12

Summary

The empirical approaches to measuring SOP detailed above vary

considerably in their operation and in the degree to which they consider the

multidimensionality of the concept. We suggest that empirical measures that

attempt to incorporate multiple dimensions of SOP are more consistent with the

theoretical strands of SOP detailed above. However, the multidimensional and

unobservable nature of SOP is considered by some theorists to preclude

empirical measurement (e.g., Altman & Low, 1992).

Although none of the researchers cited above label them so, place-related

constructs can be regarded as attitudes. Within this general framework, Sense

of Place is a complex mental structure that organizes self-referent cognitions,

emotions and behavioral commitments. Sense of Place viewed in this way is

consistent with conceptions of attitude.

In the following section we explore models of attitude structure as a basis

upon which to address the measurement of SOP. Importantly, we do not claim

that the complexity of the construct is completely accessible to quantitative

measurement. The process of measuring deeply emotional constructs (like

affect) by way of verbal reports is one that is mediated by cognitions (Heberlein,

1981). Nevertheless, some aspects of SOP are consciously held (Tuan, 1980)

and amenable to empirical investigation. We attempt to provide an empirical

exploration of the multidimensional foundation of SOP represented by its

13

B. S. Jorgensen and R. C. Stedman

affective (Place Attachment), cognitive (Place Identity), and behavioral (Place

Dependence) dimensions.

Models of Attitude Structure

Attitude can be defined as a response to an exogenous event, object or

stimulus (Fishbein & Ajzen, 1975). Affect, cognition and behavior are three

distinguishable components of response to an attitude object. Affect refers to

emotional responses or activity in the sympathetic nervous system, as reflected

in heart rate, galvanic skin response, or verbal self-reports. Beliefs, knowledge

structures, percepts, and thoughts are all representative of the cognitive

component of attitude. The behavioral (or conative) component in a tripartite

view of attitude includes verbal reports of behavioral intentions and behavioral

commitments.

Some attitude theorists assume that the distinction between the response

categories in which attitude is expressed represent different theoretical

components of attitude (see Smith, 1947; Katz & Stotland, 1959; Ostrom, 1969;

Kothandapani, 1971; Bagozzi, 1978; Bagozzi, Tybout, Craig, & Sternthal, 1979;

Breckler, 1984). Thus, instead of a unidimensional construct expressed in

beliefs, emotions, and behavioral intentions, attitude is conceived as being a

multidimensional construct. While these more or less distinct components may

vary along an evaluative continuum, it is assumed that the evaluations expressed

in each domain can potentially differ substantially for certain attitude objects. For

Measuring Sense of Place

14

example, a person may feel favorable toward their lakeshore property, but

consider it peripheral to their identity and an irrelevant domain of action.

Other conceptions of attitude posit a single dimension that is

fundamentally evaluative (Eagly & Chaiken, 1993, Fishbein & Ajzen, 1975; Dillon

& Kumar, 1985). This evaluative factor can be expressed in cognitive, affective

and conative responses, but these three classes are not regarded as separate

components of attitude. Rather, the affective, cognitive and conative realms are

regarded as domains of attitude expression or classes of observed responses to

attitude objects. A central assumption of the attitude concept in this respect is

that a common evaluative continuum is reflected in the three classes of observed

responses posited in the tripartite model.

A relatively small body of literature has attempted to confirm the

discriminant validity of measures constructed to reflect the different components

(e.g., Breckler, 1984; Breckler & Wiggins, 1989). However, the research findings

have failed to provide convincing and consistent support for one model over the

other (Bagozzi & Burnkrant, 1985; Dillon & Kumar, 1985; Widaman, 1985).

Debate between proponents of the single-factor and three-factor models has

centered on whether empirical discrimination between attitude components

results from method differences or actual differences between theoretically

independent concepts.

The debate regarding attitude structure was advanced by the introduction

of a higher-order model (Rosenberg & Hovland, 1960). This hierarchical model

posited the existence of a general evaluative dimension responsible for the

15

B. S. Jorgensen and R. C. Stedman

covariation between the cognitive, affective, and conative components of the

tripartite conception. The primary factors were theorized as separate mediating

structures between attitude and the three classes of observed responses. The

three components were defined as independent constructs that might potentially

comprise a single attitude dimension at a higher level of abstraction.

The following section applies the attitude models discussed above to

SOP, and highlights the interpretative differences following from each

conceptualization. Some additional models are also introduced as alternative

bases upon which to assess the construct validity of SOP measurements.

Sense of Place Measurement Models

Like attitude, SOP is a hypothetical construct that is not accessible to

direct observation, but can be inferred on the basis of measured responses.

When conceived as an individual's favorable or unfavorable attitude toward

spatially demarcated object, SOP can be inferred from responses of a cognitive,

affective or conative nature. When each of these classes of response is

regarded as being mediated by a distinct construct, the place concepts of

Identity, Attachment, and Dependence are evoked, respectively. Place Identity

can be regarded as an individual's beliefs, perceptions or thoughts that the self is

invested in a particular spatial setting. Place Attachment can be defined in terms

of an individual's feelings about a spatial setting. Finally, Place Dependence can

Measuring Sense of Place

16

be considered as the behavioral exclusivity of a spatial setting relative to other

settings.

Three-Factor Model

Figure 1 displays a diagram of the tripartite model of SOP. Each

component is represented as a distinct construct, although potentially correlated

with one another. This model assumes that Identity, Attachment and

Dependence can differ greatly within individuals. For example, a person may

feel content in a spatial setting, but believe that it is poorly serviced and refuse to

send their children to school there.

____________________

Insert Figure 1 about here

____________________

Single-Factor Model

When the three components of SOP correlate perfectly with one another,

the model shown in Figure 1 reduces to a single factor model (see Figure 2).

That is, all three components become indistinguishable and hence, their

originally distinct interpretations collapse into one. From a unidimensional

attitude view, this single factor would be best expressed in the Attachment items

that more directly deal with affect (Fishbein & Ajzen, 1975).

17

B. S. Jorgensen and R. C. Stedman

____________________

Insert Figure 2 about here

____________________

Empirically, the single-factor and 3-factor models are nested when the

former is expressed in a tripartite form where all factor correlations are equal to

one. Nested models can be compared using a likelihood ratio test. In this way,

models with fewer constraints are compared with models having more

constraints to ascertain whether the additional restrictions improve the

correspondence between the model-implied and sample covariance matrices.

Higher-Order Model

The higher-order interpretation of SOP is displayed in Figure 3. This

model assumes that any correlation between the primary (or specific) factors is

due to a more abstract construct. That is, Identity, Attachment, and Dependence

are regarded as structures that mediate the expression of SOP in observed

responses. However, each component may additionally reflect unique beliefs,

emotions, and behavioral preferences that are independent of one's general

evaluation of the setting. Thus, observed responses to the spatial setting reflect

(1) the indirect effects of SOP through the processes of Attachment, Identity and

Dependence, and (2) the direct effects of individual feelings, beliefs and

behavioral commitments.

Measuring Sense of Place

18

____________________

Insert Figure 3 about here

____________________

Unfortunately, higher-order models involving only 3 primary constructs are

empirically just-identified. That is, while estimates for the parameters in the

model can be derived, not enough information exists to subject the higher-order

part of the model to a goodness-of-fit test. Therefore, the chi-square and

degrees of freedom of the model shown in Figure 3 are equal to the tripartite

model in Figure 1.

In order to over-identify the higher-order level and obtain goodness-of-fit

tests, it is necessary to make additional assumptions about the model. One

assumption that follows from attitude theory is the contention that attitude is best

reflected by evaluative feelings toward an object (Fishbein & Ajzen, 1975). As an

attitude toward a spatially delimited setting, SOP might be more highly correlated

with Place Attachment than with either Identity or Dependence. Thus, equality

constraints could be placed on the paths linking SOP with Identity and

Dependence.

However, where the higher-order structure of the model is concerned, the

only thing being tested is the restriction (i.e., the equality constraint) which is

artificially imposed. One caveat arises from this strategy when the model fails to

demonstrate an adequate approximation of the data. Specifically, a poor fit for

the model could be due to the inappropriateness of the equality constraint, rather

than to the existence of a higher-order factor. One explanation would be that the

19

B. S. Jorgensen and R. C. Stedman

equivalence hypothesis was false, and another would be that the congeneric

measurement hypothesis at the second-order was false.

G+Group Factors Model

An alternative to the hierarchical higher-order model is one in which the

general and specific (or group) factors are regarded as completely independent.

The group factors are not theorized to mediate between the general factor and

the response classes as is the case in the higher-order model. That is, SOP is

not implicated in the explanation of Identity, Attachment or Dependence.

However, like the higher-order model, all the latent variables in the G+Group

model account for variance in the observed variables, but the general factor has

a wider range of influence (with respect to the response domains) than the group

factors.

____________________

Insert Figure 4 about here

____________________

All four factors in the G+Group model have neither a conceptual

relationship nor an empirical relationship with each other. Rather, SOP is

assumed to be expressed in affective, cognitive, and conative responses, and

this expression is independent of any role played by Identity, Attachment or

Dependence. In fact, Mulaik and Quartetti (1997) have pointed out that the

Measuring Sense of Place

20

independence of the factors is an untestable assumption of the model that is

necessary for its empirical identification.

Correlated Uniquenesses Model

The correlated uniqueness (CU) model provides a means of testing a fifth

explanation of SOP. This model, while similar to the G+Group model, is

distinguished by the correlations between residual components of the observed

item variances. While the CU model posits a general SOP factor, it includes the

effects of the group factors rather than the factors themselves. Because the

group factors are not modeled directly, no assumption is made regarding their

structure. These group factors may be unidimensional (as in the case of the

G+Group and higher-order models) or they may be multidimensional. To the

extent that this model fits better than the G+Group model, it can be concluded

that Identity, Attachment and Dependence do not adequately represent the

specific sources of variability influencing responses to the scale items.

____________________

Insert Figure 5 about here

____________________

In summary, attempts to measure SOP are open to a number of different

interpretations, some of which are well-established in attitude research. We do

not attempt to present a formal test of attitude structure, but to develop an easily

administered measure of Sense of Place that is informed by attitude theory.

21

B. S. Jorgensen and R. C. Stedman

Each of the five models of the latent structure of the SOP scale represent

different interpretations of the construct. In the single factor model, the

multidimensionality of SOP is absent compared with the 3-factor model. The

higher-order model attributes a multidimensional structure to SOP, but

subordinates it to a simpler interpretation at a more fundamental level. In

contrast, the general and primary components represented in the G+Group

model are independent of one another. That is, SOP is regarded as a general

evaluative dimension largely independent of unitary conceptions of Identity,

Attachment, and Dependence. Finally, the CU model differs from the G+Group

model by dispensing with the assumption that each group factor is

unidimensional.

Method

Eight lakes in Vilas County (situated in the Northern Highlands Lake

District of North Central Wisconsin) having a mix of privately and publicly owned

shorelines were selected. These types of lakes were chosen because of their

relevance to a number of research questions of interest to the Long Term

Ecological Research project being conducted in northern Wisconsin (see,

Jorgensen, Nowacek, Stedman, Brasier, and Long, in press). The eight lakes

included in the study were Big Muskellunge, Diamond, High, Plum, Razorback,

Sparkling, Trout, and Witches.

Measuring Sense of Place

22

Vilas county tax records served as a sampling frame of households

located within a mile of the shoreline of each lake. A total of 743 households

were located within these areas. Each household was sent a mail questionnaire

that was developed on the basis of field reports and preliminary trials with a small

sample of individuals from the population of households. Following Heberlein

(1978) reminder postcards and replacement questionnaires were sent to

householders who had not replied to the initial mailing.

The response rate after the first mailing was 40 percent. With further

contact, the final response rate was 66 percent after accounting for undeliverable

surveys, refusals to participate, and deceased owners. Of these respondents, 71

percent were shoreline property owners whose responses were retained for

further analysis. Property owners who did not have shoreline frontage were

excluded from the study.

The three Sense of Place components were measured with twelve self-

report items (see Table 1) and 5-point Likert response scales ranging from

"strongly disagree" to "strongly agree". A "don't know" option was available to

respondents to take into account uncertain responses that might otherwise

reduce the reliability and validity of measurement (Schuman & Presser, 1981).

These items were modified from previous research (Williams & Roggenbuck,

1989; Stedman, 1997).

____________________

Insert Table 1 about here
____________________

23

B. S. Jorgensen and R. C. Stedman

Results

Shoreline property owners in the sample were predominantly male (80%),

had gross annual household incomes in excess of $35,000 (78%), a median age

of 59 years, and were residents of either Wisconsin (59%) or Illinois (24%).

Seventeen percent of respondents were members of an environmental group,

and 30 percent were members of a lake association.

The average size of lake properties in the sample were between 1 and 5

acres. Most shoreline property owners reported having natural vegetation (89%),

trees (91%), and a dock (90%) on the lakefront. About two-thirds of all properties

in the sample tended to have a winterized house (65%), and a clear view of the

lake (69%).

Property owners nominated water quality (65%), habitat preservation

(49%), fish stocking (47%), and shoreline development (45%) as the most

important issues for their lakes. Aquatic logs (8%), property prices (15%), and

non-native plants and animals (17%) were the least reported lake issues.

After listwise deletion of missing values and "don't know" responses (n =

32), data from 282 property owners remained. A variance-covariance matrix was

computed with bootstrapping so as to obtain the best estimates given the

relatively small sample size (Joreskog & Sorbom, 1996a). The inclusion of the

asymptotic variances-covariances allowed the model parameters to be estimated

without an assumption of multivariate normality. This method enabled the

Measuring Sense of Place

24

calculation of robust chi-squares (Satorra & Bentler, 1988) and standard errors

(Yuan & Bentler, 1997).

The five measurement models were tested using LISREL 8.30 (Joreskog

& Sorbom, 1996b), and the overall fit statistics are provided in Table 2. Given

first in the table is the Satorra-Bentler scaled chi-square (SB

χ

2

) which provides a

test of the degree to which the model-based variance-covariance matrix is

consistent with the sample matrix. The Comparative Fit Index (CFI, Bentler,

1990) is not based on chi-square and indicates the extent to which the model fits

better than a baseline independence model. Next, the SRMSR is the average

standardized fitted residual and indicates the discrepancy between the sample

variance-covariance matrix and the fitted matrix. Another different type of fit

measure - the Root Mean Square Error of Approximation (RMSEA; Steiger,

1990) - takes into account the error of approximation in the population as well as

the model degrees of freedom. The last two indices given in Table 2 - the Akaike

Information Criterion (AIC; Akaike, 1987) and the Expected Cross Validation

Index (ECVI; Browne & Cudeck, 1989) - allow comparisons between non-nested

models and take into account differences in parsimony (i.e., number of

parameters) (Williams & Holahan, 1994).

Only the G+Group and CU models were associated with non-significant

chi-square statistics, indicating well-fitting models. The single-factor model in

particular showed the worst level of fit out of the five models. This model was

associated with the lowest CFI, with a value equal to 0.95 which is usually

regarded as a lower bound of good fit (Hayduk, 1996). The single-factor model

25

B. S. Jorgensen and R. C. Stedman

also had the largest standardized standardized residuals (Hu & Bentler, 1995),

and the largest RMSEA (Browne & Cudeck, 1993).

____________________

Insert Table 2 about here
____________________

The information in Table 2 indicates that the single-factor model offered

the poorest fit, while the G+Group and CU models resulted in comparably better

explanations of the observed variance. The single-factor model is empirically

nested in the 3-factor model (when the factor correlations in the latter

specification are equal to one), and a chi-square difference test can be used to

discriminate between them. Consistent with the overall fit statistics for the two

models, a single-factor explanation of the sample variance-covariance matrix

was rejected in preference for the three-factor model (

χ

2

∆

(3) = 100.83, p < .001).

The correlations between the place dimensions were 0.68 (Identity and

Dependence), 0.83 (Attachment and Dependence), and 0.72 (Identity and

Attachment).

The G+Group and CU models were the least parsimonious of the five

measurement specifications in that they have the lowest degrees-of-freedom.

Models with fewer degrees-of-freedom will result in better fit statistics, all other

things equal. However, when parsimony was taken into account via the AIC and

ECVI, the G+Group and CU models still offered better levels of fit.

The 3-factor and higher-order models achieved relatively moderate

degrees of fit, although the latter model was preferred on the basis of its lower

Measuring Sense of Place

26

AIC and ECVI. Nevertheless, both of these models offered reasonable

approximations to the data, suggesting that the domain specific components of

Identity, Attachment and Dependence were less explanatory in comparison with

the general SOP factor in this particular context.

The specific sources of variability (i.e., those independent of a general

SOP factor) were represented in the G+Group and CU models. Recall that the

latter model does not impose a structure on the specific variability, whereas each

factor is unidimensional in the G+Group model. In both models, the specific or

group factors are assumed to be orthogonal to the general SOP dimension and

to one another. The comparable performance of these two models suggests that

the assumption of unidimensional specific sources of variability is a reasonable

one in this case. That is, allowing the structure of the specific factors to go

undefined did not produce substantial differences in fit.

Comparison of the estimates of the general SOP loadings in the higher-

order, G+Group, and CU models reveals considerable consistency where the

latter two, better fitting models, were concerned. The loadings on the SOP factor

in the CU and G+Group models were virtually identical. Moreover, the loadings

for the group factors in the G+Group model reflected the pattern of error

covariances observed in the CU model. Of interest was the relatively larger

explanatory role of Identity and Dependence relative to Attachment. The latter

component accounted for little variability in observed responses compared with

the general SOP dimension. Consistent with the higher-order model, Attachment

appeared to be a redundant explanatory concept in the presence of SOP.

27

B. S. Jorgensen and R. C. Stedman

To assess the extent of variability explained by the models, eigenvalues

were calculated for each factor. Across all three models, the magnitude of the

eigenvalue for the SOP factor was reasonably consistent: 4.59 (higher-order

model), 4.75 (G+Group model), and 4.74 (CU model). There was somewhat less

consistency for the primary factor eigenvalues between the higher-order and

G+Group models. The eigenvalues for Identity, Attachment and Dependence for

the higher-order model (0.29, 0.03, 0.09, respectively) were smaller compared

with those in the G+Group model (0.77, 0.35, 0.47, respectively). Overall, the

eigenvalues for the specific factors were trivial compared with that of the SOP

dimension. Moreover, given its higher eigenvalues, the G+Group model better

explained observed variability than did the higher-order model.

Given the comparable degrees of fit associated with the G+Group and CU

models, it can be concluded that the SOP scale reflects variability among four

orthogonal dimensions. The most explanatory factor was a general evaluative

dimension (i.e., the SOP factor) summarized by owners' positive feelings toward

their lakeshore properties. Three less explanatory and largely unidimensional

factors (i.e., Identity, Attachment, and Dependence) accounted for residual

variation in responses to the scale items.

Reliability coefficients (Cronbach's alpha) were calculated for each of the

three subscales as well as the total SOP scale. These standardized coefficients

were 0.76, 0.84, 0.74, and 0.89 for Identity, Attachment, Dependence and SOP,

respectively. These statistics support the assertion that each scale reflected an

adequate degree of systematic variance.

Measuring Sense of Place

28

Summary and Discussion

Attitude theory can provide a basis for conceiving of SOP as cognitive,

affective and behavioral reactions to a spatially demarcated object. Sense of

Place was defined as a multidimensional construct comprising (1) beliefs about

the relationship between self and place; (2) feelings toward the place; and (3) the

behavioral exclusivity of the place in relation to alternatives. A 12-item SOP

scale, consistent with a multidimensional theoretical prescription, was developed

and subsequently tested in the field with a sample of lakeshore property owners

in northern Wisconsin.

A number of measurement models were posed as potential explanations

of the scale's construct validity. These models were generated on the basis of

research on attitude structure, and put forward as a means to stimulate further

thought about conceptions of SOP from a measurement perspective.

Results suggested that the scale measured a general Sense of Place

dimension that gained expression in property owners' thoughts, feelings and

behavioral commitments for their lakeshore properties. There was also clear

support for the existence of three univariate dimensions having interpretations

consistent with Place Identity, Place Attachment, and Place Dependence.

However, the general evaluative dimension better explained observed responses

than did the domain-specific constructs. The dominance of the SOP factor over

29

B. S. Jorgensen and R. C. Stedman

the narrower dimensions was prevalent in three different measurement models

that posited both general and specific factors.

The degree of covariation between the primary constructs was indicative

of shared variability with the general SOP variable that corresponded most with

feelings of Attachment and least with beliefs about Identity. That is, the concepts

of Identity and Dependence were less synonymous with the SOP factor than was

Place Attachment. This is consistent with conceptions of attitude that equate the

affective and evaluative terms (e.g., Fishbein & Ajzen, 1975).

Some caveats should be addressed with respect to the conclusions drawn

above. First, Breckler (1984) has noted that correlations among attitude

components may be inflated due to shared variability arising from common

measurement methods. However, this threat to internal validity only holds to the

extent that the measurement effects are correlated with the latent variables of

interest.

Second, the presence of measurement error that was uncorrelated with

SOP might explain the unique variability reflected in the specific place

components. However, method effects arising from the singular use of verbal

reports in this study are incompatible as an explanation for the specific sources

of variability, unless the method variability corresponded with the three

theoretical domains of Identity, Attachment, and Dependence. There was

nothing to suggest the presence of item error covariances (in the CU model),

primary factor error covariances (in the higher-order model), or group factor

correlations (in the G+Group model) that would indicate common method

Measuring Sense of Place

30

variability. Nevertheless, mixed-method approaches to measurement should be

explored in future research and are an effective means of controlling method

variance.

In conclusion, the SOP scale can provide researchers with an easily

administered and reliable means by which to measure individual attitudes toward

spatial settings. This measure can be correlated with attitudes toward policy

scenarios, group-based landscape perceptions, and characteristics of the

physical environment in order to identify consequential environmental variables

for individual wellbeing.

It is possible that the structure of the scale may not be consistent across a

range of different conditions, and correlations with other constructs may vary with

SOP components. Some attitude objects may induce disparate and conflicting

psychological responses. Future research might examine the conditions under

which multidimensional aspects of SOP are likely to be apparent. These

conditions may relate to characteristics of particular subpopulations, attitude

objects, units of analysis, and/or temporal and spatial contexts.

31

B. S. Jorgensen and R. C. Stedman

References

Akaike, H. (1987). Factor analysis and AIC. Psychometrika, 52, 317-332.

Altman, I. & Low, S. (1992). Place Attachment. NY: Plenum Press.

Bagozzi, R. P. (1978). The construct validity of the affective, behavioral, and

cognitive components of attitude by analysis of covariance structures.

Multivariate Behavioral Research, 13, 9-31.

Bagozzi, R. P. & Burnkrant, R. E. (1985). Attitude organization and the attitude-

behavior relationship: A reply to Dillon and Kumar. Journal of Personality &

Social Psychology, 49, 47-57.

Bagozzi, R. P., Tybout, A. M., Craig, C. S., & Sternthal, B. (1979). The construct

validity of the tripartite classification of attitudes. Journal of Marketing

Research, 16, 88-95.

Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological

Bulletin, 107, 238-246.

Breckler, S. J. (1984). Empirical validation of affect, behavior, and cognition as

distinct components of attitude. Journal of Personality & Social Psychology,

47, 1191-1205.

Breckler, S. J. & Wiggins, E. C. (1989). Affect versus evaluation in the structure

of attitudes. Journal of Experimental Social Psychology, 25, 253-271.

Browne, M. W. & Cudeck, R. (1989). Single sample cross-validation indices for

covariance structures. Multivariate Behavioral Research, 24, 445-455.

Browne, M. W. & Cudeck, R. (1993). Alternative ways of assessing model fit. In

K. A. Bollen & J. S. Long (Eds.), Testing structural equation models.

Measuring Sense of Place

32

London: Sage.

Cuba, L. & Hummon, D. M. (1993). A place to call home: Identification with

dwelling, community and region. Sociological Quarterly, 34, 111-131.

Dillon, W. R. & Kumar A. (1985). Attitude organization and the attitude-behavior

relationship: A critique of Bagozzi and Burnkrant's reanalysis of Fishbein and

Ajzen. Journal of Personality & Social Psychology, 49, 33-46.

Eagly, A. H. & Chaiken, S. (1993). The Psychology of Attitudes. TX: Harcourt

Brace Jovanovich.

Fishbein, M. & Ajzen, I. (1975). Belief, Attitude, Intention and Behavior: An

Introduction to Theory and Research. London: Addison-Wesley.

Hammit, W. E. & Stewart, W. P. (1996). Sense of place: A call for construct

clarity and management. Paper presented at the Sixth International

Symposium on Society & Resource Management, May 18-23, State

College, PA.

Hayduk, L. A. (1996). LISREL: Issues, Debates, and Strategies. Baltimore, MD:

Johns Hopkins.

Heberlein, T. A. (1981). Environmental attitudes. Zeitschrift fur Umweltpolitik

(Journal of Environmental Policy), 2, 241-270.

Hu, Li Tzu & Bentler, P. M. (1995). Evaluating model fit. In R. H. Hoyle (Ed.),

Structural Equation Modeling: Concepts, Issues, and Applications. Sage,

California, pp 76-99.

Joreskog, K. G. & Sorbom, D. (1996a). PRELIS 2: User's Reference Guide. IL:

SSI.

33

B. S. Jorgensen and R. C. Stedman

Joreskog, K. G. & Sorbom, D. (1996b). LISREL 8: User's Reference Guide. IL:

SSI.

Jorgensen, B. S., Nowacek, D., Stedman, R. C., Brasier, K., & Long, D. (in

press). People in a forested Lake District. In J. Magnuson & T. Kratz (Eds.),

Lakes in the Landscape, NY: Oxford University Press.

Katz, D. & Stotland, E. (1959). A preliminary statement to a theory of attitude

structure and change. In S. Kock (Ed.), Psychology: A Study of a Science, Vol

III. NY: McGraw-Hill.

Kothandapani, V. (1971). Validation of feeling, belief, and intention to act as three

components of attitude and their contribution to prediction of contraceptive

behavior. Journal of Personality & Social Psychology, 19, 321-333.

Krupat, E. (1983). A place for place identity. Journal of Environmental

Psychology, 3, 343-344.

Lalli, M. (1992). Urban-related identity: Theory, measurement, and empirical

findings. Journal of Environmental Psychology, 12, 285-303.

Lindstrom, B. (1997). A sense of place: Housing selection on Chicago's north

shore. Sociological Quarterly, 38, 19-39.

McAndrew, F. T. (1998). The measurement of 'rootedness' and the prediction of

attachment to home-towns in college students. Journal of Environmental

Psychology, 18, 409-417.

Mulaik, S. A., & Quartetti, D. A. (1997). First order or higher order general factor?

Structural Equation Modeling, 4, 193-211.

Ostrom, T. M. (1969). The relationship between the affective, behavioral, and

Measuring Sense of Place

34

cognitive components of attitude. Journal of Experimental Social Psychology,

5, 12-30.

Proshansky, H. M. (1978). The city and self-identity. Environment & Behavior,

10, 147-169.

Proshansky, H. M., Fabian, A. K., & Kaminoff, R. (1983). Place-identity: Physical

world socialization of the self. Journal of Environmental Psychology, 3, 57-83.

Relph, E. (1976). Place and Placelessness. London: Pion.

Riley, R. B. (1992). Attachment to the ordinary landscape. In I. Altman & S. M.

Low (Eds.), Place Attachment. NY: Plenum Press, pp 13-35.

Ryden, K. C. (1993). Mapping the Invisible Landscape: Folklore, Writing, and the

Sense of Place. IA: University of Iowa Press.

Rosenberg, M. J. & Hovland, C. I. (1960). Cognitive, affective, and behavioral

components of attitudes. In M. J. Rosenberg et al. (Eds.), Attitude

Organization and Change. CT: Yale.

Satorra, A. & Bentler, P. M. (1988). Scaling corrections for chi-square in

covariance structure analysis. Proceedings of the Business and Economic

Statistics Section of the American Statistics Association, pp 308-313.

Schuman, H. & Presser, S. (1981). Questions and Answers in Attitude Surveys.

NY: Academic Press.

Shamai, S. (1991). Sense of place: An empirical measurement. Geoforum, 22,

347-358.

Smith, M. Brewster (1947). The personal setting of public opinions: A study of

attitudes toward Russia. Public Opinion Quarterly, 11, 507-523.

35

B. S. Jorgensen and R. C. Stedman

Stedman, R.C. (1997). Where the north begins: Cognitive maps, sense of place

and place attachment. Paper presented at the Annual Meetings of the Rural

Sociological Society. Toronto, Ontario, August 13-17.

Steiger, J. H. (1990). Structural model evaluation and modification: An interval

estimation approach. Multivariate Behavioral Research, 25, 173-180.

Stokols, D. & Shumaker, S. A. (1981). People in places: A transactional view of

settings. In J. Harvey (Ed.), Cognition, Social Behavior, and the Environment.

NJ: Erlbaum.

Tuan, Y. F. (1979). Space and place: Humanistic perspective. In S. Gale & G.

Olson (Eds.), Philosophy in Geography. Dordrecht: D. Reidel.

Tuan, Y. F. (1980). Rootedness versus sense of place. Landscape, 24, 3-8.

Widaman, K. F. (1985). Hierarchically nested covariance structure models for

multitrait-multimethod data. Applied Psychological Measurement, 9, 1-26.

Williams, D. R., Patterson, M. E., Roggenbuck, J. W., & Watson, A. E. (1992).

Beyond the commodity metaphor: Examining emotional and symbolic

attachment to place. Leisure Sciences, 14, 29-46.

Williams, D. R. & Roggenbuck, J. W. (1989). Measuring place attachment: Some

preliminary results. Paper presented at the NRPA Symposium on Leisure

Research, San Antonio, Texas, October 20-22.

Williams, L. J. & Holahan, P. J. (1994). Parsimony-based fit indices for multiple-

indicator models: Do they work? Structural Equation Modeling, 1, 161-189.

Yuan, K. H. & Bentler, P. M. (1997). Improving parameter tests in covariance

structure analysis. Computational Statistics & Data Analysis, 26, 177-198.

Measuring Sense of Place

36

Place

Identity

IDENTITY 1

IDENTITY 2

IDENTITY 3

IDENTITY 4

ATTACHMENT 4

Place

Attachment

ATTACHMENT 1

ATTACHMENT 2

ATTACHMENT 3

Place

Dependence

DEPENDENCE 1

DEPENDENCE 2

DEPENDENCE 3

DEPENDENCE 4

FIGURE 1. Tripartite (Three-Factor) Model

37

B. S. Jorgensen and R. C. Stedman

IDENTITY 1

IDENTITY 2

IDENTITY 3

IDENTITY 4

ATTACHMENT 4

ATTACHMENT 1

ATTACHMENT 2

ATTACHMENT 3

DEPENDENCE 1

DEPENDENCE 2

DEPENDENCE 3

DEPENDENCE 4

Sense

of Place

FIGURE 2: Unidimensional (One-Factor) Model

Measuring Sense of Place

38

Place

Identity

IDENTITY 1

IDENTITY 2

IDENTITY 3

IDENTITY 4

ATTACHMENT 4

Place

Attachment

ATTACHMENT 1

ATTACHMENT 2

ATTACHMENT 3

Place

Dependence

DEPENDENCE 1

DEPENDENCE 2

DEPENDENCE 3

DEPENDENCE 4

Sense of

Place

FIGURE 3: Higher-Order Model

39

B. S. Jorgensen and R. C. Stedman

Place

Identity

IDENTITY 1

IDENTITY 2

IDENTITY 3

IDENTITY 4

ATTACHMENT 4

Place

Attachment

ATTACHMENT 1

ATTACHMENT 2

ATTACHMENT 3

Place

Dependence

DEPENDENCE 1

DEPENDENCE 2

DEPENDENCE 3

DEPENDENCE 4

Sense

of Place

FIGURE 4: G+Group Factor Model

Measuring Sense of Place

40

IDENTITY 1

IDENTITY 2

IDENTITY 3

IDENTITY 4

ATTACHMENT 4

ATTACHMENT 1

ATTACHMENT 2

ATTACHMENT 3

DEPENDENCE 1

DEPENDENCE 2

DEPENDENCE 3

DEPENDENCE 4

Sense

of Place

FIGURE 5: Correlated Uniquenesses Model

41

B. S. Jorgensen and R. C. Stedman

TABLE 1

Scale items

Factor

Item Label

Item Description

Place Identity

IDENTITY1

Everything about my lake property is a
reflection of me.

IDENTITY2

My lake property says very little about who I
am.

IDENTITY3

I feel that I can really be myself at my lake
property.

IDENTITY4

My lake property reflects the type of person I
am.

Place Attachment

ATTACH1

I feel relaxed when I'm at my lake property.

ATTACH2

I feel happiest when I'm at my lake property.

ATTACH3

My lake property is my favorite place to be.

ATTACH4

I really miss my lake property when I'm away
from it for too long.

Place Dependence

DEPEND1

My lake property is the best place for doing
the things that I enjoy most.

DEPEND2

For doing the things that I enjoy most, no
other place can compare to my lake property.

DEPEND3

My lake property is not a good place to do
the things I most like to do.

DEPEND4

As far as I am concerned, there are better
places to be than at my lake property.

Measuring Sense of Place

42

TABLE 2

Model goodness of fits statistics

Model

Overall Fit Statistics

SB

χ

2

(df)

CFI

a

SRMSR

RMSEA

(90% CI)

ECVI

(90% CI)

AIC

1-Factor

181.52 (54)***

0.95

0.07

0.09

(0.08;0.11)

0.82

(0.68;0.98)

229.52

3-Factor

80.69 (51)**

0.99

0.05

0.05

(0.02;0.06)

0.48

(0.41;0.58)

134.69

Higher-
order

79.99 (52)**

0.99

0.05

0.04

(0.02;0.06)

0.47

(0.40;0.47)

131.99

G+Group

57.33 (42)

0.99

0.04

0.04

(0.00;0.06)

0.46

(0.41;0.55)

129.33

CU

44.65 (36)

1.00

0.03

0.03

(0.00;0.05)

0.46

(0.43;0.53)

128.65

*p < .05 **p < .01 ***p < .001

a

CFI are based on the SB

χ

2

values for the target and independence models.

43

B. S. Jorgensen and R. C. Stedman

TABLE 3

Standardized parameter estimates for the Correlated Uniqueness model

SOP Loadings

Error Variances-Covariances

Item

1

2

3

4

IDENTITY1

0.57***

0.68***

IDENTITY2

-0.40***

-0.12*

0.84***

IDENTITY3

0.52***

-0.02

-0.14**

0.73***

IDENTITY4

0.61***

0.18**

-0.24***

0.19*

0.63***

ATTACH1

0.51***

0.74***

ATTACH2

0.83***

0.00

0.31***

ATTACH3

0.83***

-0.05

0.16**

0.31***

ATTACH4

0.70***

0.01

0.07

0.09

0.51***

DEPEND1

0.71***

0.50***

DEPEND2

0.67***

0.14*

0.55***

DEPEND3

-0.38***

-0.13*

-0.13*

0.86***

DEPEND4

-0.62***

-0.01

0.02

0.17**

0.62***

*p < .05 **p < .01 ***p < .001

Measuring Sense of Place

44

TABLE 4

Comparison of general SOP factor from different models

Item

Higher-order Model

G+Group Model

CU Model

Primary

Factors

SOP

H-O

Group

Factors

SOP

G+G

SOP

CU

IDENTITY1

0.26

0.49

0.28

0.54

0.57

IDENTITY2

-0.23

-0.43

-0.37

-0.39

-0.40

IDENTITY3

0.25

0.46

0.28

0.52

0.52

IDENTITY4

0.34

0.64

0.69

0.61

0.61

ATTACH1

0.05

0.43

-0.14

0.53

0.51

ATTACH2

0.10

0.86

0.22

0.86

0.83

ATTACH3

0.10

0.86

0.52

0.85

0.83

ATTACH4

0.08

0.69

0.10

0.73

0.70

DEPEND1

0.17

0.69

-0.41

0.68

0.71

DEPEND2

0.16

0.66

-0.39

0.65

0.67

DEPEND3

-0.11

-0.43

0.36

-0.38

-0.38

DEPEND4

-0.14

-0.55

0.14

-0.60

-0.62