Content uploaded by Amaury Nora
Author content
All content in this area was uploaded by Amaury Nora on Jul 11, 2018
Content may be subject to copyright.
Ma r c h /ap r i l 2009 ◆ v o l 50 n o 2 177
Conceptualization and Initial Validation of the
College Student Mentoring Scale (CSMS)
Gloria Crisp
The prevalence of conceptually valid mentoring
relationships in higher education is currently un-
known due to a lack of a valid conceptualization
within the literature. This article examines the
construct validity of College Student Mentoring
Scale (CSMS) with an eye toward identifying
developmental support functions that should be
provided to students. Participants were selected
from a stratified random sample of courses offered
in the fall of 2006 at a community college
in the south-central area of the United States
(n = 351). Results of the confirmatory factor
analysis indicated the constructs were valid and a
higher-order factor analysis revealed the existence
of a second-order construct, Mentoring. Goodness
of fit statistics suggested that the best fitting
model did not hold well across ethnic groups
however, the hypothesized factor structure was
invariant for men and women. Implications for
student affairs practice and future research are
discussed.
The value of mentoring is generally accepted
in the literature as well as in practice (Cohen,
1993). The literature indicates numerous
advantages to the student including improving
student performance, teaching specific skills,
developing intellectual and critical thinking
skills, improving student self-confidence,
discovering students’ latent abilities, self-
actualization, raising expectations and future
aspirations, and increasing student persistence
(Astin, 1999; Bank, Slavings & Biddle, 1990;
Campbell & Campbell, 1997; Freeman, 1999;
Girves, Zepeda, & Gwathmey, 2005; Johnson,
1989; Mangold, 2003; Pagan & Edwards-
Wilson, 2003; Roberts, 2000; Ross-Thomas
& Bryant; 1994).
In turn, mentoring in recent decades has
become a national priority, as demonstrated
by the hundreds of formalized programs and
institutional practices that include a mentoring
component implemented at the national,
state, and local level. The level and scope of
mentoring relationships and activities vary
(Girves et al., 2005). Similarly, the structure of
mentoring programs fluctuates by institution,
as, for example, some provide formalized
training to mentors, some provide guidelines
for meeting times, locations, frequency,
and others include students and/or staff as
mentors (Jacobi, 1991). Moreover, the goals
and objectives of mentoring initiatives differ,
as outcomes vary from improving student
achievement and retention, enhancing critical
thinking skills, increasing the number of
students who apply to graduate school, to
ensuring that graduates are able to secure
employment (Girves et al.; Jacobi).
Within the body of literature on mentoring,
researchers have attempted to address what
mentoring is and what its key components
are (e.g., Dunn & Moody, 1995; Kram, 1988;
Levinson, Carrow, Klein, Levinson, & McKee,
1978; Schockett & Haring-Hidore, 1985)
and whether mentoring positively impacts the
success of students (Astin, 1999; Bank et al.,
1990; Campbell & Campbell, 1997; Freeman,
1999; Johnson, 1989; Mangold, 2003; Pagan
& Edwards-Wilson, 2003; Ross-Thomas &
Bryant, 1994). Unfortunately, the utility of
existing studies centering on evaluating the
Gloria Crisp is Assistant Professor of Higher Education at The University of Texas at San Antonio.
178 Journal of College Student Development
Crisp
influence of mentoring on student outcomes is
extremely limited due to definitional, method-
ological, and theoretical flaws (Crisp & Cruz,
2009; Jacobi, 1991).
Currently, there is no widely accepted defi-
ni tion of mentoring (Dickey, 1996; Johnson,
1989; Miller, 2002; Rodriguez, 1995) as a recent
review identified more than 50 definitions of
mentoring, varying in scope and breadth (Crisp
& Cruz, 2009). Although some researchers
place firm boundaries on the definition of
mentoring, others define mentoring as any
relationship that teaches an individual or allows
him or her to grow. For instance, according to
Luna and Cullen (1995), mentoring can be
informal or formal, long- or short-term, and
planned or spontaneous. In contrast, Roberts
(2000) limited mentoring to
a formalized process whereby a more
knowl edgeable and experienced person
actuates a supportive role of overseeing and
encouraging reflection and learning within
a less experienced and knowledgeable per-
son, so as to facilitate that person’s career
and personal development. (p. 162)
Additionally, the definitional concept of
mentoring appears to be relative to the area
and population being studied. In the field
of psychology, Levinson et al. (1978) have
found mentoring is comprised of several
functions including teacher, sponsor, host or
guide, role model, counselor, and supporter.
In contrast, mentoring in a business context
has been found empirically to be comprised
of two broad mentoring functions, career
(aspects that facilitate career enhancement) and
psychosocial (aspects that promote a sense of
competency and identity; Kram, 1988; Noe,
1988; Schockett & Haring-Hidore, 1985).
The lack of consensus across disciplines
regarding the definition and conceptualization
of mentoring is exacerbated by weaknesses in
the research design of most published research
in the area. Existing research typically contains
low levels of external and internal validity
(Crisp & Cruz, 2009; Jacobi, 1991), as specific
mentoring models have not been tested in
various educational settings or with various
student populations (Rodriguez, 1995).
Moreover, most of the research on mentoring
has been focused on retrospective, correlational
designs (Jacobi, 1991). Therefore, additional
research is needed to develop quantitative
measures of the functions that mentors provide
(Noe, 1988).
Proposed Theoretical Framework
Despite the lack of a valid theoretically based
body of research on mentoring within the
context of college students, four major latent
variables comprising the mentoring concept
were recently identified by Nora and Crisp
(2007) who reviewed theoretical perspectives
in the education, psychology, and business
literature. As such, the four identified latent
variables combine theories by Cohen (1995),
Kram (1988), Schockett and Haring-Hidore
(1985), Levinson et al., (1978), Miller (2002),
and Roberts (2000). The four latent variables
include: (a) Psychological and Emotional Sup-
port, (b) Degree and Career Support, (c) Aca-
demic Subject Knowledge Support, and (d) the
Existence of a Role Model.
As previously described by Nora and Crisp
(2007), the first latent variable, Psychological
and Emotional Support, encompasses a sense
of listening, providing moral and emotional
support, identifying problems, and providing
encouragement as well as the establishment
of a supportive relationship in which there is
mutual understanding and link between the
student (i.e., mentee) and the mentor. Several
theoretical views were seen as comprising this
first latent variable, including Kram’s (1988)
viewpoint that mentoring must incorporate
feedback from the mentor regarding certain
fears and other issues on the part of the student.
Ma r c h /ap r i l 2009 ◆ v o l 50 n o 2 179
Conceptualization and Initial Validation
In additional, Schockett and Haring-Hidore
(1985) suggested that the discussion of fears
and uncertainties must be conducted in a safe
environment as perceived by the mentee and
that the emphasis of the relationship should
be on building the mentee’s self-confidence.
Further, Levinson et al. (1978) stipulated that
mentoring encompass moral support and Miller
(2002) suggested that listening, identification
of problems, and encouragement are part of
a mentoring experience. Moreover, this first
dimension of mentoring takes into account
active, empathetic listening and a genuine
understanding and acceptance of the mentee’s
feelings (Cohen, 1995), the development of a
positive regard conveyed by another (Kram),
a concept of buddying (Miller), and a strong
and supportive relationship (Roberts, 2000).
The second latent variable, Degree and
Career Support, includes an assessment of the
student’s strengths, weaknesses, abilities, and
includes assistance with setting academic/career
goals and decision-making. Six perspectives
provide the main focus of this latent variable:
(a) reviewing and exploring interests, abil-
ities, ideas, and beliefs (Cohen, 1995);
(b) stimulating critical thinking skills with
regard to envisioning the future and developing
personal and professional potential (Cohen,
1995); (c) reflecting and processing goals
and interests (Roberts, 2000); (d) discussing
specific suggestions regarding current plans and
progress in achieving personal, educational,
and career goals (Cohen, 1995); (e) respectfully
challenging explanations for specific decisions
or avoidance of decisions and actions relevant
to developing as an adult learner (Cohen,
1995); and (f ) facilitating the realization of
the student’s dream (Levinson et al., 1978).
The third latent variable, Academic
Subject Knowledge Support, is centered on the
acquisition of necessary skills and knowledge
(Kram, 1988); on educating, evaluating,
and challenging the student academically
(Schockett & Haring-Hidore, 1985); on
employing tutoring skills and focusing on
subject learning in contrast to mentoring that
focuses on life learning (Miller, 2002); and
on establishing a teaching-learning process
(Roberts, 2000). The fourth latent variable, the
Existence of a Role Model, is concentrated on
the presence of a role model in the student’s
life as well as the opportunity for the student
to learn from the mentor’s current and past
actions, as well as achievements and failures. In
this dimension, the emphasis is on the mentor
sharing or self-disclosing life experiences
and feelings to personalize and enrich the
relationship between himself or herself and
the student (Cohen, 1995; Kram).
Purpose
Although numerous initiatives have been
developed across the country, the prevalence
of conceptually valid mentoring relationships
in higher education is currently unknown
(Jacobi, 1991). Institutions as a whole, as well
as student affairs administrators and faculty
are currently implementing programs based
on their own definitions of mentoring or
literature that is not theoretically grounded.
One of the greatest weaknesses of mentoring
programs has been the lack of a conceptual
base to support the structure of the program
(Haring, 1999) because the theoretical base
of mentoring has not been properly examined
(Philip & Hendry, 2000). Although current
mentoring programs may be based on prior
research that connects a specific characteristic
or experience with student success (e.g., faculty
and student interaction), overall, mentoring
efforts lack a firm theoretical and conceptual
grounding. Many of these well-intended
efforts have only contributed to the confusion
and inconsistency as to what constitutes a
true mentoring experience. “Altruistic and
well-intended mentoring efforts devoid of
a substantive framework guiding program
180 Journal of College Student Development
Crisp
activities is nothing more than throwing money
at a problem or hoping that something might
stick from an array of actions. From a practical
view, engaging in the latest educational fad
is ineffective and inefficient; from a research
perspective, investigating the impact of such
programs on different student outcomes is
misleading” (Nora & Crisp, 2007, p. 338).
The design of interventions aimed at stu-
dent development and intentionally engag ing
academic and social environments must be based
on a mentoring perspective grounded on solid
theoretical perspectives underlying mentoring
experiences. Moreover, a theoretically informed
conceptualization is necessary to gain an
understanding of mentoring deeper than
simply the extent that it is beneficial to college
students (Schockett & Haring-Hidore, 1985).
More than ever student affairs administrators
need to determine how undergraduate students
establish relationships with those they view
as mentors. As many institutions of higher
education (e.g., community colleges) have
limited resources to implement mentoring
programs, it is imperative that student affairs
administrators and faculty be able to draw upon
theoretically sound research that informs the
practice of mentoring at their institutions.
Concomitantly, there is a need to under stand
how diverse groups of students conceptualize
mentoring so that more intentional efforts
are directed at designing individual programs
to meet the needs of all students (Johnson,
1989). Shultz, Colton, and Colton (2001)
concluded that a “one size fits all” approach to
mentoring may not be effective in achieving
the goal of the experience. Yet, little research
exists that empirically tests the influence of
differences, such as ethnicity and the type of
institution attended, on mentoring functions
(Jacobi, 1991). Rather, mentoring for both
minorities and non-minorities is thought to
be conceptually similar in design, suggesting
that the programmatic weaknesses in the
activities and interventions are not specific to
different groups of students (Haring, 1999)
at different types of institutions (Rodriquez,
1995). Currently, only a few formalized
mentoring programs designed specifically for
Hispanic and/or African American students
have been reported in the literature (Haring,
1999). Therefore, research is needed to test
the potential differences among these groups
to develop individualized mentoring programs
that can serve their unique needs.
The purpose of the current study was to
establish the internal consistency reliability
and construct validity of the 25-item College
Student Mentoring Scale (CSMS), designed to
assess the four interrelated latent variables in the
proposed theoretical framework. The following
research questions guided the study:
1. What is the internal consistency
reliability of the items in the CSMS?
2. To what extent is the hypothesized
measurement model an acceptable fit?
3. To what degree are the four proposed
latent variables interdependent of one
another?
4. Is the hypothesized factor structure
invariant for different groups of students?
METHOD
Participants
Although the majority of mentoring research to
date has been conducted at 4-year institutions,
a community college was used in the current
study due to the researcher’s interest in, and
access to, community college students. The
population of interest included students
enrolled in one or more core curriculum
courses (N = 7,668) at a community college in
the south-central area of the United States in
Fall 2006. Participants were identified from a
randomly generated stratified sample (n = 580)
of 20 core courses from the humanities,
Ma r c h /ap r i l 2009 ◆ v o l 50 n o 2 181
Conceptualization and Initial Validation
mathematics, and sciences. Of the classes
randomly selected, professors from 14 classes
granted permission for their class to be
surveyed (n = 436). The overall response rate
was 81% (n = 351).
Institutional data providing the partici-
pants’ gender, ethnicity, and age were not
available for 68 of the 351 participants due
to students not providing their student
identification number on their survey. Of the
283 remaining students, 54% were female,
closely paralleling the institutional population.
Similarly, the ethnic distribution of the sample
mirrored the student population overall, with
27% of survey respondents identifying as
Hispanic, 10% self-identifying as Asian, and
less than 1% as Native American. In contrast,
White students were slightly overrepresented
(52%), and African American students were
underrepresented (3%). The average survey
respondent was slightly younger at 21 years
of age than the average student attending the
college. Full-time students were significantly
overrepresented, as 74% of the sample was
classified as full-time.
Instrumentation
Participants completed the College Student
Mentoring Scale (CSMS), a 20-minute survey
administered at the beginning of class during
the first 4 weeks of the Fall term. Survey items
were developed for the purpose of this study
and were derived from factors previously
developed and validated by Cohen (1995),
Kram (1988), Schockett and Haring-Hidore
(1985), Levinson et al. (1978), Miller (2002),
and Roberts (2000). The first latent construct,
Psychological and Emotional Support, was
measured by eight items related to encouraging
the student to discuss problems, talking openly
about personal issues, providing emotional
support, and talking about social issues. The
second construct, Degree and Career Support,
was assessed using six items associated with
examining degree options, assisting the student
in making decisions associated with their
degree choice, and guiding an assessment of
the student’s skills and encouraging educational
opportunities.
Academic Subject Knowledge Support
was measured by five items related to assisting
in the achievement of academic aspirations,
encouraging discussion regarding problems
with coursework, and providing ongoing
support regarding coursework. The fourth
latent construct, the Existence of a Role
Model, was measured by six items involving
the student having someone whom they
admire and look up to regarding college
issues, someone who sets a good example,
and someone who shares personal examples of
difficulties they have overcome to accomplish
their academic goals. Participants were asked to
identify the degree to which, while in college,
they had someone in their life who provided
each of the mentoring experiences using a
5-point Likert-type scale with the following
anchors: 5 = strongly disagree, 4 = disagree,
3 = neither agree nor disagree, 2 = agree, and
1 = strongly agree (see Appendix).
DATA ANALYSIS
The internal consistency reliability of the
items measuring each of the four constructs
was established by calculating Cronbach
coefficient alphas. Confirmatory factor analysis
(CFA) was then used to address the remaining
research questions, centered on establishing the
construct validity of the items in the CSMS.
The decision to use CFA was based on the
identification of theoretical relationships
between the observed and latent variables
and the desire to determine the ability of a
hypothesized model, based on a pre-established
theory, to fit an observed set of data (Garson,
2006; Thompson, 2004). Missing data patterns
were examined using SPSS missing value
182 Journal of College Student Development
Crisp
analysis. Summary statistics were calculated
and missing values were imputed using
maximum likelihood estimation (MLE)
implemented by expectation maximization
(EM) algorithms. The fitting function used
in the analysis was maximum likelihood (ML)
which has been shown to yield consistent and
accurate estimates of population parameters
when the sample size is adequate and the
model is correctly specified (Thompson).
Model Identication
Model 1. Serving as the baseline model, Model
1 restricted factor loadings, forcing all items to
load onto the single latent variable, mentoring.
Factor loadings and errors were freely estimated
(Marsh & Hocevar, 1985). A goodness of fit
evaluation applied, as the baseline model was
overidentified, df = 276 (df = 325 – 49).
Model 2. An orthogonal four-factor
solution was hypothesized in Model 2.
Specifically, Model 2 proposed that Items 1 to
8 are congeneric and would load on Factor 1
(Psychological and Emotional Support), that
Items 9 through 14 would load on Factor
2 (Degree and Career Support), that Items
15 to 19 would load exclusively on Factor
3 (Academic Subject Knowledge Support),
and that Items 20 through 25 would load
on Factor 4 (Existence of a Role Model). As
such, Model 2 restricted factor loadings so
that each item was forced to load on only the
latent variable it was hypothesized to represent.
Moreover, errors were freely estimated and
factors were not allowed to correlate. Model 2
was overidentified, df = 279 (df = 325 – 46).
Model 3. Next, the fit of two alternative
models was tested in order to test the validity
of Model 2 (Thompson, 2004). Model 3,
similar to Model 2, required a four-factor
solution, restricting items to load onto the
respective hypothesized latent variable and
freely estimating errors. However, Model
3 differs in that it tested the potential
interrelationship between the four latent
variables by allowing each of the factors to
correlate with one another. Model 3 was also
overidentified, df = 273 (df = 325 – 52).
Model 4. Based on theoretical perspectives
of Kram (1988) and Schockett, Yoshimura,
Beyard-Tyler, and Haring (1983)—who found
mentoring to be comprised of two major
functions, career and psychosocial—Model 4
tested the validity of an orthogonal two-factor
solution. More specifically, Model 4 proposed
that Items 9 through 19 were congeneric and
loaded onto Factor 1, which was thought
to broadly measure career and academic
support. Similarly, Items 1 to 8 and 20 to 25
were forced to load exclusively on Factor 2,
hypothesized to measure psychological and
emotional support as well as aspects of social
support from the mentor. Errors were freely
estimated and factors were not allowed to
correlate. A goodness of fit evaluation applied
as the model was overidentified, df = 277
(df = 325 – 48). The identified models were
tested using AMOS 4.0.
Model Evaluation
The acceptability of the fitted CFA solution
was assessed on its overall goodness of fit,
amount of model misspecification, size,
statistical significance, and interpretability of
the parameter estimates of the model (Brown,
2006). Model fit was evaluated by examination
of the chi-square (c2), chi-square to degrees of
freedom ratio (c2/df), adjusted goodness of
fit index (AGFI), root-mean-square error of
approximation (RMSEA), root mean square
residual (RMR), normed fit index (NFI),
comparative fit index (CFI), and the Tucker-
Lewis Fit Index (TLI). The acceptable model
fit was defined by the following goodness of
fit criteria: c2 not significant (p > .05), c2/df
( < 2.0), AGFI ( > = .90), RMSEA ( < .06),
RMR ( < .08), NFI ( > = .90), CFI ( > = .95),
TLI ( > = .95) (Garson, 2006; Mulaik et al.,
Ma r c h /ap r i l 2009 ◆ v o l 50 n o 2 183
Conceptualization and Initial Validation
1989; Pedhazur, 1982; Thompson, 2004).
The amount of model misspecification
was assessed from an evaluation of the resid-
ual matrix. Standardized residuals with values
greater than 2.58 (i.e., p < .001) were considered
outside of the appropriate range (Brown,
2006; Byrne, 2001). The extent to which
the hypothesized model was appropriately
described was assessed through the examination
of modification indices. Parameters erroneously
freed with t-test statistics less than plus or
minus 2.0 and modification indices that
exceeded 15 were considered candidates to be
fixed (Thompson, 2004). Finally, assuming
the goodness of fit indices indicated a good
model fit, verified by an absence of large
residuals and modification indices, the model
evaluation concluded with an examination of
the parameter estimates. Parameter estimates
were evaluated for substan tive and statistical
sense. The out-of-range values were defined
as standardized factor correlations that exceed
1.0, negative indicator error variances, or
negative factor variances, identified as the
model specification error (Brown).
Higher Order Factor Analysis
Next, the higher order structure of the alter-
native hypothesized measurement model was
examined using hierarchical factor analysis.
Specifically, higher order CFA was used to test
whether a second-order factor “Mentoring”
had a direct effect on the four proposed latent
variables (i.e., first-order factors; Brown, 2006).
The higher order factor model restricted the
first-order factors to load onto the second-
order factor, allowing residual variances (i.e.,
disturbances) to be freely estimated. The higher
order portion of the model was overidentified
and therefore was subject to a goodness of
fit evaluation. The higher order model was
fit to the data, comparing the model fit to
that of the revised version of Model 3. The
target coefficient (T) was then calculated to
determine the degree to which the correlation
among the first-order factors was explained by
the second-order factor. The target coefficient
was found by computing the ratio of the chi-
square obtained in Model 3 compared with the
chi-square found for the higher order solution
(Marsh & Hocevar, 1985).
Testing Invariance Among Ethnic
Groups and Gender
Determination of whether the survey items
perform the same for different groups of students
was assessed by conducting simultaneous
factor analyses for each group (e.g., ethnic
group and gender). A series of CFAs were
conducted, examining similarity among
groups by imposing increasingly stringent
criteria. The first CFA examined whether the
same items loaded on the different factors for
each of the groups. If a similar fit was found,
the magnitudes of the factor loadings were
restricted to be equal for all groups. Finally,
the variances of the items across groups were
examined if the factor loadings were found
to be similar for each. If a fit of the model
was found under this third condition, it was
determined that the survey items performed
similarly for each group (Streiner, 2006).
RESULTS
Data Screening
Prior to the CFA analysis, the mentoring items
were evaluated for univariate normality. Item
means ranged between 3.53 and 4.07 and the
standard deviation for the items ranged between
.86 and 1.01. Univariate normality for each item
was analyzed to determine the level of skewness
and kurtosis. Skewness values were found to be
nonsignificant, ranging between –.752 and –.163
(p > .05). Similarly the level of flatness within the
distribution, or kurtosis, for each of the items was
found to be nonsignificant, with kurtosis values
ranging between –.509 and .275 (p > .05).
184 Journal of College Student Development
Crisp
Data outliers were identified using SPSS
13.0 software through the examination of
histograms and boxplots. All data points were
found to lie within 2.5 standard deviations of
the regression line. However, nine responses
were identified as having Malanobis d-squared
distance values greater than 75 and were
therefore removed from subsequent analysis to
prevent their potential negative effect on the
correlation among variables (n = 342). Data
were found to have moderate to moderately
high intercorrelations, ranging from .33 to .77.
The Keiser-Meyer-Olkin (KMO) statistic was
.962, indicating moderate-high intercorrelation
among the items without mulitcollinearity.
Internal Consistency Coefcients
Cronbach coefficient alphas for each of the
latent variables were found to be substantial
(i.e., greater than .70). The value of coefficient
alpha for Psychological and Emotional Support
was .912, indicating the factor was highly
reliable. Substantial reliability results for
latent variables were also found for Degree
and Career Support (α = .903), Academic
Subject Knowledge Support (α = .883) and
the Existence of a Role Model (α = .845).
Model Evaluation
Models 1, 2, and 4 were not found to provide
an acceptable fit, as evidenced by goodness of
fit indices well outside the standard values,
standardized residuals outside the appropriate
range, and large modification indices (Table
1). However, Model 3, which tested the
interrelationship between the four latent
variables by allowing each of the factors to
correlate with one another, was found to
be a plausible fit. An examination of the
standardized residual covariances indicated
that all but the covariance for Mentor24
and Mentor17 (3.96) were acceptably small.
Modification indices and expected parameter
change (EPC) values were examined in an
attempt to improve the fit of the model.
Nineteen covariances with modification
indices ranging between 15.42 and 58.34 were
identified for respecification.
Model 3 was revised, allowing the identified
error terms to correlate, which resulted in an
overall improvement in fit of the model as
the RMR and the NFI statistics were within
the appropriate range: c2(249) = 639.613,
p > .001, c2/df (2.569), AGFI (.826), RMSEA
(.068), RMR (.032), NFI (.908), CFI (.941),
TLI (.929). Standardized residual covariances
were calculated for the revised model. A
good model fit was verified by an absence
of large residuals, as all residuals were found
to be acceptably small, ranging between
–1.6 and 1.5. The evaluation of the revised
Model 3 concluded with an examination of
parameter estimates. No evidence was found
of specification error within the model,
as correlations among the factors ranged
between –.252 and .965. Strong positive
correlations were found between each of the
factors (r = .882 to .965) and weak positive
relationships were found between the error
terms that were allowed to correlate (r = .141
to .405).
Subsequently, the direction of the para-
meter estimates was examined. All of the
factor and error variances were found to
be positive, providing substantive evidence
to the validity of the model. Moreover,
all variances were found to be statistically
significant (p < .001). Confidence intervals
were calculated to determine the magnitude
of the unstandardized parameter estimates.
Standard errors were small (.02 to .06),
indicating the model’s parameter estimates
accurately approximated the population. All
standardized estimates were found to range
between .63 and .86, indicating they were
substantively meaningful.
Ma r c h /ap r i l 2009 ◆ v o l 50 n o 2 185
Conceptualization and Initial Validation
TABLE 1.
Fit Statistics for Maximum Likelihood Conrmatory Factor Analysis
Absolute Fit Indices Comparative Fit Indices
Model c2c2/df AGFI RMSEA RMR NFI CFI TLI
11257.020*** 4.571 0.715 0.101 0.047 0.816 0.850 0.836
22407.344*** 8.754 0.578 0.151 0.377 0.653 0.678 0.649
31057.098*** 3.930 0.736 0.093 0.041 0.847 0.881 0.867
3 revised 639.613*** 2.569 0.826 0.068 0.032 0.908 0.941 0.929
41713.500*** 6.231 0.695 0.124 0.310 0.753 0.783 0.763
Notes. c2 = chi square; c2/df = chi square to degree of freedom ratio; AGFI = adjusted goodness of t index;
RMSEA = root mean square of approximation; RMR = root mean square residual; NFI = normed t index;
CFI = comparative t index; TLI = Tucker-Lewis index.
*p < .05. **p < .01. ***p < .001.
Higher Order Factor Analysis
In response to the high positive correlation
found to exist among the four latent variables
(r = .882 to .965), a higher order factor
analysis was employed to test whether the
covariance among the latent variables could be
explained by the existence of a general factor,
“Mentoring.” The revised version of Model 3
provided a fit comparison for the higher order
factor model. The fit of the solution was found
to be comparable to the best-fitting model:
c2(251) = 646.106, p > .001, c2/df(2.574),
AGFI (.826), RMSEA (.068), RMR (.032),
NFI (.907), CFI (.940), TLI (.929). Moreover,
the target coefficient (T) was .989, indicating
that the higher order factor model was valid.
Testing Invariance by Ethnic Groups
Prior to conducting multiple group CFAs, the
best-fitting model (i.e., revised Model 3) was
estimated individually for White, Hispanic,
and Asian respondents. Data for African
American students were found to be linearly
dependent due to an inadequate sample size
(n = 11) and therefore were excluded from
further analyses. Overall fit statistics suggested
that the hypothesized model did not hold
well across ethnic groups as the c2 value was
found to be significant (p < .001) and c2/df
value was larger than 2.0 for Asian students.
Furthermore, although the RMR value was
less than .08, all other goodness of fit indices
were outside of the appropriate range for all
groups.
In an attempt to find the source of the
lack of invariance, simultaneous CFA’s were
estimated for each ethnic group. In the first
step, analyses of equal form were conducted.
Serving as the least restricted model, factor
loadings, variances, and covariances were
not constrained. Results indicated that the
model produced an adequate fit for White,
Hispanic, and Asian students (Table 2).
Factorial invariance was tested in the next
step, restricting the magnitude of factor
loadings (i.e., regression weights) to be equal
for all groups, while different unique variances,
common factor variances, and covariances
were permitted for all groups. An omnibus
test of invariance was conducted to determine
whether restricting factor loadings produced
a significant increase in the model chi-square
value. No significant difference was found
between the first least restrictive model and the
second model that restricted factor loadings
186 Journal of College Student Development
Crisp
TABLE 2.
Tests of Measurement Invariance Among Ethnic Groups and Gender
Absolute Fit Indices Comparative Fit
Indices
c2c2/df AGFI RMSEA RMR NFI CFI TLI
Single Group Solutions
White 475.289*** 1.909 .747 .079 .039 .848 .920 .903
Hispanic 493.405*** 1.982 .581 .118 .062 .715 .829 .794
Asian 615.888*** 2.473 .346 .234 .073 .499 .605 .524
Measurement Invariance (Ethnicity)
Equal Form 1598.444*** 2.140 .630 .068 .060 .738 .836 .803
Equal Factor Loading 1646.164*** 2.032 .649 .065 .084 .730 .839 .821
Single Group Solutions
Males 435.550*** 1.749 .736 .065 .039 .832 .919 .902
Females 503.596*** 2.022 .749 .083 .040 .852 .918 .901
Measurement Invariance (Gender)
Equal Form 939.156*** 1.886 .743 .057 .039 .844 .918 .902
Equal Factor Loading 974.136*** 1.804 .755 .054 .055 .838 .920 .911
Equal Loadings and
Variances 1029.308*** 1.721 .770 .051 .070 .829 .920 .920
Notes. c2 = chi square; c2/df = chi square to degree of freedom ratio; AGFI = adjusted goodness of t index;
RMSEA = root mean square of approximation; RMR = root mean square residual; NFI = normed t index;
CFI = comparative t index; TLI = Tucker-Lewis index.
*p < .05. **p < .01. ***p < .001.
(Δdf = 63, Δc2 = 47.72, p > .05), indicating
the factor loadings were not significantly
different for each ethnic group (Table 2).
In contrast however, an inspection of the
factor loadings for certain variables indicated
substantial differences for each ethnic group.
Testing Invariance by Gender
The invariance among gender was also exam-
ined. As a preliminary assessment of invariance,
the best-fitting model (i.e., revised Model 3) was
individually estimated for both male (n = 126)
and female (n = 150) respondents. Overall fit
statistics suggested that the hypothesized model
held well across both genders (Table 2). Next,
simultaneous CFAs were estimated for both
groups, to test form and factorial invariance.
In the first step, analyses of equal form were
conducted to confirm the hypothesized lack of
form invariance among gender. Serving as the
least restricted model, all parameter estimates
were freely estimated. Results indicated that the
model held well across both genders (Table 2).
Factorial invariance was tested next,
Ma r c h /ap r i l 2009 ◆ v o l 50 n o 2 187
Conceptualization and Initial Validation
restrict ing the magnitude of factor loadings
(i.e., regression weights) to be equal for both
genders. The omnibus test of invariance
was once again conducted to determine
whether restricting factor loadings produced
a significant increase in the model chi-square
value. No significant difference was found
between the first least restrictive model and the
second model that restricted factor loadings
(Δdf = 42, Δc2 = 34.98, p > .05), indicating
the constraint of the factor loadings did not
significantly degrade the fit of the model.
Finally, the variances of the items across
groups were examined to see if the factor load-
ings were found to be similar for each. An
acceptable fit of the model was found under
this third condition. Once again, no significant
difference was found between the first least
re strictive model and the third model that
restric ted factor loadings and variances (Δdf =
100, Δc2 = 90, p > .05). Therefore, it was
determined that the survey items performed
similarly for each group (Streiner, 2006).
DISCUSSION
The following section addresses the implications
of the findings within the context of each of the
three research questions guiding the study.
Research Question 1
The items measuring each of the four constructs
were found to be reliable, indicating the parti-
cipants responded consistently across the 25
items designed to measure the four latent
constructs.
Research Question 2
Additional research is needed to confirm the
construct validity of the correlated four-factor
model (i.e., revised Model 3). Model evaluation
is one of the most difficult issues related to
CFA (Thompson, 2004), and many researchers
have questioned the adequacy of the goodness
of fit standards due to a lack of empirical
evidence for many standards (Hu & Bentler,
1999). In turn, as Brown (2006) recommended,
the fit of the model was assessed by a number
of indicators including the amount of model
misspecification, the size, statistical significance,
and interpretability of the parameter estimates
of the model, and a variety of absolute and
comparative fit indices.
Many of the indicators used to assess the
solution provided support for the construct
validity of the correlated four-factor structure.
Although none of the fit indices indicated a
perfect fit for revised Model 3, both an abso-
lute (i.e., RMR) and comparative fit statistic
(i.e., NFI) were above the recommended values
indicating a reasonable model fit (Thomp son,
2004). Moreover, support for the construct
validity of the revised Model 3 was provided
by acceptably small residuals and parameter
estimates that failed to reveal any specification
error within the model (i.e., correlations among
the factors were all within the appropriate
range). Substantive evidence to the fit of the
model was also shown by positive factor and
error variances that were statistically significant
(p < .001; Thompson). Moreover, the construct
validity of a four-factor structure was supported
by the poor fit of the one and two-factor
models (i.e., Model 1 and Model 4).
The construct validity of the correlated
four-factor model was not supported however,
by the chi-square (c2), chi-square to degrees
of freedom ratio (c2/df ), adjusted goodness
of fit index (AGFI), root-mean-square error of
approximation (RMSEA), comparative fit
index (CFI), and the Tucker-Lewis Fit Index
(TLI). Although the chi-square was found to
be significant, researchers have widely recog-
nized that the chi-square statistic is sensitive
to sample size, and therefore was likely biased
(Hu & Bentler, 1999; Thompson, 2004).
Similarly, the AGFI statistic (i.e., .826) may
have been negatively biased by the large num-
188 Journal of College Student Development
Crisp
ber of parameters in the model (Schumacker
& Lomax, 2004). Moreover, the remainder of
the fit statistics would have been considered
acceptable if conventional standards had been
used. The RMSEA value was .068, which was
extremely close to the standard of .06 and
would be considered a reasonable fit by Hu
and Bentler. Furthermore, although the CFI
and TLI statistics did not meet more restrictive
standards (i.e., > .95), both values would have
been considered acceptable if conventional
cutoff values were used (i.e., > .90) (Bentler
& Bonnet, 1980; Kline, 1998). Regardless of
the possible justification for many of the out-
of-range fit indices, I recommend that the fit
of the model undergo further validation before
the model is applied to practice.
Research Question 3
An extremely high degree of correlation was
found among each of the latent variables, seri-
ously challenging the discriminant validity of
the latent variables. As a result, the hypothesized
mentoring model (i.e., Model 2) was found to
be an extremely poor fit. At first glance, the
survey items appeared to have been measuring
a single latent construct. However, evidence
to the construct validity of a single factor was
not supported by Model 1, which forced all of
the items to load onto a single factor. A higher
order factor analysis was conducted in response
to this seemingly contradictory finding.
Results support the possibility of the existence
of a second-order construct, Mentoring. It
appears that although respondents may have
considered psychological and emotional
support, degree and career support, academic
subject knowledge support, and the existence of
a role model to be separate types of mentoring
experiences, collectively these aspects of
support were perceived as a single overarching
construct of mentoring.
A visual inspection of the raw data revealed
that 6% of the participants either responded
that they “strongly agreed” or “agreed” to
all of the mentoring items while 3% of the
respondents indicated that they “strongly
disagreed” to all of the items. In these cases,
it was not clear whether students perceived
the four types of support as a single construct,
whether they had been provided all or none of
the types of support, or whether they failed to
accurately fill out the survey. It was also not
clear how these responses may have biased
or inflated the correlation between the latent
variables.
Research Question 4
I hypothesized that the survey items would
be invariant for different groups of students.
However, overall fit statistics suggest that the
best-fitting model did not hold well across
ethnic groups. Although restricting factor
loadings did not significantly increase the chi-
square in the model, it was determined that the
chi-square change statistic was likely biased due
to the large discrepancy in sample size among
the three groups (Thompson, 2004). In other
words, the White respondents contributed
twice the amount to the equal form than
Hispanic respondents and considerably more
than Asian American students. Moreover,
factor loadings for certain variables indicated
substantial differences for each ethnic group,
providing additional evidence to the lack of
invariance among White, Hispanic, and Asian
American students.
In contrast, the hypothesized factor
structure was found to be invariant for men
and women suggesting that both genders
perceived mentoring in the same way. However,
additional research using a larger sample size
and data from different student populations is
needed to substantiate this finding.
IMPLICATIONS FOR RESEARCH
Mentoring often occurs gradually and ran-
domly, making it hard to understand (Boice,
Ma r c h /ap r i l 2009 ◆ v o l 50 n o 2 189
Conceptualization and Initial Validation
1992). In response to the glaring absence of
a valid definition of mentoring within the
higher education literature (Dickey, 1996;
Johnson, 1989; Miller, 2002; Rodriguez,
1995), a definition, which centers on broad
forms of assistance provided to students rather
than programmatic issues, developed from
the findings of the current study is offered.
Findings suggest that mentoring, within the
context of undergraduate college students,
may be defined as: Support provided to
college students that entails emotional and
psychological guidance and support, help
succeeding in academic coursework, assistance
examining and selecting degree and career
options, and the presence of a role model by
which the student can learn from and copy
their behaviors relative to college going.
Prior to this study, the theoretical base of
mentoring had not been properly examined.
In addition, the existing mentoring research
has been limited to studies investigated within
the context of students attending 4-year
institutions. Results of the current study
contribute to educational research by providing
preliminary evidence that mentoring, within
the context of community college students,
consists of four interrelated latent variables,
which encompass an overarching concept
of mentoring. Moreover, the study provides
evidence of the construct validity of the
proposed four latent variables.
Overall, results provide evidence to
the validity of the four latent constructs
identified within the psychological, business,
and education literature (i.e., Cohen, 1995;
Kram, 1988; Schockett & Haring-Hidore,
1985; Levinson et al., 1978; Miller, 2002; and
Roberts, 2000). Furthermore, results suggest
that mentoring scales and/or interventions
that are not specifically designed for different
groups of students (i.e., minority, community
college) might be invalid, supporting the work
of Haring (1999) and Rodriquez (1995).
Similarly, findings support research by Shultz,
Colton and Colton (2001) who concluded
that a “one size fits all” approach to mentoring
may not be effective for a diverse group of
students.
Contrary to previous work however, the
poor fit of Model 4 indicates that the two-
factor model (i.e., career and psychosocial)
supported by Kram (1988) and Schockett
et al. (1983) may not be an acceptable fit for
community college students. Though it should
be noted that the questions developed and
used in the present study did not exclusively
use items previously validated by Kram and
Schockett et al. Therefore, a lack of proper
specification may have contributed to the poor
model fit.
Implications for Future Research
Although the majority of indicators supported
the validity of the four constructs, additional
research is needed to substantiate the acceptable
fit of the model. Similarly, although preliminary
evidence suggests that the proposed mentoring
scale may be reliable for one sample at a
single point in time, further evaluation of
the proposed mentoring model is suggested.
Currently, results are preliminary and limited
to a single study with a small sample size
at a community college, likely not to be
representative of others around the country.
Additional research is therefore needed to assess
the external validity of the mentoring model
on other student populations. Specifically, it is
not clear whether the items and constructs on
the CSMS will perform similarly for students
attending community colleges located in urban
or rural settings, as the institution used for the
current study is located in an affluent suburban
area. Therefore, future research should examine
the generalizability of the results using students
attending community college in urban and
rural settings.
Future research should also examine how
190 Journal of College Student Development
Crisp
different student populations conceptualize
mentoring. Potential differences among upper
level and graduate students as well as students
attending different types of 4-year institutions
should be investigated. In addition, research is
needed to understand the specific conceptual
differences that appear to exist among students
of different ethnic groups. In turn, future
research that includes an adequate sample of
African American students is needed to be able
to disaggregate this group from other minority
and nonminority students. Moreover, results of
the current study suggest that Asian American
students, a group commonly excluded from
disaggregated analyses, may perceive mentoring
in a very different way than do White or
Hispanic students. Thus, future research should
be sure to include this group of students.
Use of qualitative methods, including
focus groups and/or interviews with outlier
students, are also needed to better assess how
mentoring was perceived by the participants
(9%) who selected the same response for all of
the mentoring items. Of particular importance
is the need to understand whether, due to the
wording of the survey items, these students
perceive mentoring as a single construct or
whether they view mentoring as distinct types
of support which students did or did not
receive. Moreover, students may not receive all
aspects of mentoring within their first semester
or year of college. Therefore, future research
should include students who have had enough
time to receive mentoring (e.g., upper level
students).
Beyond a theoretical understanding of the
various components involved in mentoring, it
is also necessary to examine the roles various
individuals play in students’ mentoring
experiences. Overall, the literature has defined
and measured mentoring as a one-on-one
relationship between a student and a faculty
member or student affairs professional.
However, it is possible that other individuals,
such as upper division students, peers, friends,
and family, can and do, contribute to students’
mentoring experiences (Zalaquett & Lopez,
2006). For instance, it is reasonable that
students may best obtain subject knowledge
support from a faculty mentor and/or an
upper division student. Therefore, the idea
that various components of mentoring should
be provided by more than one individual also
deserves exploration (Wallace, Abel, & Ropers-
Huilman, 2000).
The significance of the current study will
be realized once the mentoring model is found
to be reliable and valid for specific student
populations. At that time, the theoretical
model should be incorporated within the
current persistence models and used to test
the causal relationship between mentoring and
student persistence (or other outcome variables
of interest). Various components of mentoring
may influence student outcomes in different
ways. The experience or background of mentors
may also influence the impact of mentoring.
Moreover, various mentoring experiences may
impact diverse groups of students differently.
Therefore, future researchers should explore
the differential impact of the various mentoring
components as well as the student and mentor’s
background/training on student success.
On a final note, the possibility of equiva-
lent solutions should not be ignored (Brown,
2006). Although Model 3 was found to be
an acceptable fit, this does not mean that
additional equivalent models would not also
be acceptable or even a better fit to the data.
Although current software does not have
the ability to generate equivalent model,
alternative models may exist. Therefore, future
research should not rule out this possibility.
Implications for Student Affairs
Practice
Although the major contribution of the current
study was to research and theory, the study’s
Ma r c h /ap r i l 2009 ◆ v o l 50 n o 2 191
Conceptualization and Initial Validation
potential implications for student affairs are
also worthy of note. The study contributes to
student affairs practice by offering a mentoring
scale that, once found to be an acceptable fit
for other samples, can be used in several ways
to aid in the development and evaluation of
mentoring programs. Many studies to date
have limited the definition of mentoring to
contact with faculty rather than attempting
to measure the impact of a broader mentoring
experience. In turn, the CSMS could aid
in program development by identifying the
mentoring components that students at a
particular institution are not receiving from
informal mentoring experiences in and outside
the institution. This would allow student
affairs professionals to focus their efforts
toward developing a mentoring program
around the particular needs of students at
their institution.
To that end, mentoring programs also
must be as individualized as possible to meet
the needs of a diverse student population.
Therefore, the CSMS could also aid in
identifying the particular needs of traditionally
underserved or at-risk students. For instance,
survey results may reveal that first-generation
college students have a greater need for degree
and career support, which could be used
to develop a mentoring program designed
specifically for that group of students. Finally,
an externally valid version of the CSMS
could be used by student affairs professionals
to assess programmatic efforts, comparing
the experiences and perceptions of students
before and after the mentoring intervention.
Moreover, the scale could potentially be used
to investigate the causal relationship between
programmatic efforts and various measures
of student success (e.g., retention, grades,
graduation rates).
Correspondence concerning this article should be addressed
to Gloria Crisp, Assistant Professor, Department of
Educational Leadership & Policy Studies, The University
of Texas at San Antonio, One UTSA Circle, San Antonio,
TX 78249-0654; gloria.crisp@utsa.edu
192 Journal of College Student Development
Crisp
APPENDIX.
College Student Mentoring Scale (CSMS)
While in college, I have had someone in my life who. . . .
(strongly agree = 5, agree = 4, neutral = 3, disagree = 2, strongly disagree = 1)
I look up to regarding college-related issues
helps me work toward achieving my academic aspirations
helps me realistically examine my degree or certicate options
I can talk with openly about social issues related to being in college
I admire
helps me perform to the best of my abilities in my classes
encourages me to consider educational opportunities beyond my current plans
I want to copy their behaviors as they relate to college-going
provides ongoing support about the work I do in my classes
gives me emotional support
encourages me to talk about problems I am having in my social life
sets a good example about how to relate to other people
helps me to consider the sacrices associated with my chosen degree
expresses condence in my ability to succeed academically
serves as a model for how to be successful in college
discusses the implications of my degree choice
makes me feel that I belong in college
encourages me to use him or her as a sounding board to explore what I want
shares personal examples of difculties they have had to overcome to accomplish academic goals
helps me carefully examine my degree or certicate options
I can talk with openly about personal issues related to being in college
encourages me to discuss problems I am having with my coursework
questions my assumptions by guiding me through a realistic appraisal of my skills
recognizes my academic accomplishments
provides practical suggestions for improving my academic performance
Ma r c h /ap r i l 2009 ◆ v o l 50 n o 2 193
Conceptualization and Initial Validation
REFERENCES
Astin, A. W. (1999). Student involvement: A developmental
theory for higher education. Journal of College Student
Development, 40(5), 518-529.
Bank, B. J., Slavings, R. L., & Biddle, B. (1990). Effects of
peer, faculty, and parental influence on students’ persistence.
Sociology of Education, 63, 208-225.
Bentler, P. M., & Bonnet, D. G. (1980). Significance tests
and goodness of fit in the analysis of covariance structures.
Psychological Bulletin, 88, 588-606.
Boice, R. (1992). Lessons learned about mentoring. In M.D.
Sorcinelli & A. E. Austin (Eds.), Developing New and Junior
Faculty (New Directions for Teaching and Learning No. 50,
pp. 51-71). San Francisco: Jossey-Bass.
Brown, T. (2006). Confirmatory factor analysis for applied
research. New York: Guilford Press.
Byrne, B. M. (2001). Structural equation modeling with AMOS:
Basic concepts, applications, and programming. Mahwah, NJ:
Erlbaum.
Campbell, T. A., & Campbell, D. E. (1997). Faculty mentor
programs: Effects on academic performance and retention.
Research in Higher Education, 35(6), 727-742.
Cohen, N. H. (1993). Development and validation of the
principles of adult mentoring scale for faculty mentors in
higher education. (Doctoral dissertation, Temple University,
1993). Dissertation Abstracts International, 54/02, 400. (UMI
No. AAT 9316468).
Cohen, N. H. (1995). The principles of adult mentoring
scale. New Directions for Adult and Continuing Education,
66, 15-32.
Crisp, G. & Cruz, I. (2009). Mentoring college students: A critical
review of the literature between 1990 and 2007. Scheduled to
be published in Research in Higher Education, 50(60).
Dickey, C. (1996). Mentoring women of color at the University
of Minnesota: Challenges for organizational transformation.
Minneapolis: University of Minnesota. (ERIC Document
Reproduction Service No. ED399838).
Dunn, R., & Moody, J. (1995). Mentoring in the academy:
A survey of existing programs. Washington, DC: U.S.
Department of Education. (ERIC Document Reproduction
Service No. ED396599)
Freeman, K. (1999). No services needed?: The case for
mentoring high-achieving African American students.
Peabody Journal of Education, 74(2), 15-26.
Garson, D. (2006). Structural Equation Modeling. Retrieved
May 22, 2006, from: http://www2.chass.ncsu.edu/garson/
pa765/structur.htm
Girves, J. E., Zepeda, Y., & Gwathmey, J. K. (2005). Mentoring
in a post-affirmative action world. Journal of Social Issues,
61(3), 449-479.
Haring, M. J. (1999). The case for a conceptual base for
minority mentoring programs. Peabody Journal of Education,
74(2), 5-14.
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indices
in covariance structure analysis: Conventional criteria versus
new alternatives. Structural Equation Modeling, 6, 1-55.
Jacobi, M. (1991). Mentoring and undergraduate academic
success: A literature review. Review of Educational Research,
61(4), 505-532.
Johnson, C. S. (1989). Mentoring programs. In M. L. Upcraft
& J. Gardner (Eds.), The freshman year experience: Helping
students survive and succeed in college, (pp.118-128). San
Francisco: Jossey-Bass.
Kline, R. B. (1998). The principles and practice of structural
equation modeling. New York: Guilford Press.
Kram, K. E. (1988). Mentoring at work: Developmental
relationships in organizational life. Lanham, MD: University
Press of America, Inc.
Levinson, D. J., Carrow, C. N., Klein, E. B., Levinson, M.
H., & McKee, B. (1978). The seasons of a man’s life. New
York: Ballentine.
Luna, G., & Cullen, D. L. (1995). Empowering the faculty:
Mentoring redirected and renewed. ASHE-ERIC Higher
Education Reports, 3, 1-87.
Mangold, W. D. (2003). Who goes who stays: An assessment
of the effect of a freshman mentoring and unit registration
program on college persistence. Journal of College Student
Retention, 4(2), 95-122.
Marsh, H. W., & Hocevar, D. (1985). Application of
confirmatory factor analysis to the study of self-concept: First
and higher order factor models and their invariance across
groups. Psychological Bulletin, 97(3), 562-582.
Miller, A. (2002). Mentoring students & young people: A handbook
of effective practice. London: Kogan Page.
Mulaik, S. A., James, L. R., Van Alstine, J., Bennett, N., Lind,
S., & Stillwell, C. D. (1989). Evaluation of goodness-of-fit
indices for structural equation models. Psychological Bulletin,
105(3), 430-445.
Noe, R. A. (1988). An investigation of the determinants
of successful assigned mentoring relationships. Personnel
Psychology, 41, 457-479.
Nora, A., & Crisp, G.(2007). Mentoring students: Conceptual-
izing and validating the multi-dimensions of a support
system. Journal of College Retention: Research, Theory, and
Practice, 9(3), 337-356.
Pagan, R., & Edwards-Wilson, R. (2003). A mentoring program
for remedial students. Journal of College Student Retention,
4(3), 207-225.
Pedhazur, E. J. (1982). Multiple regression in behavioral research
(2nd ed.). New York: Holt, Rinehart, & Winston.
Philip, K., & Hendry, L. B. (2000). Making sense of mentoring
of mentoring making sense? Reflections on the mentoring
process by adult mentors with young people. Journal of
Community and Applied Social Psychology, 10, 211-223.
Roberts, A. (2000). Mentoring revisited: a phenomenological read-
ing of the literature. Mentoring and Tutoring, 8(2), 145-170.
Rodriguez, Y. E. (1995). Mentoring to diversity: A multicultural
approach. In M.W. Galbraith & N.H. Cohen (Eds.),
Mentoring: New strategies and challenges (New Directions for
Adult and Continuing Education No. 66, pp. 69-77). San
Francisco: Jossey-Bass.
Ross-Thomas, E., & Bryant, C. E. (1994). Mentoring in
higher education: A descriptive case study. Education,
115(1), 70-76.
Schockett, M. R., & Haring-Hidore, M. (1985). Factor analytic
support for psychosocial and vocational mentoring functions.
Psychological Reports, 57, 627-630.
194 Journal of College Student Development
Crisp
Schockett, M. R., Yoshimura, E., Beyard-Taylor, K., &
Haring, J. J. (1983, April). Proposed model of mentoring.
Paper presented at the annual meeting of the American
Psychological Association, Anaheim, CA.
Schumacker, R. E., & Lomax, R. G. (2004). A beginner’s
guide to structural equation modeling. Mahwah, New Jersey:
Erlbaum.
Shultz, E. L., Colton, G. M., & Colton, C. (2001). The
adventor program: Advisement and mentoring for students of
color in higher education. Journal of Humanistic Counseling,
Education and Development, 40(2), 208-218.
Streiner, D. L. (2006). Building a better model: An introduction
to structural equation modeling. Canadian Journal of
Psychiatry, 51(5), 317-324.
Thompson, B. (2004). Exploratory and confirmatory factor anal-
ysis. Washington, DC: American Psychological Association.
Wallace, D., Abel, R., & Ropers-Huilman, B. R. (2000).
Clearing a path for success: Deconstructing borders through
undergraduate mentoring. The Review of Higher Education,
24(1), 87-102.
Wirt, J., Choy, S., Rooney, P., Provasnik, S., Sen, A., & Tobin,
R. (2004). The condition of education 2004, (NCES 2004-
077). U.S. Department of Education, National Center for
Education Statistics. Washington, DC: U.S. Government
Printing Office.
Zalaquett, C. P., & Lopez, A. D. (2006). Learning from the
stories of successful undergraduate Latina/Latino students:
The importance of mentoring. Mentoring & Tutoring, 14(3),
337-353.
