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.
bjorgensen@facstaff.wisc.edu
Fax: (608) 262-6022
Ph: (608) 262-6049
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