SENSE OF PLACE AS AN ATTITUDE: LAKESHORE OWNERS

Journal of Environmental Psychology (2001) 21, 233^248
# 2001 Academic Press
doi:10.1006/jevp.2001.0226, available online at http://www.idealibrary.com on
0272 - 4944/01/030233 + 16$35.00/0
SENSE OF PLACE AS AN ATTITUDE:
LAKESHORE OWNERS ATTITUDES TOWARD THEIR PROPERTIES
BRADLEY S. JORGENSEN1
AND
RICHARD C. STEDMAN2
1
Department of Psychology, University of Bath, UK
Canadian Forest Service, Edmonton, Alberta, Canada
2
Abstract
Existing attempts to measure Sense of Place (SOP) are open to a number of di¡erent interpretations, some of
which are well established in attitude research. Attitude theory can provide a basis for conceiving of SOP as
cognitive, a¡ective and conative relationships with human environments. In this study, Sense of Place was
de¢ned 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 ¢eld with a sample of lakeshore property owners in northern Wisconsin (n = 282). A number of
measurement models based on attitude structure 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.
This general evaluative dimension was more explanatory of observed responses than were the three univariate
dimensions having interpretations consistent with place identity, place attachment, and place dependence.
The dominance of the SOP factor over the narrower dimensions was prevalent in three di¡erent measurement
models that posited both general and speci¢c factors. Future research in this vein could be oriented towards
re£ecting the domains of attitude more closely, rather than being organized around the domains of sense of
place as described in the literature.
# 2001 Academic Press
Introduction
There are a plethora of concepts describing the relationship between people and spatial settings, but
Sense of Place (SOP) is perhaps the most general.
Our research uses an attitude framework to assess
whether SOP encompasses place concepts commonly
addressed in environmental psychology: Attachment, Identity, and Dependence. It is useful to consider sense of place as an attitude towards a spatial
setting especially since the constructs noted above
share strong similarities to the a¡ective, cognitive
and conative components of attitude, respectively.
An attitude approach o¡ers place research a number of bene¢ts: (1) organization of rather disorganized constructs, (2) linkage to established
literature, and (3) established research methods.
Sense of Place has been referred to as an overarching concept which subsumes other concepts describing relationships between human beings and
spatial settings (Shamai, 1991). In general, SOP is
the meaning attached to a spatial setting by a person or group. Tuan (1979) has provided the most oftcited de¢nition, declaring that a place is a center of
meaning or ¢eld 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.
For theorists such as Canter (1991) places represent a con£uence of cognitions, emotions and actions organized around human agency. In this
respect, Canter recognized that places could be conceptualized as an integrated system comprising
three attitude domains. He further ventured that
developing an understanding of the processes involved in the integration of these domains would
234
B. S. Jorgensen and R. C. Stedman
enable stronger theoretical coherence between
various threads of environmental psychology
theory (e.g., environmental cognition and environmental evaluation). Similarly, greater coherence
might be obtained across di¡erent place constructs
by considering them within a tripartite framework
comprising cognitive, a¡ective and conative processes.
Three place constructs appear in the environmental psychology literature with some regularity: place
identity, place dependence, and place attachment.
There is a considerable degree of overlap among
these concepts, but they have distinctive characteristics, also (Bonnes & Secchiaroli, 1995). Place identity involves ‘those dimensions of self that de¢ne 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 tendencies and skills relevant to this environment’
(Proshansky, 1978, p. 155). As a cognitive structure,
‘place identity’ is a substructure of a more global
self-identi¢cation in the same way that one might
consider
gender
identity
and
role-identity
(Proshansky et al., 1983).
Place attachment is described as a positive bond
that develops between groups or individuals and
their environment (Altman & Low, 1992; Williams
et al., 1992). It explicitly contains emotional content.
For example, Riley (1992) emphasized Attachment as
the ‘a¡ective relationship between people and the
landscape that goes beyond cognition, preference,
or judgement’ (p. 13). Others have suggested that attachment ‘involves an interplay of a¡ect and emotions, knowledge and beliefs, and behaviours and
actions in reference to a place’ (Altman & Low,
1992, p. 5).
Finally, place dependence is de¢ned by Stokols
and Shumaker (1981, p. 457) as an ‘occupant’s perceived strength of association between him or herself and speci¢c places.’ This strength of
association is not necessarily positive, based on Thibaut and Kelly’s (1959) comparison level/comparison
level for alternatives model. This process involves a
comparison of the current outcomes to those that
would be obtained by selecting an alternative
course of action. Each option may be negative; the
chosen option may simply be the best among poor
alternatives. Thus, place dependence concerns how
well a setting serves goal achievement given an existing range of alternatives (‘how does this setting
compare to others for what I like to do?’). Place dependence appears to di¡er from attachment in two
ways. First, it can be negative to the extent that a
place limits the achievement of valued outcomes.
Second, the ‘strength of connection’ of the social actor to the setting may be based on speci¢c behavioral goals rather than general a¡ect.
Some researchers have claimed that the relationship between place constructs is poorly understood
(Hammitt & Stewart, 1996; Kaltenborn, 1998).
Although particular relationships between place
constructs have been examined (e.g., Moore &
Graefe, 1994; Williams et al., 1992) the issue has not
been the subject of serious sustained study. Regardless, there is hardly universal agreement on the relationships between concepts. For example, in an
attempt to generate debate, Altman and Low (1992,
p. 3) asserted that ‘place attachment subsumes or is
subsumed by a variety of analogous ideas, including
topophilia (Tuan, 1974), place identity (Proshansky
et al., 1983), insideness (Rowles, 1980), genres of
place (Hu¡ord, 1992), sense of place or rootedness
(Chawla, 1992), environmental embeddedness, community sentiment and identity (Hummon, 1992).’ This
articulation suggests the need for additional study
on the interrelationships between variables. In our
research, we seek to bring clarity to these relationships by testing the hypothesis that they can be represented as speci¢c dimensions of the more
general ‘sense of place’.
The conceptual confusion noted above may in
part be due to the diversity of approaches utilized
in understanding SOP. Lalli (1992) divides SOP theory and research into phenomenological and positivistic approaches. Positivistic research on SOP is
characterized by researcher-de¢ned 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’ (p. 286).
Many place theorists (e.g., Relph, Tuan) either explicitly identify place research as a phenomenological endeavor or otherwise do not use empirical
methods to ‘test hypotheses’ in any formal sense.
Nevertheless, 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’
(Relph, 1976, p. 56). Many of these statements suggest testable hypotheses about the nature of place.
Measuring Sense of Place
However, it is possible that the phenomenological
nature of much sense of place thinking has contributed to the feeling that concepts should be treated
holistically, that dissecting a multidimensional concept is perilous business and may lead to losing the
essence of the overall concept (Hummon, 1992;
Krueger, 1996).
Approaches to measuring sense of place
Perhaps owing in part to the phenomenological emphasis, empirical investigations of SOP utilizing
quantitative methods have been relatively few in
number and have generally lagged behind theory
(Shamai, 1991; Lalli, 1992) despite calls for better
measurement (Krupat, 1983). More recently, examples, of quantitative measurement of place concepts
have been more numerous, but we suggest that
these attempts have not re£ected theoretical imperatives well, speci¢cally the multidimensionality
of the sense of place concept. In the next section of
the paper, we review several empirical studies.
These approaches vary to the extent to which they
are theoretically grounded, and range from unidimensional conceptions to those that explore potential multidimensionality.
In a relatively early example of quantitative place
research, Burdge and Ludtke (1972) employed an
‘identi¢cation with place’ scale to measure individuals’ a¡ective attachments to their residential
area. The instrument included 12 items rated on
Likert scales, and demonstrated very high reliability
(Cronbach’s a = 099). Respondents rated their level of
agreement with attitudinal statements encompassing Identity (e.g., ‘This area is in my blood, it is
really a part of me’) and Attachment (e.g., ‘I don’t
really feel any strong attachment to this place’).
Burdge and Ludtke reported that strong identi¢cation with place correlated positively with feelings of
apprehension about both relocating from the area
and having to establish new relationships in another community.
Shamai (1991) provided 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 re£ected their relationship to the place in question. The scale measured
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
235
to sacri¢ce their interests for it. In other words, respondents who were prepared to endure some form
of behavioral cost would also hold developed beliefs
and strong feelings about the place. The scale was
conceived to be unidimensional, which contrasts
with suggestions made by other theorists that emphasize the multidimensional nature of SOP.
Cuba and Hummon (1993) focused on the concept
of Place identity, which they de¢ned as an expression of ‘at homeness’. Three elements of ‘Identity’
were conceived and measured: (1) its existence (‘Do
you feel at home here?’); (2) its a⁄liations (‘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 speci¢c place loci. However,
respondents’ intensity of feelings toward a place are
not addressed.
Hay (1998) created a scale measuring the intensity of sense of place for Banks Peninsula, New
Zealand. The summed scale (Cronbach’s a = 070)
was based on reported attachment level, motivation
to remain in the setting, ancestry, and insider feelings. The resulting scale addressed only the intensity of sense of place based on these four equally
weighted variables. Hay did not conduct an analysis
of the scale’s dimensionality, but assumed that it
measured a single factor.
McAndrew (1998) sought to measure the concept
of ‘rootedness’ which he equated with an a¡ective
interpretation of place attachment. Rootedness was
assumed to have positive and negative components
that de¢ned the ends of a general bipolar concept.
The positive dimensionölabeled ‘Desire for
Change’öwas 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 was a child’. Due to McAndrew’s
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 079 and 070, 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
236
B. S. Jorgensen and R. C. Stedman
components technique in which item communalities
are set to unity. The reliability coe⁄cients also decreased to 056 (Desire for Change) and 051
(Home/Family). Further, there was a signi¢cant negative correlation between the two subscales, despite the assumption of independence manifest in
the orthogonal rotation method. McAndrew (1998)
noted 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), suggesting a single place dimension at the theoretical level despite the
multidimensionality implied in the empirical model.
More 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 ¢ve subscales of urban-related identity:
Evaluation, Familiarity, Attachment, Continuity,
and Commitment. The ¢rst subscale referred to evaluative comparisons regarding the uniqueness of the
town relative to other towns (e.g., ‘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
de¢nitions in the literature. Lalli de¢ned Attachment in terms of feelings of belonging, and measured it with items like ‘I feel really at home at
Heidelberg.’ Continuity was de¢ned 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 signi¢cance of the town in
one’s future and was measured by items that re£ected a commitment to a future relationship with
the town (e.g., ‘I would like to stay in Heidelberg
inde¢nitely’).
All ¢ve subscales showed strong correlations with
one another that ranged from 038 to 082. The highest average inter-scale correlation occurred for the
General Attachment dimension (average r = 071).
These summary statistics suggest that the ¢ve 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 con-
¢rmatory factor analysis. The scale loadings on a
general factor ranged from 032 (t = 769, p50001)
for Evaluation to 087 (t = 1540, p50001) for General
Attachment.
Williams et al. (1992) also adopted a multidimensional approach. These authors conceptualized place
attachment as an overarching concept describing an
emotional bond between an individual and a particular spatial setting. Identity and Dependence were
conceived as subdomains of attachment. Dependence was conceived in terms of how the setting
compares with other alternatives supporting similar
behavioral (recreational) goals. Identity referred to
a form of attachment resulting from the symbolic
importance of the physical environment to self-de¢nition. 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
a = 093) reported in the study.
Following the work of Williams et al. (1992),
Moore and Graefe (1994) tested a model suggesting
that place attachment to a recreational trail setting
is composed both of place dependence (how well
does the place serve instrumental values or goal
achievement) and place identity (a more a¡ect-based
statement about symbolic values of the setting). The
researchers’ principal components analysis produced two factors that were characterized by high
cross loadings and a between-scale correlation of
062, suggesting a considerable degree of overlap between constructs at the empirical level.
Towards an attitude-based conception
of sense of place
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
Measuring Sense of Place
multiple dimensions of SOP are more consistent
with the theoretical strands of SOP detailed above.
In particular, the studies by Williams et al. (1992)
and Moore and Graefe (1994) explore multidimensionality and incorporate some theoretical imperative of how the di¡erent place concepts (Identity,
Dependence, and Attachment) interrelate.
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 psychosocial structure
that organizes self-referent cognitions, emotions
and behavioral commitments. Sense of Place viewed
in this way is consistent with conceptions of attitude. Attitude theory can therefore provide a theoretical framework for organizing the relationships
between place components.
In the following section, we further explore the
multidimensional foundation of SOP with each of
the dimensions representing di¡erent dimensions
or components of attitude. Place attachment is
equated with the a¡ective (or emotional) component
of attitude; place identity as the cognitive domain
whereby a place is part of the social actor’s sense
of self; and, place dependence represents the conative domain of attitude in which the dependence expressed for one’s setting is relative to the behaviors
performed there (e.g., Moore & Graefe, 1994). The
question we ask in our research is whether the interrelationship between these components corresponds to conventional models of attitude structure.
Models of attitude structure
An attitude can be de¢ned as a response to an exogenous event, object or stimulus (Fishbein & Ajzen,
1975). Therefore, spatial settings may themselves
serve as attitude objects. Further, a¡ect, cognition
and behavior are three distinguishable components
of response to an attitude object. A¡ect refers to
emotional responses or activity in the sympathetic
nervous system, as re£ected 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 conative component in a tripartite view of attitude includes reports of behavioral intentions and
behavioral commitments, but not actual behavior.
A number of studies have been undertaken to establish how these di¡erent characteristics of attitude ¢t together. Some attitude theorists assume
that the distinction between the response categories
in which attitude is expressed represent di¡erent
237
theoretical components of attitude (see Smith, 1947;
Katz & Stotland, 1959; Ostrom, 1969; Kothandapani,
1971; Bagozzi, 1978; Bagozzi et al., 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 di¡er substantially
for certain attitude objects. For example, a person
may feel favorable toward their lakeshore property,
but consider it peripheral to their identity and a
poor place to perform certain behaviors.
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, a¡ective 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 re£ected in
the three classes of observed responses posited in
the tripartite model.
A relatively small body of literature has attempted to con¢rm the discriminant validity of
measures constructed to re£ect the di¡erent components (e.g., Breckler, 1984; Breckler & Wiggins,
1989). However, the research ¢ndings 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 di¡erences or actual di¡erences 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 covariation between
the cognitive, a¡ective, 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 de¢ned as
independent constructs that might potentially
comprise a single attitude dimension at a higher
level of abstraction.
B. S. Jorgensen and R. C. Stedman
238
The following section applies the attitude models
discussed above to SOP, and highlights the interpretative di¡erences 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, a¡ective 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
cognitions, beliefs, perceptions or thoughts that the
self is invested in a particular spatial setting. Place
attachment can be de¢ned in terms of an individual’s a¡ective or emotional connection to a spatial
setting, and place dependence can be considered as
the perceived behavioral advantage of a spatial setting relative to other settings.
FIGURE 1. Tripartite (Three-Factor) Model
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 di¡er greatly within individuals. For example, a person may strongly identify
with a setting, and be attached to it (perhaps as a
result of years of association) but feel that it is a
poor place to earn a living or raise children.
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 likely
be best expressed in the Attachment items that
more directly deal with a¡ect and express an overall evaluation of the object (Fishbein & Ajzen, 1975).
FIGURE 2. Unidimensional (One-Factor) Model
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 speci¢c) factors is due to a
Measuring Sense of Place
239
FIGURE 4. G+Group Factor Model
FIGURE 3. Higher-Order Model
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
re£ect unique beliefs, emotions, and behavioral preferences that are independent of one’s general evaluation of the setting. Thus, observed responses to
the spatial setting re£ect (1) the indirect e¡ects of
SOP through the processes of Attachment, Identity
and Dependence, and (2) the direct e¡ects of individual feelings, beliefs and behavioral commitments.
Unfortunately, higher-order models involving only
three primary constructs are empirically just-identi¢ed (Bollen, 1989). 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-¢t 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-¢t tests, it is necessary to
make additional assumptions about the model (Bollen, 1989). One assumption that follows from attitude theory is the contention that attitude is best
re£ected 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 arti¢cially imposed. One caveat arises from this
strategy when the model fails to demonstrate an
adequate approximation of the data. Speci¢cally, a
poor ¢t 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 equivalence hypothesis was
false, and another would be that the 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 speci¢c (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 this
G+Group (i.e., general and group factors) model account for variance in the observed variables, but the
general factor has a wider range of in£uence (with
respect to the response domains) than the group
factors.
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 a¡ective, 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 independence of the factors is an untestable assumption of the model that
is necessary for its empirical identi¢cation. By separating all four potential sources of variation,
240
B. S. Jorgensen and R. C. Stedman
their unique in£uences on the measured place beliefs can be gauged.
Correlated uniquenesses model
The correlated uniqueness (CU) model provides a
means of testing a ¢fth 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 e¡ects 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 ¢ts better than
the G+Group model, it can be concluded that Identity, Attachment, and Dependence do not adequately
represent the speci¢c sources of variability in£uencing responses to the scale items.
In summary, attempts to measure SOP are open to
a number of di¡erent 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.
Each of the ¢ve models of the latent structure of
the SOP scale represents di¡erent interpretations of
the construct. The three factor model suggests that
Attachment, Identity, and Dependence represent unique constructs that are not reducible to each other.
In the single factor model, however, the hypothesised multiple dimensions of SOP are absent and
only a unitary conception of the construct is posed.
FIGURE 5. Correlated Uniquenesses Model
The higher-order model attributes a multidimensional structure to SOP, but subordinates it to a
simpler interpretation (the e¡ects of a general SOP
factor) at a more fundamental level. In contrast, the
general and speci¢c 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 di¡ers from the G+Group model by
dispensing with the assumption that each group factor is unidimensional. The speci¢c place dimensions
have a more ambiguous interpretation in the CU
framework since they lie outside this measurement
model.
In both the CU and G+Group models the a¡ective, cognitive and conative categories of responses
are represented as a function of a general evaluation of a place as well as the relevant attitude component. These two models imply that some aspect of
the three response domains may be due to feelings
of attachment, beliefs about identity, and behavioral
assessments that are not synonymous with an overall evaluation of a place.
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 (see Figure 6). 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, et
al., forthcoming). The eight lakes included in the
study were Big Muskellunge, Diamond, High, Plum,
Razorback, Sparkling, Trout, and Witches.
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 ¢eld reports and preliminary trials
with a small sample of individuals from the population of households. The questionnaire comprised a
number of sections, beginning with questions about
environmental quality, environmental values and behaviours with respect to riparian forest and woody
debris destribution, recreational behaviors, perceptions of lakeshore development, attitudes toward
the lakes and properties (including the sense of
place items) and demographic characteristics.
Measuring Sense of Place
241
FIGURE 6. A shoreline property in Northern Wisconsin
TABLE 1
Scale items
Factor
Item label
Item description
Place identity
IDENTITY1
Everything about my lake property
is a re£ection of me.
My lake property says very little
about who I am.
I feel that I can really be myself at my lake property.
My lake property re£ects the type of person I am.
I feel relaxed when I’m at my lake property.
I feel happiest when I’m at my lake property.
My lake property is my favorite place to be.
I really miss my lake property when I’m away
from it for too long.
My lake property is the best place for doing the
things that I enjoy most.
For doing the things that I enjoy most, no other place
can compare to my lake property.
My lake property is not a good place to do the things
I most like to do.
As far as I am concerned, there are better places to
be than at my lake property.
IDENTITY2
Place attachment
Place dependence
IDENTITY3
IDENTITY4
ATTACH1
ATTACH 2
ATTACH3
ATTACH4
DEPEND1
DEPEND2
DEPEND3
DEPEND4
Following Heberlein and Baumgartmer (1978) reminder postcards and replacement questionnaires
were sent to householders who had not replied to
the initial mailing. The response rate after the ¢rst
mailing was 40 per cent. With further contact, the
¢nal response rate was 66 per cent after accounting
B. S. Jorgensen and R. C. Stedman
242
for undeliverable surveys, refusals to participate,
and deceased owners. Of these respondents, 71 per
cent were shoreline property owners whose responses were retained for further analysis. Property
owners who did not have shoreline frontage were excluded from the analysis because (1) it could not be
reasonably assumed that the structure of sense of
place was comparable across the two populations,
and (2) the sampling strategy was not intended to
deliver an appropriate sample of nonshoreline owners. Stedman (2000) explicitly compares sense of
place constructs between lakeshore property owners
and nonowners.
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 nonattitudes that might otherwise reduce the
reliability and validity of measurement (Schuman &
Presser, 1981). These items were modi¢ed from previous research (Williams & Roggenbuck, 1989; Stedman, 1997) and interspersed among items
concerning the lakes in the region.
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%). Nearly one-quarter of the
shoreline owners listed their property as their permanent residence. On average, the remaining 75
per cent visited their properties for 75 days out of
the year (42 days in Summer, 18 days in Autumn, 5
days in Winter and 10 days in Spring). Seventeen per
cent of respondents were members of an environmental group, and 30 per cent were members of a
lake association.
The average size of lake properties in the sample
was between 4 and 20 hectares, approximately. 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 house insulated against
the winter cold (65%), and a clear view of the lake
(69%).
Property owners nominated water quality (65%),
habitat preservation (49%), ¢sh stocking (47%), and
shoreline development (45%) as the most important
issues for their lakes. Tree logs in the water (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 shoreline
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 calculation of robust chisquares (Satorra & Bentler, 1988) and standard errors (Yuan & Bentler, 1997).
Overall, strong, positive beliefs about lakeshore
properties are prevalent (see Table 2). This sentiment
is most apparent in the relatively high levels of
mean agreement for all four Attachment items.
Where the Identity items are concerned, however,
there is less support on average for one’s property
to be implicated in beliefs about identity. These ¢ndings substantiate the use of the asymptotic variances-covariances since some item means revealed
the presence of non-normal distributions.
Examination of the correlations among scale
items failed to identify a pattern suggesting three
speci¢c attitude domains (see Appendix 1). For example, the Identity items did not show higher within components correlations than correlations with
items from the other place domains.
TABLE 2
Item means and standard deviations
Item label
IDENTITY1
IDENTITY2
IDENTITY3
IDENTITY4
Identity Subscale
ATTACH1
ATTACH2
ATTACH3
ATTACH4
Attachment Subscale
DEPEND1
DEPEND2
DEPEND3
DEPEND4
Dependence Subscale
Total Scale
Mean*
S.D.
377
221
403
390
1560
464
399
406
409
1686
411
398
181
209
1631
4877
094
100
092
086
269
054
080
083
078
238
082
096
099
100
281
678
*Negatively worded items were inversely recoded for inclusion in the summed scales. Higher numbers can be interpreted as higher levels of the construct in each case.
243
Measuring Sense of Place
multidimensional 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 speci¢cation are equal to one), and a chisquare di¡erence test can be used to discriminate
between them. Consistent with the overall ¢t statistics for the two models, a single-factor explanation
of the sample variance-covariance matrix was rejected in preference for the three-factor model
(w2D(3) = 100083, p50001). The correlations between
the place dimensions were 068 (Identity and Dependence), 083 (Attachment and Dependence), and 072
(Identity and Attachment). Given the use of a single
measurement method (i.e., Likert scales) in this
study, these correlations may be in£ated to a degree.
The G+Group and CU models were the least parsimonious of the ¢ve measurement speci¢cations in
that they have the lowest degrees-of-freedom. Models with fewer degrees-of-freedom will result in better ¢t statistics, all other things equal. However,
when parsimony was taken into account via the
AIC and ECVI, the G+Group and CU models still
o¡ered better levels of ¢t.
The 3 -factor and higher-order models achieved relatively moderate degrees of ¢t, although the latter
model was preferred on the basis of its lower AIC
and ECVI. Nevertheless, both of these models offered reasonable approximations to the data, suggesting that the domain speci¢c components of
Identity, Attachment and Dependence were less explanatory in comparison with the general SOP factor in this particular context.
The speci¢c sources of variability (i.e., those independent of a general SOP factor) were represented
in the G+Group and CU models. Recall that the
The ¢ve measurement models were tested using
LISREL 830 (Joreskog & Sorbom, 1996b), and the
overall ¢t statistics are provided in Table 3. Given
¢rst in the table is the Satorra-Bentler scaled chisquare (SBw2) 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 ¢ts better than a baseline independence model. Next, the SRMSR is the average standardized
¢tted residual and indicates the discrepancy between the sample variance^covariance matrix and
the ¢tted matrix. Another di¡erent type of ¢t measureöthe Root Mean Square Error of Approximation (RMSEA; Steiger, 1990)ötakes into account
the error of approximation in the population as well
as the model degrees of freedom. The last two indices given in Table 3öthe Akaike Information Criterion (AIC; Akaike, 1987) and the Expected Cross
Validation Index (ECVI; Browne & Cudeck, 1989)ö
allow comparisons between non-nested models and
take into account di¡erences in parsimony (i.e.,
number of parameters) (Williams & Holahan, 1994).
Only the G+Group and CU models were associated with nonsigni¢cant chi-square statistics, indicating well-¢tting models. The single-factor model
in particular showed the worst level of ¢t out of the
¢ve models. This model was associated with the lowest CFI, with a value equal to 095 which is usually
regarded as a lower bound of good ¢t (Hayduk,
1996). The single-factor model also had the largest
standardized residuals (Hu & Bentler, 1995), and
the largest RMSEA (Browne & Cudeck, 1993).
The information in Table 3 indicates that the single-factor model o¡ered the poorest ¢t, while the
TABLE 3
Model goodness of ¢ts statistics
Model
1-Factor
3 -Factor
Higher-order
G+Group
CU
Overall ¢t statistics
2
SBw (df)
CFI{
SRMSR
18152 (54)***
095
007
8069 (51)**
7999 (52)**
5733 (42)
4465 (36)
099
099
099
100
005
005
004
003
RMSEA (90% CI)
009
(008;
005
(002;
004
(002;
004
(000;
003
(000;
011)
006)
006)
006)
005)
*p5005 **p5001 ***50001
{CFI are based on the SBw2 values for the target and independence models.
ECVI (90% CI)
AIC
082
(068; 098)
048
(041; 058)
047
(040; 047)
046
(041; 055)
046
(043; 053)
22952
13469
13199
12933
12865
B. S. Jorgensen and R. C. Stedman
244
latter model does not impose a structure on the
speci¢c variability, whereas each factor is unidimensional in the G+Group model. In both models, the
speci¢c 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 speci¢c sources of variability is a reasonable
one in this case. That is, allowing the structure of
the speci¢c factors to go unde¢ned did not produce
substantial di¡erences in ¢t.
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 ¢tting models, were concerned
(see Table 4). 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 re£ected 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.
Given the comparable degrees of ¢t associated
with the G+Group and CU models, it can be concluded that the SOP scale re£ects variability among
four orthogonal dimensions in this particular research context. The most explanatory factor was a
general evaluative dimension (i.e., the SOP factor)
summarized by owners’ positive emotion 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. However,
the three domain speci¢c factors accounted for relatively trivial amounts of variability in the observed
items.
Reliability coe⁄cients (Cronbach’s a) were calculated for each of the three subscales as well as the
total SOP scale. The scales were constructed by simply summing the scores of the relevant items. The
standardised reliability coe⁄cients were 076, 084,
074, 089 for Identity, Attachment, Dependence and
SOP, respectively. These statistics support the assertion that each scale re£ected an adequate degree of
systematic variance.
Summary and discussion
Attitude theory can provide a basis for conceiving of
SOP as potentially encompassing cognitive, a¡ective
and conative reactions to a spatially based object.
Our research explored the utility of considering
Sense of Place as a multidimensional construct
comprising (1) Identity (beliefs about the relationship between self and place); (2) Attachment (emotional connection to place); and (3) Dependence, or
the degree to which the place in relation to alternative places is perceived to underpin behaviour.
A 12-item SOP scale, consistent with the multidimensional theoretical prescription described above,
was developed and subsequently tested in the ¢eld
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
TABLE 4
Comparison of general SOP factor from di¡erent models
Item
IDENTITY1
IDENTITY2
IDENTITY3
IDENTITY4
ATTACH1
ATTACH2
ATTACH3
ATTACH4
DEPEND1
DEPEND2
DEPEND3
DEPEND4
Higher-order model
G+Group model
CU model
Primary factors
SOPH7O
Group factors
SOPG+G
SOPCU
026
7023
025
034
005
010
010
008
017
016
7011
7014
049
7043
046
064
043
086
086
069
069
066
7043
7055
028
7037
028
069
7014
022
052
010
7041
7039
036
014
054
7039
052
061
053
086
085
073
068
065
7038
7060
057
7040
052
061
051
083
083
070
071
067
7038
7062
Measuring Sense of Place
thought about conceptions of SOP from a measurement perspective.
Results suggest that the scale measures a general
‘Sense of Place’ dimension that is expressed in property owners’ thoughts, emotions, and behavioral beliefs regarding their lakeshore properties. Although
there was also clear support for the existence of
three univariate dimensions consistent with place
identity, place attachment, and place dependence,
the general evaluative dimension better explained
observed responses than did the domain-speci¢c
constructs. The dominance of the SOP factor over
the narrower dimensions was prevalent in three different measurement models that posited both general and speci¢c 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 Attachment. Interestingly, this is
consistent with conceptions of attitude that equate
the a¡ective and evaluative terms (e.g., Fishbein &
Ajzen, 1975), further supporting our contention that
SOP can be addressed by researchers using attitude
terminology and theory. However, this outcome may
be due to the fact that all items measuring ‘Attachment’ were positively worded. This approach was
adopted so as to accurately re£ect views in the literature that emphasis a positive bond between humans and their environments (e.g., Altman & Low,
1992; Williams, et al., 1992). Subsequent research
might include both positively and negatively worded
items to explore the possibility of two unipolar constructs versus a bipolar model.
Some additional caveats should be addressed with
respect to the conclusions drawn above. First,
Breckler (1984) has noted that correlations among
attitude components may be in£ated 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 re£ected in the speci¢c place components. However, method e¡ects arising from the
singular use of verbal reports in this study are incompatible as an explanation for the speci¢c
sources of variability, unless the method variability
corresponded with the three theoretical domains of
Identity, Attachment, and Dependence. There was
245
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 variability. Nevertheless,
mixed-method approaches to measurement should
be explored in future research as a means of controlling method variance. Other approaches have
been expressed with respect to place that o¡er an
alternative avenue for attitude measurement and
an understanding of human-environment relationships (e.g., Canter, 1997).
The utility of our approach ultimately rests on the
validity of the concepts of Dependence, Identity, and
Attachment as attitude components. Of the concepts
illustrated, the most potentially problematic is the
concept of place-dependence as representative of
the conative domain of attitude. In this respect,
our measures are not behavioral intentions, but
rather reports of the degree to which the setting is
perceived to serve as a basis for chosen behaviors.
We suggest that future research in this vein could
be oriented towards re£ecting the domains of attitude more closely, rather than being organized
around the domains of sense of place as described
in the literature. For example, an attitude-based exploration might include (hypothetically), (1) an overall evaluation of the setting; (2) descriptive
cognitions about the setting (what are its perceived
attributes, what characteristics does it have); (3) behavioral intentions associated with the place; (4) the
quality of the attitude-object relationship characterized by the various dimensions of attitude strength;
and, (5) actual behaviors that might be reciprocally
associated with the attitude. Thus, a sense of place
becomes a true ‘sense’ of place in that it includes descriptive elements about ‘what kind of place is this’
as well as being more explicitly behavior-related.
The particular measurement structure reported
here should not be viewed as a property of the scale,
independent of the context in which the approach
was implemented. It is reasonable to expect that
the structure of the scale may not be consistent
across a range of di¡erent conditions, and correlations with other constructs may vary with SOP components. Some attitude objects may be associated
with disparate and con£icting psychological responses while others may demonstrate consistency
among response domains. The conditions under
which cognitive, a¡ective, and conative domains
might show inconsistency will most likely vary
across populations, settings, and objects, as well as
a range of individual and group level characteristics. In addition, a more heterogeneous population
B. S. Jorgensen and R. C. Stedman
246
(with respect to age and income, for example) might
demonstrate greater variation within response domains than was observed for the shoreline property
owners living on the type of lakes included in the
present sample.
Finally, the items comprising the scale may require alterations in other contexts. In this respect,
adherence to the domains of attitude may be relevant despite restrictions in the scale’s applicability
that may arise given the speci¢c content of the
items. For example, the notion that one’s property
in northern Wisconsin is ‘relaxing’ is in-keeping
with the salient discourse among owners for whom
there are ‘restorative qualities’ (Korpela & Hartig,
1996) associated with the place. Thus, while the attitude approach may be shown to have some general
applicability to place research of the type presented
here, researchers may need to contextualize the theory and methods to better re£ect the unique aspects
of their respective substantive domains.
In conclusion, this present investigation provides
an initial endeavour to relate attitude theory and
place research. The SOP scale can provide researchers with an easily administered and reliable means
by which to measure 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 psychological wellbeing.
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 a¡ective, behavioural, and cognitive components of attitude by analysis of covariance structures.
Multivariate Behavioural Research, 13, 9^31.
Bagozzi, R. P. & Burnkrant, R. E. (1985). Attitude organization and the attitude-behaviour 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 classi¢cation of attitudes. Journal of Marketing Research, 16,
88^95.
Bentler, P. M. (1990). Comparative ¢t indexes in structural
models. Psychological Bulletin, 107, 238^246.
Bollen, K. A. (1989). Structural Equations with Latent
Variables. NY: Wiley.
Breckler, S. J. (1984). Empirical validation of a¡ect, behaviour, and cognition as distinct components of attitude. Journal of Personality & Social Psychology, 47,
1191^1205.
Bonnes, M. & Secchiaroli, G. (1995). Environmental Psychology: A Psycho-Social Introduction. London: Sage.
Breckler, S. J. & Wiggins, E. C. (1989). A¡ect versus evaluation in the structure of attitudes. Journal of Experimental Social Psychology, 25, 253^271.
Browne, M. W. & Cudeck, R. (1989). Single sample crossvalidation indices for covariance structures. Multivariate Behavioural Research, 24, 445^455.
Browne, M. W. & Cudeck, R. (1993). Alternative ways of assessing model ¢t. In K. A. Bollen & J. S. Long, (Eds),
Testing Structural Equation Models. London: Sage.
Burdge, R. J. & Ludtke, R. L. (1972). Social separation
among displaced rural families: The case of £ood control reservoirs. In W. R. Burch, Jr., N. H. Cheek, Jr., &
L. Taylor, (Eds), Social Behaviour, Natural Resources,
and the Environment. NY: Harper & Row.
Canter, D. (1991). Understanding, assessing and acting in
places: Is an integrative framework possible? In T.
Garling & G. Evans, (Eds), Environmental Cognition
and Action: An Integrated Approach. NY: Oxford University Press.
Canter, D. (1997). The facets of place. In G. T. Moore & R.
W. Marans, (Eds), Advances in Environment, Behaviour
and Design, Volume 4: Towards the Integration of Theory, Methods, Research and Utilisation. NY: Plenum.
Chawla, L. (1992). Childhood place attachments. In I.
Altman & S. M. Low, (Eds), Place Attachment. NY:
Plenum Press.
Cuba, L. & Hummon, D. M. (1993). A place to call home:
Identi¢cation with dwelling, community and region.
Sociological Quarterly, 34, 111^131.
Dillon, W. R. & Kumar, A. (1985). Attitude organization
and the attitude^behaviour 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 Behaviour: 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.
Hay, R. (1998). Sense of place in developmental context.
Journal of Environmental Psychology 18, 5^29.
Hayduk, L. A. (1996). LISREL: Issues, Debates, and Strategies. Baltimore, MD: Johns Hopkins.
Heberlein, T. A. & Baumgartner, R. (1978). Factors a¡ecting response rates to mailed questionnaires: a quantitative analysis of the published literature. American
Sociological Review,, 43, 447^462.
Hu, Li Tzu & Bentler, P. M. (1995). Evaluating model ¢t. In
R. H. Hoyle, (Ed.), Structural Equation Modeling: Concepts, Issues, and Applications. Sage, California, pp.
76^99.
Hu¡ord, M. (1992). Thresholds to an alternate realm: Mapping the chaseworld in New Jersey’s Pine Barrens. In
I. Altman & S. M. Low, (Eds), Place Attachment. NY:
Plenum Press.
Hummon, D. M. (1992). Community attachment: Local sentiment and sense of place. In I. Altman & S. M. Low,
(Eds), Place Attachment. NY: Plenum Press.
Measuring Sense of Place
Joreskog, K. G. & Sorbom, D. (1996a). PRELSI 2: User’s
Reference Guide. IL: SSI.
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. (forthcoming). 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.
Korpela, K. & Hartig, T. (1996). Restorative qualities of favorite places. Journal of Environmental Psychology, 16,
221^233.
Kothandapani, V. (1971). Validation of feeling, belief, and
intention to act as three components of attitude and
their contribution to prediction of contraceptive behaviour. Journal of Personality & Social Psychology, 319,
321^33.
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 ¢ndings. 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.
Moore, R. L. & Graefe, A. R. (1994). Attachments to recreation settings: the case of rail-trail users. Leisure
Sciences, 16, 17^31.
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 a¡ective, behavioural, and cognitive components of attitude. Journal of Experimental Social Psychology, 5, 12^
30.
Proshansky, H. M. (1978). The city and self-identity. Environment & Behaviour, 10, 147^169.
Proshansky, H. M., Fabian, A. K. & Kamino¡, 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.
Rosenberg, M. J. & Hovland, C. I. (1960). Cognitive, a¡ective, and behavioural components of attitudes. In M.
J. Rosenberg et al. (Eds), Attitude Organization and
Change. CT: Yale.
247
Rowles, G. D. (1980). Growing old ‘inside’: Aging and attachment to place in an Appalachain community. In
N. Datan & A. Lahmann, (Eds), Transitions in Aging.
NY: Academic Press.
Ryden, K. C. (1993). Mapping the Invisible Landscape: Folklore, Writing, and the Sense of Place. IA: University of
lowa Press.
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.
Stedman, R. C. (2000). ‘Up North’: Measuring Sense of
Place in Wisconsin’s Lake Country. Ph.D. Dissertation, Department of Rural Sociology, University of
Wisconsin-Madison.
Stedman, R. C. (1997). Where the north begins: Cognitive
maps, sense of place and place attachment. Paper presented at the Annual Meeting of the Rural Sociological Society. Toronto, Ontario, August 13^17.
Steiger, J. H. (1990). Structural model evaluation and modi¢cation: An interval estimation approach. Multivariate Behavioural Research, 25, 173^180.
Stokols, D. & Shumaker, S. A. (1981). People in places: A
transactional view of settings. In J. Harvey (Ed.), Cognition, Social Behaviour, 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 covariances
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 ¢t
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.
B. S. Jorgensen and R. C. Stedman
248
APPENDIX
Pearson product moment correlation matrix
l1
l2
l3
l4
A1
A2
A3
A4
D1
D2
D3
D4
I1
I2
I3
I4
A1
A2
A3
A4
D1
D2
D3
D4
100
7036
029
053
028
042
042
042
051
042
7015
7035
100
7035
7049
7023
7029
7029
7035
7033
7027
015
027
100
051
030
045
047
041
037
030
7021
7031
100
029
053
052
052
041
031
7019
7040
100
042
038
037
039
033
7023
7027
100
085
065
058
058
7033
7054
100
067
057
059
7032
7054
100
046
045
7028
7042
100
062
7040
7045
100
7038
7041
100
040
100