Two Dimensions of Professor-Student Rapport Differentially Predict Student Success Outcomes: A Replication of Wilson and Ryan (2013)

Abstract

Background

Professor-student rapport is associated with various important student success outcomes, highlighting the need for a good measure of this construct.

Objective

The current study attempted to replicate the components of the Professor-Student Rapport Scale (PSRS) reported in Wilson and Ryan (2013).

Method

760 undergraduate students completed the PSRS and several student outcome measures. The sample was split in half to conduct both an exploratory factor analysis and confirmatory factor analysis. The predictive validity of the factors that emerged was then assessed.

Results

The exploratory factor analysis (EFA) did not replicate the components previously reported. Instead, we describe the emergence of two distinct factors: “Professor Cares about Students” and “Professor Creates an Engaging and Constructive Atmosphere.” The factor structure exhibited adequate model fit in the confirmatory factor analysis (CFA) and significantly predicted five of the six student and course outcomes examined.

Conclusion

We argue that the two factors reported herein better capture and elucidate professor-student rapport than the components previously identified.

Teaching Implications

There appear to be two critical pieces to establishing professor-student rapport. Students must perceive their instructor cares about their individual success/well-being, and is competent at creating an overall atmosphere conducive to engagement/learning.

Keywords

professor-student rapport student success scale development factor analysis

An important, although sometimes overlooked aspect of effective teaching is the instructor’s ability to form positive relationships with their students that are defined by rapport, or “an overall feeling between two people encompassing a mutual, trusting, and prosocial bond” (Frisby & Martin, 2010, p. 147). Students have reported that professor-student rapport is established when they perceive that a professor is accessible, approachable, fair, interesting, and elicits feelings of mutual trust, respect, and care (Benson et al., 2005; Faranda & Clarke, 2004). Evidence suggests that greater professor-student rapport is associated with a number of positive student and course outcomes (in both in-person and online formats), including perceptions/enjoyment of the course, motivation, learning, and pro-academic behaviors such as attendance, attentiveness, and studying (Benson et al., 2005; Buskist & Saville, 2004; Frisby & Martin, 2010; Frisby & Myers, 2008; Lammers & Gillaspy Jr., 2013; Lammers et al., 2017). Each of these outcomes is related to overall student success, which is an important consideration for students concerned with completing their degrees as well as institutions concerned with retention (Krumrei-Mancuso et al., 2013). Prior research also suggests that professor-student rapport predicts student ratings of instruction (Richmond et al., 2015), a measure of teacher effectiveness that impacts decisions related to hiring, promotion, and tenure at institutions of higher learning.

Given the association between professor-student rapport and impactful student, institutional, and instructor outcomes, it is important that an adequate measure of this construct exist. To this end, at least two different sets of researchers have constructed scales to measure professor-student rapport. Frisby and Myers (2008) modified an existing scale of customer–employee rapport (Gremler & Gwinner, 2000), rewording it to make an instructor the object of evaluation rather than an employee. By contrast, Wilson et al.’s (2010) Professor-Student Rapport Scale (PSRS) was constructed based on input from undergraduate students. In comparing the two scales, Ryan and Wilson (2014, p. 72) point out that their approach, unlike a direct adaptation from a different context, “values the students’ perspective.”

Because the PSRS (Wilson et al., 2010) was developed specifically for an educational context, and because it appears to be more widely used between the two (e.g., Demir et al., 2019; Richmond et al., 2015; 2016; Rogers, 2015; Wilson et al., 2014) it is the focus of the current study. More specifically, the current study is a replication and extension of Wilson and Ryan (2013), which aimed to develop a brief version of the original 34-item PSRS. We attempt to replicate its components through exploratory factor analysis (EFA), test the stability of the factor structure through confirmatory factor analysis (CFA), and test the predictive validity of the factors with regard to student and course outcomes through multiple regressions.

Development of the Professor-Student Rapport Scale

The items for the PSRS were constructed based on input from undergraduate students (Wilson et al., 2010). A principal component analysis (PCA) revealed a single component on which 34 of an original 44 constructed items significantly loaded. Convergent validity for the 34-item PSRS was established (Ryan et al., 2011; Wilson et al., 2010) through significant correlations with conceptually related constructs, including immediacy (i.e., psychological availability; Mehrabian, 1968), perceived social support (Zimet et al., 1988), working alliance (Horvath & Greenberg, 1989), and verbal aggressiveness (Infante & Wigley, 1986). Further, Wilson and colleagues demonstrated that the PSRS predicted unique variance in student outcomes (motivation, perceived learning, and self-reported grade) beyond that explained by immediacy (Wilson et al., 2010), and established adequate test-retest reliability over a 22-day interval (Ryan et al., 2011).

Wilson and Ryan (2013) next sought to reduce the size of the PSRS to make it more manageable and to eliminate any redundancies within the scale. Using data from 192 undergraduate students, they conducted a PCA which suggested two components according to Cattell’s scree test. Using a minimum loading cutoff of .50 on the rotated component matrix, the following two components were identified: a nine-item “Perceptions of Teacher” component and six-item “Student Engagement” component. In hierarchical linear regressions comparing the first and second components with the full scale, only the second component was a significant predictor of student outcomes. Ryan and Wilson (2014) designated the items of the second component (i.e., “Student Engagement”) the Professor-Student Rapport Scale – Brief (PSRS-B) and submitted it to a similar validation process to that used for the full scale.

A burgeoning body of research utilizing the scale created by Wilson and colleagues (Wilson et al., 2010) has linked professor-student rapport to multiple student outcomes: students with higher levels of rapport with their professors report higher levels of attendance (Demir et al., 2019; Wilson & Ryan, 2013), learning (Rogers, 2015; Ryan & Wilson, 2014; Schriver & Kulynych, 2021; Wilson et al., 2010), investment in the course (Rogers, 2015), and motivation (Estepp & Roberts, 2015; Wilson et al., 2010; Ryan & Wilson, 2014; Schriver & Kulynych, 2021). Also, perceived professor-student rapport predicts expected (Wilson et al., 2010) and actual grades for the course (Rogers, 2015; Wilson & Ryan, 2013). Finally, rapport was reported to be a consistent predictor of teacher effectiveness (Demir et al., 2019; Richmond et al., 2015; Rogers, 2015; Ryan & Wilson, 2014; Wilson et al., 2010).

The Current Study

The development of the PSRS and PSRS-B provides valuable tools for researchers studying the consequences of professor-student rapport as well as instructors more broadly. Our goal in the current study was to build on the foundation laid by Wilson and colleagues and to further contribute to the psychometric development of the PSRS by assessing the replicability of the components that emerged in Wilson and Ryan (2013). There were several features of the previous work developing the PSRS and PSRS-B that suggested the components of the PSRS might be an important consideration for scholars and instructors wishing to use the scale. First, the reliability of the components is arguably called into question by the fact that a single component best captured variance in the items during the scale’s initial development (Wilson et al., 2010) whereas a two-component solution emerged in later work (Wilson & Ryan, 2013). Because this ambiguity exists, a further concern is the sample size used (N = 192). Tabachnick and Fidell (2013) recommend using a sample size of at least 300 for PCA and factor analysis (FA) unless factors are well-determined, which they describe as factors with many high-loading items including marker variables with loadings greater than .80, in which case they say a sample of 100–200 subjects is acceptable. Because the “Student Engagement” component is defined by only six indicators, with the highest loadings reaching .65, it appears as though the components may not be sufficiently well-defined for a sample smaller than 300 to be appropriate.

Wilson and Ryan (2013) also raise concerns as to how well the “Student Engagement” component (i.e., the PSRS-B) captures the construct it is intended to. As they point out, the “Perceptions of Teacher” component better fits the conceptual definition of professor-student rapport, whereas “Student Engagement” may be tapping into the friendliness of a teacher the extent to which students are interested in attending class (Wilson & Ryan, 2013, p.133). Another potential problem with the “Student Engagement” component is the lack of thematic consistency across all six items. For example, one of the six items, “My professor’s body language says, ‘Don’t bother me,’” had a component loading (.50) in the original PCA (Wilson & Ryan, 2013), and does not seem to align with the theme of student engagement. Instead, it seems thematically more consistent with the “Perceptions of Teacher” component. Additional evidence suggesting that the different versions of the PSRS may not be measuring the same psychological construct comes from disparities that emerge across studies (e.g., Demir et al., 2019; Richmond et al., 2015; Rogers, 2015; Ryan & Wilson, 2014; Wilson et al., 2010; Wilson & Ryan, 2013) with respect to how well each version of the scale predicts certain student outcomes. For example, the ability of each version of the scale to predict students’ perception of the course varies widely. Reported standardized regression coefficients predicting this outcome range from β = −0.03 (Demir et al., 2019) to β = 0.21 (Wilson & Ryan, 2013) for the 9-item “Perceptions of Teacher” component, as compared to a range of β = 0.51 (Wilson & Ryan, 2013) to β = 0.75 (Demir et al., 2019) for the 6-item “Student Engagement” component and β = 0.79 (Wilson et al., 2010) for the full 34-item PSRS. Collectively, these findings suggest that even though the “Perceptions of Teacher” component may better represent the construct of professor-student rapport, the “Student Engagement” component appears to be a better predictor of the student outcomes investigated. Additional associations between student outcomes and different versions of the PSRS found in previous studies are displayed in Table 1.

Table 1.

Summary of Results (Standardized Regression Coefficients, β) Across Studies Using Some Form of the Professor-Student Rapport Scale.

Study	N	Version of PSRS^a	Perception of Course^b	Perception of Teacher	Motivation	Perceived Learning	Self-Reported Grade	Teacher Effectiveness	Classes Missed	Actual Final Grade
Wilson et al. (2010)	195	34-item	.79***	.79***	.67***	.49***	.46*	-	-	-
Wilson and Ryan (2013)	192	9-item	.21	−.11	.10	.26	−.04	.12	−.09	−.38
		6-item	.51***	.49***	.29*	.39*	.34*	.63***	−.35*	.39*
Ryan and Wilson (2014)	130	6-item	.68***	.69***	65***	.48***	.39***	-	-	-
Richmond et al. (2015)	252	45-item	-	-	-	-	-	.74***	-	-
Rogers (2015)	280	34-item	-	-	-	.43***	-	-	-	.15*
Demir et al. (2019)	412	15-item	.44***	-	-	.40***	.06	.49***	−.21*	.01
		9-item	−.03	-	-	.06	-	.35**	.02	-
		6-item	.75**	-	-	.49**	-	.52**	–.24*	-

^a9-item and 6-item (15-item = combination) scales = “Perceptions of Teacher” and “Student Engagement,” respectively (Wilson & Ryan, 2013).

^bThough outcome names varied all were included if considered effectively equivalent (e.g., SRIs and Teacher Effectiveness were considered equivalent).

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

The current study sought to determine whether the components found in Wilson and Ryan (2013) are replicable using a sample size greater than 300 and exploratory factor analysis (EFA) as opposed to PCA alone. EFA attempts to reach a parsimonious solution by explaining only common variance rather than the total variance as PCA does. Conceptually, this means in EFA the factors cause the variables whereas in PCA the variables cause the components. For this reason, EFA is preferable for theory building and inferences. PCA, on the other hand, is an empirical summary of the data and is less suited to these purposes. Further, our aim was to provide additional empirical support for the factor structure that emerged in the EFA through the use of confirmatory factor analysis (CFA). Another advantage of EFA is that it generally explores multiple solutions, seeking out a stable structure, which CFA in turn either supports or refutes.

After determining the factor structure, the predictive validity of the factors was tested through several multiple linear regressions using previously examined student and course outcomes similar to those used in the original Wilson and Ryan (2013) study and previous studies using these scales (e.g., Richmond et al., 2015; Rogers, 2015). In the case of a successful replication of the previous components, this would determine the extent to which the associations between the factors and outcomes were replicable as well. In the case of an emergence of a new factor structure, it would allow for testing of the predictive validity of the new factor structure.

Method

Participants and Procedure

Data were collected from 760 undergraduate students attending a midsized southwestern university. The survey was distributed to students in 18 different face-to-face classes taught by 10 different instructors (3 female, 7 male). Classes varied in size (range = 24–170 students), course level (100-, 200-, and 300-level courses), and subject (Biology, Criminology and Criminal Justice, English, and Psychological Sciences). Data collection occurred the week before finals, allowing students the entire semester to establish rapport with the instructor. Students were instructed to complete the survey with the instructor of the course they were currently attending in mind. To reduce bias, instructors exited the room during survey completion. All participants provided written informed consent. The consent form included a FERPA waiver, and most students consented to releasing their final grade in the course for use in the study (Yes = 760, No = 26, No mark/not used = 12). Participation was voluntary and without compensation. The study was approved by the Institutional Review Board at Northern Arizona University.

Table 2 includes demographic information on characteristics of the students (e.g., age, gender, and ethnicity/race). Table 3 provides information on course characteristics. Instructors were both faculty members (tenured/tenure-track = 10%, lecturer = 20%, instructor = 20%) and graduate teaching assistants (GTA = 50%; both master’s and Ph.D. candidates).

Table 2.

Sample Demographics on Individual Characteristics.

Total Sample Size: N = 760
Mean age (standard deviation, range)		19.4 (0.13, 18–44)
Gender		754 (99%)
	Female	534 (70%)
	Male	219 (29%)
	“N/A”	1 (<1%)
Class standing		753 (99%)
	Freshman	397 (52%)
	Sophomore	153 (20%)
	Junior	131 (17%)
	Senior	68 (9%)
	Other	4 (1%)
First-generation college student		721 (95%)
	Yes	287 (38%)
	No	434 (57%)
Ethnicity and race		760 (100%)
	American Indian or Alaskan Native	30 (4%)
	Asian	17 (2%)
	Black or African American	36 (5%)
	Hispanic or Latino	105 (14%)
	Multiracial or Other	82 (11%)
	Native Hawaiian or Pacific Islander	9 (1%)
	White	481 (63%)

Table 3.

Sample Demographics on Course Characteristics.

Course Characteristic		n (%)
Instructor
	Female	173 (23%)
	Male	587 (77%)
Department
	Biology (1 instructor, 3 classes)	60 (8%)
	Criminology/criminal justice (1 instructor, 3 classes)	110 (15%)
	English (4 instructors, 4 classes)	76 (10%)
	Psychological sciences (4 instructors, 8 classes)	514 (68%)
Course level
	100-Level (6 instructors, 10 classes)	483 (64%)
	200-Level (3 instructors, 6 classes)	233 (31%)
	300-Level (1 instructor, 2 classes)	44 (6%)
Course size
	Small (24–25 students; 6 instructors, 9 classes)	180 (24%)
	Medium (44–76 students; 3 instructors, 6 classes)	233 (31%)
	Large (100–180 students; 1 instructor, 3 classes)	347 (46%)

Note. Total N = 760.

Measures

Professor-Student Rapport Scale

The full, 34-item PSRS was used. In prior research, it has demonstrated strong internal consistency reliability (α = .96; Wilson et al., 2010). Sample items included, “My professor and I get along,” and, “My professor makes class enjoyable.” Participants were asked to indicate their agreement with each statement using a Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). Internal consistency reliability for the current study was strong and consistent with previous work using the scale (full scale: α = .97; “Perceptions of Teacher” component: α = .93; “Student Engagement” component: α = .88).

Outcome Measures

Teacher effectiveness. Consistent with past research (Demir et al., 2019), eight items from the university’s student rating of instruction (SRI) were averaged to measure teacher effectiveness (α = .94). Participants rated their agreement with statements such as “Instructor was effective in creating a climate open to student questions,” and, “Instructor motivated me to be successful in this course” on a scale from 1 (strongly disagree) to 5 (strongly agree); higher scores indicated higher ratings of teacher effectiveness.

Perception of course. Three items (α = .92) were averaged to measure students’ perception of the course. For the first item (Demir et al., 2019), participants were asked: “What is your overall perception of the course?” (1, unacceptable to 5, excellent). For the other two items (Filak & Sheldon, 2008), participants rated their agreement with the following statements: “Overall, this course was excellent,” and, “I would recommend this course to a friend” (1, strongly disagree to 5, strongly agree). Higher scores indicated more positive perceptions of the course.

Perception of teacher. Two items previously used by Demir et al. (2019) were averaged to measure students’ perception of the teacher (r = .89, p < .01): “Overall, this teacher was excellent,” and, “I would recommend this teacher to a friend” (1, strongly disagree to 5, strongly agree). Higher scores indicated more positive perceptions of the teacher.

Perceived learning. Consistent with previous research (e.g., Wilson & Ryan, 2013; Demir et al., 2019), a single item assessed perceived learning. Participants rated the amount they had learned in the course (1 = very little to 5 = a great deal). Higher scores indicated higher perceptions of learning.

Expected final grade. Participants indicated what percentage of the total points possible in the course they expected to earn in the class as a measure of expected final grade. All instructors used the standard 90–100% = A, 80–89% = B, etc. grading scale.

Actual final grade. Most students (95.24%) waived their FERPA rights and agreed to allow their professor to provide final course grades to the researchers. These were provided to one of the authors by the course instructors at the end of the semester.

Results

Exploratory Factor Analysis

Exploratory analyses were conducted in IBM SPSS Statistics Version 28 unless otherwise specified. A dataset containing 760 cases was randomly split into two datasets—one for each factor analysis—of 379 (exploratory) and 381 (confirmatory) cases. A missing values analysis indicated less than 5% missing data for all 34 PSRS items included in the EFA; thus, 13 cases with missing data were deleted listwise (n = 366). Descriptive analysis revealed considerable negative skew across all items (skew > |1| for ∼74% of items), so an average of the full scale was used to screen for outliers rather than assessing each variable individually for univariate outliers and using regression to screen for multivariate outliers¹. A z score cutoff of 3.29 (p < .001) indicated exclusion of three outlier cases (n = 363).

First, a PCA with varimax rotation was conducted. The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy indicated sufficiently strong relationships between items (KMO = .96), and Bartlett’s Test of Sphericity was significant (p < .001). Eigenvalues did not approach zero (lowest value = .14), suggesting no evidence of multicollinearity or singularity. Communalities were sufficiently high to warrant inclusion of all items in additional analyses (range = .25 – .75).

The first four components displayed eigenvalues greater than 1.00 (15.27, 2.50, 1.51, and 1.21). After extraction, these four components explained 60% of the available variance, with the first component responsible for 45% of this. After rotation, the components explained 20%, 17%, 13%, and 10% of the variance, respectively. Examination of the scree plot suggested two to four factors. Based on these evaluations, two-, three-, and four-factor solutions were explored in the EFAs.

Multiple solutions were run that varied according to number of factors, extraction technique, and rotation technique. Oblique rotation was applied across all combinations because factor correlations exceeded .32. Two types of oblique rotation were compared: direct oblimin (delta = 0) and promax. Unweighted least squares (ULS) extraction was used due to the non-normal nature of the data and its being less limited by distributional assumptions (Zygmont & Smith, 2014), and these were compared alongside solutions using generalized least squares (GLS) extraction. Overall, four combinations of extraction and rotation were explored for each possible number of factors.

A two-factor structure with ULS extraction and promax rotation provided the best solution in terms of the conceptual consistency and meaningfulness of the factors as well as with respect to the goal of creating a shorter version of the scale. Compared to the other solutions, it was the only one that displayed stability across all four combinations of extraction and rotation.² After extraction, the two factors explained 50% of the variance in the correlation matrix. A minimum loading value of 0.55 (30% of variance) was chosen for inclusion because it provided a solution with a simple structure (i.e., no substantial cross-loadings between factors). Though this is a relatively stringent cutoff, in addition to providing a simple structure, selecting a high cutoff was consistent with the goal of creating a shorter version of the scale. The high loadings of many of the items on their respective factor suggest the factors are well defined (see Table 4). Additionally, the reproduced correlations matrix indicated an adequate fit (i.e., approximately 75% of the residuals were smaller than .05). The two factors were strongly associated (r = .70).

Table 4.

Factor Loadings and Communalities (h2) for Unweighted Least Squares Extraction and Promax Rotation on PSRS Items.

Item	F₁^a	F₂	h ²
My professor is eager to help students (Q21)	.84	-^b	.70
My professor maintains eye contact with me (Q18)	.82	−.34	.39
My professor encourages me to succeed (Q23)	.81	-	.66
My professor and I communicate well (Q20)	.80	-	.57
My professor is compassionate (Q22)	.78^c	-	.66
My professor cares about students (Q27)	.73^c	-	.69
My professor and I get along (Q1)	.69	-	.44
My professor is aware of the amount of effort I am putting into this class (Q8)	.66	-	.32
My professor is approachable (Q13)	.65	-	.42
My professor is understanding (Q4)	.64	-	.43
My professor is receptive (Q31)	.62^c	-	.44
My professor wants to make a difference (Q30)	.59^c	-	.59
My professor makes class enjoyable (Q15)	-	.95^d	.70
I want to take other classes taught by my professor (Q16)	-	.95^d	.71
I dislike my professor’s class (Q14)	-	.92^d	.59
I really like to come to class (Q19)	-	.84^d	.61
My professor is not helpful (Q2)	-	.75	.45
I feel I have learned much less form this professor compared to others I have had in the past (Q24)	-	.67	.44
I understand what my professor expects of me (Q7)	-	.65	.47
My professor is reliable (Q32)	-	.61^c	.63

^aF₁ – “Professor Cares about Students,” F₂ – “Professor Creates an Engaging and Constructive Atmosphere.”

^bLoadings below |.30| not displayed.

^cItems also contained in “Perceptions of Teacher” component.

^dItems also contained in “Student Engagement” component. Items are from Wilson & Ryan (2013), p. 131.

The factor structure that emerged was not the same as the components found in previous work (Wilson & Ryan, 2013). Rather, two new factors emerged. The first factor contained 12 items consisting of behaviors and qualities indicating an instructor cares about students and their success. Because this factor describes an instructor who cares enough about students to pay attention to them, invest energy into communicating well with them, and offer them assistance when needed, we called this first factor “Professor Cares about Students.” It displayed high internal consistency reliability (α = .92), and all 12 items appeared to be conceptually consistent with the theme of caring about students. As depicted in Figure 1a, the items in this factor did overlap somewhat with those in the “Perceptions of Teacher” component found by Wilson and Ryan (2013). However, the items loading on the new factor appear to focus on qualities and behaviors that are arguably more relevant to interpersonal relationships than the items of the “Perceptions of Teacher” component. Specifically, the items of the old component that did not overlap with the new factor appear to focus more on perceptions of competence that are not necessarily related to positive interpersonal interactions.

Figure 1.

(a) Venn diagram depicting shared and unshared items between the “Perceptions of Teacher” component and “Professor Cares about Students” factor. (c) Venn diagram depicting shared and unshared items between the “Student Engagement” component and “Professor Creates an Engaging and Constructive Atmosphere” factor.

The second factor was very similar to Wilson and Ryan’s (2013) second component, “Student Engagement.” As seen in Figure 1b, both have items in common that describe a student who enjoys and wants to attend class. However, the second factor that emerged in the current study has additional items that move beyond whether or not students feel engaged. This suggests that it is not only engagement that is important for rapport, but also a constructive atmosphere in which the student feels the course has a stable, clear, and supportive structure conducive to actual learning. Therefore, we called this 8-item second factor, “Professor Creates an Engaging and Constructive Atmosphere.” It also displayed high internal consistency reliability (α = .91), and all items appeared to be conceptually consistent.

Confirmatory Factor Analysis

The CFA was conducted within the other half of the split data set (381 cases) to determine the stability of the new factor structure that emerged in the EFA. A missing values analysis indicated less than 5% missing data for all variables; thus, 4 cases with missing data were deleted listwise (n = 377). Descriptive analysis revealed non-normal data across all variables, each exhibiting significant negative skew. In order to be consistent, outliers were screened for and eliminated according to the same procedures used for the EFA. Based on average scores of each factor, four cases were identified as outliers and excluded (n = 373).

The CFA was conducted in MPlus Version 7.1. Maximum likelihood parameter estimates with standard errors and a mean-adjusted chi-square statistic that are robust to non-normality (MLM) were used due to the non-normal nature of the data (Byrne, 2012). The highest loading items on each respective factor from the EFA (i.e., Q21 for the first factor and Q15 for the second factor, Table 4) were fixed to one for the CFA for purposes of identification. The chi-Square Test of Model Fit was significant (χ²(169) = 393.86, p < .001, scaling correction factor = 1.37), suggesting poor model fit. However, this fit statistic is very sensitive to sample size, and thus “decisions regarding adequacy of model fit are typically based on alternate indices of fit” (Byrne, 2012, p. 69). In line with recommendations made by Jackson et al., (2009), the determination for adequacy of model fit was based on the Root Mean Square Error of Approximation (RMSEA), Comparative Fit Index (CFI), and Tucker–Lewis Index (TLI). The value for the CFI was 0.938³ while the TLI was 0.930, both of which suggested acceptable model fit (Hu & Bentler, 1999). The value for the RMSEA was 0.060 (90% CI = 0.052–0.067, p = .019), which also suggested acceptable model fit (Browne & Cudeck, 1993; Byrne, 2012). All parameter estimates were statistically significant (p < .001) and all standard errors appeared to be of appropriate size, suggesting that all items fit well within the model (Byrne, 2012). Based on the acceptable values yielded for the model fit indices and statistical significance of the parameter estimates, the factor structure appeared to demonstrate adequate stability and this model was retained (see Figure 2)⁴.

Figure 2.

Model yielded by the confirmatory factor analysis. F1 represents the first factor, “Professor Cares about Students.” F2 represents the second factor, “Professor Creates an Engaging and Constructive Atmosphere.” For wording of items please refer to Table 4. Values in parentheses are standard errors.

Predictive Validity of the PSRS Factors

For the multiple linear regressions assessing the predictive validity of the PSRS factors, we returned to the full data set containing 760 cases. A missing values analysis revealed less than 5% missing data on all variables except for actual final grade. Independent samples t-tests revealed no significant differences on any of the predictors or other outcomes due to presence or absence of final grade. Therefore, cases with missing data were deleted listwise, as was done for the factor analyses, across all variables (N = 655). See Table 5 for correlations between the two factors and six outcomes, as well as descriptive statistics.

Table 5.

Correlations Between Factors and Outcomes: Means, Standard Deviations, and Medians.

Measures	1	2	3	4	5	6	7	8
F₁^a	1
F₂^b	.65**	1
Perception of Teacher	.64**	.87**	1
Perception of Course	.55**	.86**	.84**	1
Perceived Learning	.51**	.82**	.80**	.85**	1
Teacher Effectiveness	.72**	.88**	.88**	.83**	.80**	1
Expected Final Grade	.22**	.08*	.07	.15**	.12**	.11**	1
Actual Final Grade	.13**	−.01	−.03	.03	.03	.02	.84**	1
Mean (SD)	4.32 (0.59)	4.33 (0.80)	4.53 (0.91)	4.34 (0.93)	4.31 (0.98)	4.52 (0.70)	88.31 (7.32)	87.69 (9.58)
Median	4.50	4.63	5.00	4.67	5.00	4.88	90.00	90.00

^aF₁ – “Professor Cares about Students.”

^bF₂ – “Professor Creates an Engaging and Constructive Atmosphere.”

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

In order to circumvent overpowering the set of analyses, the dataset without any missing data (N = 655) was split into six sub-datasets (Ns = 109–110), one for each outcome being examined. These subsamples were sufficiently large to provide adequate power, as calculated using G*Power 3 (i.e., power = .80, α = .05, and effect size f² = 0.15; Faul et al., 2007).

The distribution of all DVs and IVs within the six sub-datasets was non-normal, and characterized by negative skew. However, examination of scatter plots and residual scatterplots, histograms, and P-P Plots indicated sufficient linearity and homoscedasticity between the predictors and all outcomes (Field, 2013, p. 348). Within each regression, correlations between predictors did not exceed .90, tolerance statistics were above 0.20, and variance inflation factor (VIF) values were less than 10.00, suggesting no major issues of multicollinearity or singularity (Field, 2013, p. 325). Due to the non-normality of all variables, bias-corrected accelerated (BCa) bootstrapping with 1,000 samples and 95% confidence intervals was used. BCa does not rely on assumptions of normality and accounts for skewness and bias in the bootstrap distribution (Efron, 1987); thus, this method is robust to the negative skew in the data as well as any undue influence exerted by possible outliers. Within the sample, outliers were simply the minority of participants who did not perceive their instructors as fantastic. Therefore, because any undue influence on their part was accounted for through BCa bootstrapping, and because their designation as outliers was purely an artefact of the considerable negative skew of the data, “outliers” were retained for the regression analyses.

All assumptions of multiple regression were met or accounted for in some way. Despite the factor correlation of .70 exhibited in the EFA as well as the moderately high correlation between the average scores for the two factors displayed in the correlation table (r = .65, p < .01, Table 5), both factors were entered simultaneously into each regression. Although this approach facilitates replication of previous methods (Wilson & Ryan, 2013), it was also deemed appropriate because it allowed for the assessment of each factor’s unique contribution to each outcome as well as the amount of overlap between the two factors with respect to each outcome. Though one downside of entering the two simultaneously is the possibility that high amounts of shared variance between the two factors might obscure an otherwise significant univariate relationship with an outcome for one or both factors, we were of the opinion that the insights yielded by this multivariable method outweighed this potential drawback.

The results of the regressions indicate the differential importance of the two factors (see Table 6). Both factors were significant predictors of perception of teacher and teacher effectiveness, and for both outcomes the second factor accounted for just over 25% of the unique variance, indicating a considerable overlap (over 40%) between the two factors in predicting the outcome. Only the second factor predicted perception of course and perceived learning. For these outcomes, about 60% of the total variance accounted for was unique, while the rest was shared with the first factor. Indeed, this pattern of high shared variance was evident across all outcomes that were significantly predicted, a finding of which readers are advised to remain cognizant. For expected final grade, only the first factor was a significant predictor. Interestingly, the two factors did not significantly predict actual final grade.

Table 6.

Regression Analyses Predicting Outcomes from Professor-Student Rapport Factors.

		Semi-partial ²	B	β	t	95% CI	BCa Bootstrapped 95% CIs
Perception of teacher
R² = .78	F ₁ ^a	.04	0.40	.27	4.53***	0.23, 0.58	0.04, 0.67
	F ₂ ^b	.26	0.78	.68	11.17***	0.64, 0.92	0.55, 1.02
Perception of course
R² = .80	F ₁	<.01	−0.06	−.03	−0.61	−0.24, 0.13	−0.25, 0.11
	F ₂	.50	1.03	.92	16.36***	0.91, 1.16	0.88, 1.17
Perceived learning
R² = .73	F ₁	<.01	−0.11	−.06	−0.88	−0.36, 0.14	−0.38, 0.10
	F ₂	.44	1.16	.89	13.01***	0.98, 1.34	0.98, 1.36
Teacher effectiveness
R² = .74	F ₁	.05	0.30	.28	4.42***	0.16, 0.43	0.17, 0.44
	F ₂	.27	0.50	.66	10.36***	0.40, 0.59	0.40, 0.61
Expected final grade
R² = .13	F ₁	.13	5.51	.44	3.95***	2.75, 8.28	2.92, 8.48
	F ₂	.02	−1.96	−.19	−1.73	−4.21, 0.29	−3.92, 0.09
Actual final grade
R² = .03	F ₁	.02	3.47	.22	1.57	−0.91, 7.86	−0.71, 7.53
	F ₂	<.01	−0.62	−.05	−0.38	−3.86, 2.61	−4.64, 3.48

Note. Significance was retained for all bootstrapped solutions, but was attenuated, going from p < .001 to p < .01 (range = .001-.002) in all cases but F₁ as a predictor of Perception of Teacher (p = .03).

^aF₁ – “Professor Cares about Students.”

^bF2 – “Professor Creates an Engaging and Constructive Atmosphere.”

***p < .001.

Discussion

The purpose of this study was to clarify the factor structure of the PSRS and to establish its predictive validity with regard to previously examined student and course outcomes (e.g., Demir et al., 2019; Richmond et al., 2015; Rogers, 2015; Wilson et al., 2010; Wilson & Ryan, 2013). The components from Wilson and Ryan (2013) were not replicated. Instead, two new factors emerged: “Professor Cares about Students” and “Professor Creates an Engaging and Constructive Atmosphere.” A two-factor solution demonstrated stability across four different combinations of extraction and rotation and was supported by the results of a parallel analysis. Additionally, the two-factor structure displayed adequate model fit in the CFA, further supporting its stability. While the model fit statistics in the CFA for the new factor structure were virtually equivalent to those found for the two components from Wilson and Ryan’s (2013) PCA (see footnote 4 above), we argue that the factor structure that emerged from the current work demonstrates better conceptual consistency/clarity and better captures the construct of professor-student rapport as a whole.

Multiple regressions provided some preliminary evidence for predictive validity with regard to all outcomes examined other than actual final grade. Comparison of the standardized regression coefficients contained in Tables 1 and 5 suggests that the association between the new factors and outcomes is comparable to or greater in magnitude than those previously found, with the exception of actual final grade. However, results for this outcome have been inconsistent in previous studies (Table 1; Demir et al., 2019). The current study makes a valuable contribution to the growing literature on professor-student rapport by extending previous findings to students and instructors outside of psychology, making the results more generalizable.

One crucial question we raised concerned whether the PSRS-B adequately represented the broader construct of professor-student rapport. As has previously been pointed out (Wilson & Ryan, 2013), it appears to capture only whether the student enjoys and wants to attend class, which does not seem representative of the construct as a whole. The two factors reported herein appear to better capture and clarify professor-student rapport. The first factor, “Professor Cares about Students,” retains items that better capture the actual interpersonal relationship between the professor and student (e.g., “My professor and I communicate well”) that are lost in the PSRS-B. Though it is not part of the PSRS-B, even when comparing the “Professor Cares about Students” factor to the “Perceptions of Teacher” component, it is clear that the interpersonal relationship element of rapport is much better captured by the first factor (i.e., “Professor Cares about Students”) that emerged in the current study (Figure 1a). The nine items in the “Perceptions of Teacher” factor do not appear to capture very well any sort of one-on-one interaction between the professor and student from which interpersonal rapport might be derived. The wording of some items in the “Professor Cares about Students” factor, on the other hand, do capture this essential element of one-on-one interaction between a professor and individual student rather than the entire class (e.g., “My professor encourages me to succeed”). The second factor, “Professor Creates an Engaging and Constructive Atmosphere,” contains items in common with the “Student Engagement” component that fully capture this theme. Additionally, it contains items that better encompass the overall atmosphere the professor creates. Thus, the second factor describes a professor’s interactions with the entire class rather than interpersonal relationships between a professor and individual students.

Our results suggest that professor-student rapport is established when two main ingredients are present. First, students must perceive the professor cares about them and their success. This idea is consistent with previous findings emphasizing the importance of behaviors that communicate a genuine investment in students’ well-being (personally and academically), including listening to them, being approachable and accessible outside of the classroom, and learning/remembering names (Faranda & Clarke, 2004). Second, students must perceive that the professor is effective in creating a classroom atmosphere in which there is structure, learning, and they feel engaged. As Wilson and Ryan (2013) point out, rapport that is useful for student outcomes entails not only being a likable teacher but also creating an engaging classroom environment for students. Previous research suggests professor-student rapport can enhance indicators of engagement, including pro-academic behaviors (Benson et al., 2005) and class performance (Rogers, 2015; Wilson & Ryan, 2013). Richmond et al. (2015), among others (Table 1), demonstrated that professor-student rapport is a strong predictor of SRIs (i.e., teacher effectiveness), but it seems likely that the reverse is also true: effective teaching contributes to rapport.

Other scholars have also highlighted the importance of having both ingredients to create rapport. Meyerberg and Legg (2015) posit that professor-student rapport is an integral part of establishing a learning alliance with a student, that is, purposive and collaborative behaviors on the parts of students and teachers toward the common goal of student learning (Rogers, 2012). There is a strong positive correlation between the PSRS and the Learning Alliance Inventory (Rogers, 2015), suggesting that students seek to collaborate with instructors that they perceive not only care about them but that will be effective in advancing them toward their goals. Though instructors can begin to demonstrate their investment in students before courses even begin—for example, with learner-centered syllabi (Richmond et al., 2016)—demonstrating that they will be effective in advancing students toward their goals may depend more on early perceptions of the atmosphere instructors create within the classroom. It is possible that the first factor, “Professor Cares about Students,” is an integral part of establishing interpersonal rapport, but that the second factor, “Professor Creates an Engaging and Constructive Atmosphere,” introduces the element of mentorship that moves the relationship beyond interpersonal rapport to professor-student rapport specifically.

Our results provide potential insights into which of the two aspects of professor-student rapport is most important for each outcome examined. Perception of teacher was predicted by both factors, but the second factor explained approximately six times more unique variance than the first, suggesting the atmosphere an instructor creates may impressions of them more than perceptions of much they care about students. This was true for teacher effectiveness as well. Because most participants were rating instructors with whom they had large- or medium-sized courses, it may be the case that one-on-one interactions were rare and thus perceptions of teachers and their effectiveness were necessarily based on interactions between them and the entire class, which were better captured by the second factor. Perceiving that an instructor cares, however, was still an important component, and probably more so than the results suggest due to the high amount of shared variance between the two factors.

The second factor emerged as the only significant predictor for perception of course as well as perceived learning, suggesting that the atmosphere an instructor creates may be more important for these outcomes than perceptions that they care. With respect to expected final grade, only the first factor emerged as a significant predictor. This might suggest that perceiving a professor cares about students is associated with higher student performance if not for the fact that actual final grade was not predicted by either factor. In light of this, it is perhaps the case that perceiving a professor cares is associated with some other indicator of student success that expected final grade is indirectly measuring. In the presence of mutual care, students may feel more motivated to perform well, which could result in more favorable estimations of one’s performance.

As for actual final grade, it could be the case that these two aspects of professor-student rapport do not influence a student’s grade in a course, or that within the current sample students felt strong rapport regardless of grades. However, there was considerable restriction in range for the final grades within our sample, the vast majority earning a B or higher in their classes. It is possible that the first factor, which explained more unique variance than the second, may significantly predict actual final grade in a sample with more variability in class performance.

Conclusion

Limitations and Future Directions

Future studies should provide additional psychometric validation for the new factors by establishing convergent validity, establishing test-retest reliability, and predictive validity using both previously used and new outcomes. More research is also needed on the mechanism by which professor-student rapport fosters student success. Self-determination theory provides one potential explanation with previous research demonstrating positive associations between autonomy support and student success (Black & Deci, 2000; Williams & Deci, 1996). Demir et al. (2019) found autonomy support and professor-student rapport were strongly correlated but differentially predictive of student and course outcomes. This suggests that while professor-student rapport may function in part by supporting the psychosocial need for autonomy, it may support other psychosocial needs as well.

Although we believe the current study represents a more appropriate solution with regard to factor structure and reduction of the PSRS, several limitations should be addressed. First, the professor-student rapport data displayed high negative skew. Thus, it might be the case that the factor structure reported herein is reflective only of instructors already motivated to establish rapport with students. Future studies should attempt to replicate the factor structure in a more normally distributed sample by seeking out a group of instructors with a greater range of likability, possibly through blind recruitment (i.e., asking the instructor for permission to collect data from their class without telling them the nature of the study).

It is also possible that there were moderating effects of course level and course size. However, the associations found in the regressions in the current study were quite robust, which lessens concern that any such moderating relationship considerably decreased these associations. Another possible confound includes students’ prior experience with an instructor, which was not assessed in the current investigation. Rapport may increase when students take multiple classes with an instructor, and they may enroll in additional classes with instructors they like. Familiarity with how an instructor organizes their classes, moreover, might facilitate student performance independently of rapport.

Predictor-criterion overlap is also a concern in the current study due to high correlations (>.80) between predictors and outcomes. Discriminant validity may be the most important next step with regard to psychometric validation, and researchers should be cognizant of the potential presence of items measuring rapport within scales one might use for this purpose. Another limitation was the use of outcome variables with few (1–3) items. Future research should also assess whether the trajectory of these two factors over the course of the semester predicts student outcomes, as previous research suggests that maintaining or increasing rapport over time may be important considerations for student success (Lammers et al., 2017).

Another possible concern regarding the factor structure reported in the current study is that it explained 50% of the variance in the items retained, which may fall short of some metrics of adequacy. However, whereas the percentage of explained variance in a PCA is a direct index of the quality of the analysis (Jolliffee & Cadima, 2016), this is not as clearly the case in EFA, especially in the social sciences where 60% or even less is considered satisfactory (Hair et al., 2014).

Finally, it remains to be seen whether the findings obtained in the current study with face-to-face classes would be replicated for online courses. Future research should investigate the challenges associated with establishing perceptions that the instructor cares and creates an engaging atmosphere in online formats. Considering the findings of the current study showing that the second factor was more important than the first in predicting many of the outcomes, it would be ideal for instructors to learn and practice how to create an engaging and constructive climate in their online courses (e.g., through courses such as those provided by Quality Matters; https://www.qualitymatters.org/).

Available research clearly demonstrates that perceptions of professor-student rapport have implications for student outcomes (Benson et al., 2005; Buskist & Saville, 2004; Demir et al., 2019; Frisby & Martin, 2010; Frisby & Myers, 2008; Lammers & Gillaspy Jr., 2013; Lammers et al., 2017; Richmond et al., 2015; Rogers, 2015; Ryan & Wilson, 2014; Wilson et al., 2010; Wilson & Ryan, 2013). A good measure of professor-student rapport, therefore, is an important tool for scholars of teaching and learning as well as instructors wishing to assess their own ability to establish rapport with students. To this end, we tested the replicability of the PSRS (Wilson et al., 2010), a measure of professor-student rapport derived from student feedback. Though the components previously reported for the PSRS (Wilson & Ryan, 2013) were not replicated, a distinct two-factor structure emerged in the current study that we argue better captures and elucidates the construct of professor-student rapport. The first factor, “Professor Cares about Students,” reflects students’ perceptions that the instructor cares about their well-being and success. The second factor, “Professor Creates an Engaging and Constructive Atmosphere,” reflects students’ perceptions of how well the instructor interacts with the class overall. In other words, instructors high on this factor create an atmosphere that not only holds students’ attention, but is clear in structure and conducive to learning. Our recommendation to researchers/teachers wishing to use these two factors in their own research or teaching is to consider them separately as they may be differentially predictive of various student success outcomes. An exciting direction for future research will be to further enhance our understanding of which dimension is more critical for promoting which student outcomes: when does student success depend on that student feeling cared for as an individual, and when does it depend on being part of an overall environment in which rapport is built through engagement and clear expectations?

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Timothy W. Broom

Notes

References

Benson

A. T.

Cohen

L. A.

Buskist

(2005). Rapport: Its relation to student attitudes and behaviors toward teachers and classes. Teaching of Psychology, 3(4), 237–239. https://doi.org/10.1207/s15328023top3204_8

Black

A. E.

Deci

E. L

. (2000). The effects of instructors' autonomy support and students' autonomous motivation on learning organic chemistry: A self-determination theory perspective. Science Education, 84(6), 740–756. https://doi.org/10.1002/1098-237X(200011)84:6<740::AID-SCE4>3.0.CO;2–3

Browne

M. W.

Cudeck

(1993). Alternative ways of assessing model fit. In Bollen

K. A.

Long

J. S.

(Eds.), Testing structural equation models (pp. 136–162). SAGE.

Buskist

Saville

B. K.

(2004). Rapport building: Creating positive emotional contexts for enhancing teaching and learning. In Perlman

McCann

L. I.

McFadden

S. H.

(Eds.), Lessons learned: Practical advice for the teaching of psychology (Vol. 2, pp. 149–155). American Psychological Society.

Byrne

B. M.

(2012). Structural equation modeling with Mplus: Basic concepts, applications, and programming. Routledge.

Demir

Burton

Dunbar

(2019). Professor-student rapport and perceived autonomy support as predictors of course and student outcomes. Teaching of Psychology, 46(1), 22–23. https://doi.org/10.1177/0098628318816132

Efron

. (1987). Better bootstrap confidence intervals. Journal of the American Statistical Association, 82(397), 171–185. https://doi.org/10.1080/01621459.1987.10478410

Estepp

C. M.

Roberts

T. G.

(2015). Teacher immediacy and professor/student rapport as predictors of motivation and engagement. NACTA Journal, 59(2), 155–163. https://www-jstor-org-443.web.bisu.edu.cn/stable/nactajournal.59.2.155

Faranda

W. T.

Clarke

(2004). Student observations of outstanding teaching: Implications for marketing educators. Journal of Marketing Education, 26(3), 271–281. https://doi.org/10.1177/0273475304268782

10.

Faul

Erdfelder

Lang

A.-G.

Buchner

(2007). G*Power 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39(2), 175–191. https://doi.org/10.3758/BF03193146

11.

Field

. (2013). Discovering statistics using IBM SPSS Statistics (4th. ed.). SAGE.

12.

Filak

V. F.

Sheldon

K. M.

(2008). Teacher support, student motivation, student need satisfaction, and college teacher course evaluations: Testing a sequential path model. Educational Psychology, 28(6), 711–724. https://doi.org/10.1080/01443410802337794

13.

Frisby

B. N.

Martin

M. M.

(2010). Instructor–student and student–student rapport in the classroom. Communication Education, 59(2), 146–164. https://doi.org/10.1080/03634520903564362

14.

Frisby

B. N.

Myers

S. A.

(2008). The relationships among perceived instructor rapport, student participation, and student learning outcomes. Texas Speech Communication Journal, 33(1), 27–34.

15.

Gremler

D. D.

Gwinner

K. P.

(2000). Customer-employee rapport in service relationships. Journal of Service Research, 3(1), 82–104. https://doi.org/10.1177/109467050031006

16.

Hair

J. F.

Black

W. C.

Babin

B. J.

Anderson

R. E.

(2014). Multivariate data analysis. Pearson.

17.

Horvath

A. O.

Greenberg

L. S.

(1989). Development and validation of the Working Alliance Inventory. Journal of Counseling Psychology, 36(2), 223–233. https://doi.org/10.1037/0022-0167.36.2.223

18.

L. T.

Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1–55. https://doi.org/10.1080/10705519909540118

19.

Infante

D. A.

Wigley

C. J.

(1986). Verbal aggressiveness: An interpersonal model and measure. Communication Monographs, 53(1), 61–69. https://doi.org/10.1080/03637758609376126

20.

Jackson

D. L.

Gillaspy

J. A.

Jr Purc-Stephenson

(2009). Reporting practices in confirmatory factor analysis: An overview and some recommendations. Psychological Methods, 14(1), 6–23. https://doi.org/10.1037/a0014694

21.

Jolliffe

I. T.

Cadima

(2016). Principal component analysis: A review and recent developments. Philosophical Transactions. Series A, Mathematical, Physical, and Engineering Sciences, 374(2065), 20150202. https://doi.org/10.1098/rsta.2015.0202

22.

Krumrei-Mancuso

E. J.

Newton

F. B.

Kim

Wilcox

(2013). Psychosocial factors predicting first-year college student success. Journal of College Student Development, 54(3), 247–266. https://doi.org/10.1353/csd.2013.0034

23.

Lammers

W. J.

Gillaspy

J. A.

Jr (2013). Brief measure of student-instructor rapport predicts student success in online courses. International Journal for the Scholarship of Teaching and Learning, 7(2), 1–13. https://doi.org/10.20429/ijsotl.2013.070216

24.

Lammers

W. J.

Gillaspy

J. A.

Jr Hancock

(2017). Predicting academic success with early, middle, and late semester assessment of student–instructor rapport. Teaching of Psychology, 44(2), 145–149. https://doi.org/10.1177/0098628317692618

25.

Mehrabian

(1968). Some referents and measures of nonverbal behavior. Behavior Research Methods & Instrumentation, 1(6), 203–207. https://doi.org/10.3758/BF03208096

26.

Meyerberg

J. M.

Legg

A. M.

(2015). Assessing professor-student relationships using self- report scales. In Jhangiani

R. S.

Troisi

J. D.

Fleck

Legg

A. M.

Hussey

H. D.

(Eds.), A compendium of scales for use in the scholarship of teaching and learning (pp. 149–160). Division 2 of the American Psychological Association.

27.

Richmond

A. S.

Berglund

M. B.

Epelbaum

V. B.

Klein

E. M.

(2015). a + (b1) professor-student rapport + (b2) humor + (b3) student engagement = (Ŷ) student ratings of instructors. Teaching of Psychology, 42(2), 119–125. https://doi.org/10.1177/0098628315569924

28.

Richmond

A. S.

Slattery

J. M.

Mitchell

Morgan

R. K.

Becknell

(2016). Can a learner-centered syllabus change students’ perceptions of student–professor rapport and master teacher behaviors? Scholarship of Teaching and Learning in Psychology, 2(3), 159–168. https://doi.org/10.1037/stl0000066

29.

Rogers

D. T.

(2012). The learning alliance inventory: Instrument development and initial validation. International Journal for the Scholarship of Teaching and Learning, 6(1), 1–16. https://doi.org/10.20429/ijsotl.2012.060109

30.

Rogers

D. T.

(2015). Further validation of the learning alliance inventory the roles of working alliance, rapport, and immediacy in student learning. Teaching of Psychology, 42(1), 19–25. https://doi.org/10.1177/0098628314562673

31.

Ryan

Wilson

J. H.

(2014). Professor-student rapport scale: Psychometric properties of the brief version. Journal of the Scholarship of Teaching and Learning, 14(3), 64–74. https://doi.org/10.14434/josotl.v14i3.5162

32.

Ryan

R. G.

Wilson

J. H.

Pugh

J. L.

(2011). Psychometric characteristics of the professor-student rapport scale. Teaching of Psychology, 38(3), 135–141. https://doi.org/10.1177/0098628311411894

33.

Schriver

J. L.

Kulynych

R. H.

(2021). Do professor-student rapport and mattering predict college student outcomes? Teaching of Psychology. Advance online publication. https://doi.org/10.1177/00986283211037987

34.

Tabachnick

B. G.

Fidell

L. S.

(2013). Using multivariate statistics (6th ed.). Pearson.

35.

Williams

G. C.

Deci

E. L.

(1996). Internalization of biopsychosocial values by medical students: A test of self-determination theory. Journal of Personality and Social Psychology, 70(4), 767–779. https://doi.org/10.1037//0022-3514.70.4.767

36.

Wilson

J. H.

Beyer

Monteiro

(2014). Professor age affects student ratings: Halo effect for younger teachers. College Teaching, 62(1), 20–24. https://dx.doi.org.libproxy.nau.edu/10.1080/87567555.2013.825574

37.

Wilson

Ryan

(2013). Professor-student rapport scale: Six items predict student outcomes. Teaching of Psychology, 40(2), 130–133. https://doi.org/10.1177/0098628312475033

38.

Wilson

J. H.

Ryan

R. G.

Pugh

J. L.

(2010). Professor–student rapport scale predicts student outcomes. Teaching of Psychology, 37(4), 246–251. https://doi.org/10.1080/00986283.2010.510976

39.

Zimet

G. D.

Dahlem

N. W.

Zimet

S. G.

Farley

G. K.

(1988). The multidimensional scale of perceived social support. Journal of Personality Assessment, 52(1), 30–41. https://doi.org/10.1207/s15327752jpa5201_2

40.

Zygmont

Smith

M. R.

(2014). Robust factor analysis in the presence of normality violations, missing data, and outliers: Empirical questions and possible solutions. The Quantitative Methods for Psychology, 10(1), 40–55. https://doi.org/10.20982/tqmp.10.1.p04