Introduction
The concept of trust is a key feature of theories and models focusing on functioning and well-being in romantic relationships (, ; ). The work of is a landmark contribution that provided a theoretically rich foundation for capturing the trust people place in their intimate partners. Their theoretically derived measure and model for the development of trust is the primary source for modern thinking and research on the topic. However, research has not yet systematically evaluated the proposed factor structure of trust scale, nor the similarities and differences in the nature of trust at different relationship stages. Additionally, the little research that has assessed the development of trust presents a potential paradox: people in new relationships report placing similarly high levels of trust in their romantic partners compared to those in long-lasting relationships (e.g., ; ).
With the present research, we conducted two measurement-focused studies to attain deeper insight into the nature of trust and explore an account of trust that reframes its development as a process of construction, rather than addition. In Study 1, we evaluated the soundness of trust scale with a large sample of romantically-attached individuals to first establish how to best model trust’s elements. In Study 2, we then used bifactor modeling and invariance testing techniques with large, independent samples of individuals involved in relationships at different stages to capture divergences in the conceptualization of trust as a function of relationship maturation.
Instantiation of trust
The concept of trust
Trust can stem from evidence—data collected from past experiences—and/or faith— conviction taken as truth without any supporting evidence. In romantic relationships, the components of trust that concern evidence include perceptions of the partner’s predictability and dependability. Predictability refers to the perceived likelihood of successfully foretelling the future actions of the partner. Perceptions concerning a partner’s predictability are thought to be based on observations of the nature and consistency of the partner’s behaviours (). Dependability refers to the degree to which the partner is perceived to embody a set of qualities approximating a “dependable archetype” (, p. 191). Specifically, people would be more likely to regard their partner as dependable when their partner has solidified themselves as, for example, honest, cooperative, and benevolent. Although dependability, much like predictability, bears on the behavioural outputs of the partner, dependability is slightly more nuanced in that it involves the abstraction of underlying narratives and trait attributions from the partner’s outputted behaviours (; ).
Unlike predictability and dependability, faith captures a sense of conviction untethered to evidence, enabling one to dismiss doubts concerning the unknown future (; ). With faith in their partner and the future of the relationship, people can transcend the need for supportive evidence and act as if the future will unfold as desired ().
The measurement of trust
To establish their measure, first created a compilation of 26 items presumed to reflect trust in a partner across the theorized dimensions and administered them to a sample of 47 long-term couples. The authors conducted an exploratory factor analysis (EFA) to analyze item responses and removed items on the grounds of (1) low item-total correlations and (2) poor loadings. The resulting 17-item questionnaire comprised a 5-item subscale reflecting predictability, a 5-item subscale reflecting dependability, and a 7-item subscale reflecting faith.
The authors regarded their research as an initial step towards gaining deeper insight into the nature of the theoretical entity of trust. However, the soundness and validity of their framework for capturing trust has undergone little psychometric investigation as most research emanating from their perspective has prioritized calculating an overall score to correlate with other measures of relationship functioning.
The development of trust
The concept of trust in romantic relationships has been approached from at least two major theoretical perspectives: attachment theory and interdependence theory (). According to attachment theory, trust is one component of individual differences in attachment orientations that begin developing in childhood. Early interactions with core attachment figures during infancy are thought to later be crystalized into lasting mental representations of “self” and “others” (; ). The responsiveness of the attachment figures to the infant’s needs is thought to shape the infant’s mental representations of “self” and/or “others”, leading to the formation of either a more secure or more insecure attachment orientation. Such attachment (in)security furnishes intuitions and beliefs about what should be expected from an intimate partner(ship)—whereby individuals with a more secure attachment orientation tend to have greater ease in being vulnerable and extending trust to close others, and those with a more insecure attachment orientation tend to be generally less trusting in their relationships (e.g., ).
Whereas individual differences (e.g., attachment orientations) may represent a readiness to trust, an interdependence theoretical perspective suggests that experiences within a given relationship regulate trust’s development (; ). Trust’s growth in relationships is theorized to be linked to the increase in interdependence between those involved—a view that guided approach to the construct. framed the development of trust in terms of uncertainty reduction whereby “… trust evolves out of successfully confronting increasing concerns about dependency as relationships develop” (p. 190). As partners become more interdependent, successfully reconciling issues raised by this increasing interdependency allows for trust to grow. Such issues take the form of, for example, finding oneself in positions of vulnerability or conflicting interests (; ). In a similar vein, the development of trust is believed to be rooted in diagnostic situations—moments when the partner must navigate personal interests and those of the relationship (). Successfully confronting diagnostic situations can provide insight into the promise of the partner(ship) and, in turn, enrich trust. From this perspective, therefore, the trust people place in their romantic partners ought to grow over time as experiences accumulate and issues of interdependency are confronted ().
Paradox of trust
This account of trust however presents a paradox when compared with existing data: from the earliest through to the deepest stages of a romantic relationship, the amount of trust people place in their partner appears to remain appreciably high (e.g., ; ). Therefore, trust in a romantic partner might inherently exist at (or somewhere near) the maximum amount from the earliest stages of the relationship onward, potentially rendering accumulated experiences with the romantic partner an unnecessary precondition for developing trust over time.
Reimagining the development of trust
The pursuit of a deeper understanding of the nature of trust’s development is predicated on the manner in which growth is defined (). In the literal sense, the growth of trust can reflect a positive linear trend from relatively low to high amounts over time; in the figurative sense, the growth of trust can reflect a metamorphic construction of a person’s conceptualization of trust over time (). In our current research, we endeavoured to explore an account of trust that frames trust’s development as a process of construction in meaning, as opposed to addition in amount.
A defining feature of this view of trust is that it is responsive to important events and activities occurring within the confines of the relationship (). That is, the major issues and themes in focus at the given period of a relationship’s development might give insight into the contents by which people construct their trust for their partner. Therefore, as an initial foray into reconciling the paradox of trust, we focused on investigating the (in)variability in the constructions of trust of individuals involved in newly-formed relationships versus those involved in long-term relationships.
Description of research objectives and study
With our present research, we endeavoured to garner deeper insight into the nature of the construct of trust in romantic relationships. In Study 1, using data collected from a large sample of individuals involved in romantic relationships, we explored the form and essence of the construct of trust as put forward by . In Study 2, we confirmed our findings from Study 1, and then using our refined version of assessment of trust, proceeded to capture the extent to which conceptualizations of trust across individuals involved in newly-formed relationships versus those involved in long-term relationships reflect different forms of the construct. We analyzed all our data using the R programming language (). All our pre-registrations, data, analytic scripts, and supplementary materials are available on our Open Science Framework repository (https://osf.io/4km73/).
Study 1
For Study 1, we evaluated measurement framework for assessing trust as a way to gather deeper insights into the construct.
Method
Participants and procedure
Study 1 included 494 individuals recruited from Prolific (https://www.prolific.co/). Eligible participants needed to be of legal age to consent, speak English fluently, and be in an ongoing romantic relationship for at least 2 years. Consenting participants were given a link to a Western Qualtrics survey. Upon completion, participants received a code to enter into the Prolific application to receive compensation (i.e., £2.00). We removed individuals who provided incorrect answers to any of the four attention check questions (n = 2). Participants had a mean age of 40.80 years (Median = 38, SD = 12.90, Min = 19, and Max = 79). Participants were based in either the United Kingdom (82.20%), Canada (11.90%), or the United States (4.25%). Most participants were non-student individuals (85.80%) and were employed in full-time (49.60%) or part-time work (18.60%). See Table 1 for a demographic breakdown of the Study 1 participants.
Predictability, dependability, and faith
We used 26-item scale to capture participants’ trust in their romantic partner across the dimensions of predictability, dependability, and faith (see Supplementary Materials; https://osf.io/f32ds/). All items were evaluated on a 1 (strongly disagree) to 7 (strongly agree) scale and presented to participants in a randomized fashion.
Data analysis strategy
EFA
We used EFA, via the psych package (), to determine a psychometrically optimal and meaningful number of factors to represent items. We employed maximum likelihood extraction with oblimin rotation to estimate factor solutions. EFA model selection was based on a combination of parallel analysis, nested model comparisons, model fit indices, and conceptual interpretability of factor loading patterns.
Construct multidimensionality due to fallible indicators
The construct multidimensionality present in assessments of psychological processes can emerge from: (1) the natural tendency of indicators to capture unintended processes; and/or (2) overarching latent constructs. These competing sources of construct multidimensionality, indeed, create ambiguity about whether the multidimensionality captured within a given assessment is best conceptualized as a consequence of the innate fallibility of the item contents at hand (e.g., presence of cross-loading indicators) or the hierarchical nature of the construct under study (e.g., presence of general factors) ()—potentially undermining, in turn, the assessment’s construct validity. As remarked, the major threats to the construct validity of assessments of psychological processes are construct-irrelevant variance and construct underrepresentation. Specifically, construct-irrelevant variance describes the degree to which the assessment and items therein tap into factors or processes that are adjacent or extraneous to the target construct (e.g., methodological artifacts), whereas construct underrepresentation describes the degree to which the assessment fails to capture the construct’s theoretical form and aspects (; see , for a review).
Given that assessment is multidimensional yet designed to capture an overarching construct of trust, we used strategies described in to investigate the sources of construct multidimensionality present in our chosen exploratory measurement model and compilation of indicators. For our chosen model, we contrasted an Exploratory Structural Equation Model (ESEM) variant comprised of cross-loading indicators, versus a Confirmatory Factor Analysis (CFA) model variant comprised of fully discriminatory indicators. If the model’s fit and parameter estimates (i.e., factor correlations) for its CFA variant do not differ substantially from its corresponding ESEM variant, then the presence of construct multidimensionality due to the fallible nature of the model’s indicators can be deemed trivial. For further details regarding this approach, see as well as the Supplementary Materials.
The ESEM and CFA model variants, programmed via lavaan (), were identified using the fixed-factor scale-setting and identification technique (), and fitted using robust maximum-likelihood estimation (MLR). Missing data were handled using full-information maximum-likelihood (FIML). Missingness accounted for approximately 0.13% of the analyzed data (see Supplementary Materials for further details). Though test indicated our missing data did not meet the MCAR (i.e., missing completely at random) assumption (χ2 (246) = 329, p < .05), FIML’s performance and handling of missing data (e.g., production of unbiased parameter and error estimates) has proven to be highly effective under missing data conditions driven by non-MCAR mechanisms—especially when the extent of the missingness is under 25% (; ), as in our case (0.13% of missing data).
Results
Observed means and standard deviations of trust, predictability, dependability, and faith for Study 1 are available in the Supplementary Materials.
EFA
The results of the parallel analysis suggested four factors were sufficient (see Supplementary Materials). Therefore, we evaluated solutions comprising of one-to-four factors (see Table 2). The evidence from nested model comparisons and model fit indices led us to further interpret and examine the four-factor solution (see Table 3) as a final candidate measurement model.
Close inspection and refinements
Loadings of |.35| or greater were deemed substantial. Recognizing that cross-loading indicators are not solely a threat to simple structure and can potentially contribute meaningfully to multidimensional construct spaces, selectively cross-loading indicators—substantially loading onto only a particular set of factors—were retained and permitted to be associated with the most conceptually compatible factor, whereas completely cross-loading indicators—(un)substantially loading onto all or no factors—were discarded (see ). Items 3 and 15 were identified as selectively cross-loading and retained, whereas items 1 and 23 were identified as completely cross-loading and discarded—leaving 24 indicators loading onto four factors. Compatible with framework, one factor reflected indicators addressing trait attributions of the romantic partner’s character (dependability), whereas another factor reflected indicators addressing aspirational convictions in the romantic partner/relationship (faith). Of the remaining two factors, one factor reflected indicators addressing predictability whereas another factor reflected purely negatively worded indicators, potentially highlighting a demarcation of bona fide predictability content from general distrust content.
We decided to discard the indicators from the general distrust factor for two reasons. First, given that this family of indicators are all reverse-worded items their constellated factor might address methodological artifacts. Second, trust and distrust, though related, can be examined independently. Assuming trust and distrust form endpoints of a common continuum potentially goes beyond the scope of proposed framework that was designed to capture the concept of trust.
A factor reflecting an abundance of negatively worded indicators was consistent with the findings of . characterized their factor comprised of negatively worded items as a lack of predictability rather than the presence of predictability. However, they did not test a competitor 4-factor model as an underlying structure of their items and, thus, were unable to evaluate the degree to which bona fide predictability-related item content can be effectively demarcated from distrust-related item content (and/or methodological artifacts). Taken together, the final measurement structure suggested by our analyses contained 17 indicators capturing perceived partner predictability, perceived partner dependability, and faith (see Tables 2 and 4).
Construct multidimensionality due to fallible indicators
Results from our investigations of sources of construct multidimensionality in Study 1 are summarized in Table 4 and the text below, but see Supplementary Materials for more details. In terms of goodness-of-fit indexes, both model variants—following standards outlined in (e.g., ≥.95 for the Comparative Fit Index [CFI] and Tucker-Lewis Index [TLI]; ≤.06 for the Root Mean Square Error of Approximation [RMSEA])—provided a tenable degree of fit to the data. In terms of information criteria, whereas the ESEM variant showed a slight improvement in model fit according to the Akaike Information Criterion (AIC), the CFA model variant showed a slight improvement in model fit according to the Bayesian Information Criterion (BIC). Model fit was therefore ambivalent for which variant ought to be favoured. Concerning parameter estimates, though the factor correlations differed across the ESEM (r = .69 to .84) and CFA variants (r = .79 to .83), the differences were not substantial, suggesting our chosen measurement structure may host a non-critical mass of indicators tapping multiple adjacent, non-target constructs (see ; as well as Supplementary Materials). Altogether, these results suggest the fallible nature of the indicators can be deemed a minimal source of the construct multidimensionality present in our chosen measurement structure.
Discussion
In Study 1, we identified an interpretable and coherent compilation of items organized by underlying latent factors that appear to address constructs akin to framework (see Table 5). One factor reflected the perceived predictableness of a romantic partner’s behavioural output; a second factor reflected trait attributions of the romantic partner’s dependability; and a third factor reflected faith in the romantic partner/relationship. Furthermore, the selected item compilation garnered support that precluded trust’s multidimensionality being sourced from the presence of items addressing multiple adjacent construct spaces. This finding may therefore hint at an overarching construct of trust as a potential source of trust’s multidimensionality—a possibility we address in Study 2.
Study 2
The accounts of trust theorists and lay notions assuming time as a necessary precondition for the development of trust seem to run paradoxical to the existing data (e.g., ; ). For Study 2, using our revised version of assessment, we confirmed our findings from Study 1 and investigated (a) the appropriateness of an overarching general construct of trust, and (b) the extent to which the trust of individuals involved in new versus mature romantic relationships reflect different constructions.
Bifactor modelling
Most researchers use a composite score of measure to represent overall trust. In Study 2, we therefore investigated, via bifactor modelling (), the appropriateness of a psychometric structure of trust reflecting an overarching conceptualization (i.e., general factor) and specialized conceptualizations (i.e., specific factors). Bifactor modelling allows us to directly evaluate the nature of (a) variation across the full array of indicators caused by an overarching general construct of trust (i.e., general factor); and (b) variation across cluster sets of meaningfully grouped items caused by specific constructs concerning trust (i.e., predictability, dependability, and faith). See Supplementary Materials for a visual representation of the parameterization of a prototypical bifactor model.
Invariance testing to investigate construction divergences
Measurement invariance techniques have profound utility for investigating (in)variability in the construction of psychological concepts and the nature of how such construction may unfold (see ; ). To examine the nature of trust’s construction as a function of romantic relationship maturation, we therefore contrasted, in a multigroup CFA framework, the trust of individuals involved in newly-formed relationships and those involved in long-term relationships, and demarcated construction divergences at the levels of configural, metric, and scalar invariance.
The test of configural invariance can capture (dis)similarities in trust’s form (i.e., number of factors and general pattern of loadings) across the modeled samples. The test of metric invariance—also often referred to as weak invariance or loading invariance—assesses whether the relations between the (observed) items and latent constructs (i.e., factor loadings) are equivalent across the modeled groups. Examinations into a measurement instrument’s metric invariance can, therefore, help to establish the extent to which the construct under study might carry (dis)similar meaning across the modeled groups (). The test of scalar invariance, also often referred to as strong invariance or intercept invariance, assesses whether the indicator intercepts (i.e., expected value of an item when its latent variable[s] is[are] at zero) are equivalent across the modeled groups (). Said differently, examinations into a measurement instrument’s scalar invariance can help to establish “whether an item has the same point of origin across different groups” (, p. 1006).
Method
Participants and procedure
Study 2 included 847 individuals—nNewly-formed = 387 and nLong-term = 460—recruited from Prolific (https://www.prolific.co/). Eligible participants needed to be of legal age to consent, speak English fluently, and be in a romantic relationship at the time of participation. The study was only made available to individuals reporting (via Prolific pre-screeners) being involved in an ongoing romantic relationship for 5 or more years (long-term relationships), or 6 months or less (newly-formed romantic relationships). Upon providing informed consent, participants were given a Western Qualtrics survey which included 26-item trust scale as well as questionnaires assessing the nature of their relationship and demographic backgrounds. Upon completion, participants received a code to enter into the Prolific application to receive compensation (i.e., £3.00). We removed individuals who provided incorrect answers to any of the four attention check questions (n = 16). Participants had a mean age of 35.26 years (Median = 33, SD = 12.19, Min = 18, and Max = 79). Participants were based in either the United Kingdom (61.74%), the United States (32.59%) or Canada (5.67%). Most participants were non-student individuals (75.91%) and were employed in full-time (64.34%) or part-time work (21.25%). See Table 1 for a demographic breakdown of the Study 2 participants.
Predictability, dependability, and faith
Predictability, dependability, and faith were assessed using 26-item Trust Scale—following the same administration method as Study 1.
Data analysis strategy
All analyses were performed using the lavaan () and semTools () packages. All models were identified using the fixed-factor scale-setting and identification technique () (see Supplementary Materials), and fitted using MLR; and missing data were handled using FIML. Missingness accounted for approximately 0.12% of the analyzed data (see Supplementary Materials for further details) and was—via test—demonstrated to be MCAR (χ2 [425] = 456, p = .145).
Investigation of sources of construct multidimensionality
Sources of construct multidimensionality due to fallible indicators
Sources of construct multidimensionality due to fallible indicators were investigated using the same data analysis strategy described in Study 1.
Sources of construct multidimensionality due to an overarching construct
Following a similar strategy described in , we contrasted, via a CFA statistical framework, the correlated three-factor model variant and bifactor model variant on two levels of model evaluation: model fit and, more critically, the presence of a well-defined general factor. To the extent that the bifactor model variant shows improved model fit indices and the presence of a well-defined general factor, then the appropriateness of a bifactor structure and presence of construct multidimensionality due to an overarching construct would be supported. To appraise model fit of the three-factor and bifactor model variants, we placed a heavier premium on interpreting model fit indicators that included a correction for parsimony (i.e., TLI, RMSEA, AIC, and BIC) as the bifactor model is a heavily parameterized variant and subject to overfitting. To appraise the presence of a well-defined general factor, we took particular care in examining the extent to which the strength of the loadings onto the designated general factor could be demonstrated across a range of item contents.
Investigation of measurement model (in)variability
We tested the state of invariability of the final candidate measurement model between individuals in newly-formed relationships versus individuals in long-term relationships—examining measurement invariance at both the model level (i.e., configural, metric, and scalar invariance) and parameter level (i.e., invariability of particular items and model parameters) (; ; ). Measurement invariance was evaluated via fitting a series of increasingly constrained multi-group CFA models—incrementally constraining model parameters to equality across groups—and inspecting changes in model fit between each level of invariance constraints imposed. We began with a constraint of an identical quantity of latent factors and broad factor loading pattern (test of configural invariance), then proceeded to constrain the factor loading estimates to equality (test of metric invariance), and, lastly, proceeded to constrain the indicator variable intercepts to equality (test of scalar invariance).
Configural invariance was appraised using the permutation randomization method (see ). Specifically, we used I = 1000 permutations via the semTools package () to compute p values of statistical significance associated with ΔCFI and ΔRMSEA. Metric and scalar invariance were examined within a nested model comparison framework. Our adjudications of metric and scalar invariance were guided by the likelihood ratio test (i.e., Δχ2), though—following the recommendations described in —we also computed ΔΓ and ΔCFI for descriptive purposes. Omnibus detections of metric and scalar non-invariance were further investigated by testing a series of models wherein the equality constraint of each indicator variable was released, and model fit was compared (i.e., via Δχ2) against the corresponding fully constrained model (; ; ).
Results
Observed means and standard deviations of trust, predictability, dependability, and faith for each sample are available in the Supplementary Materials.
Investigation of sources of construct multidimensionality
Summaries of our ESEM and CFA models are presented in Table 6.
Construct multidimensionality due to fallible indicators
In terms of goodness-of-fit, each of the correlated three-factor ESEM (i.e., containing estimated cross-loadings) and CFA model (i.e., fully discriminatory indicators) variants provided a tenable degree of fit to the data—according to cut-offs. However, whereas the ESEM variant was suggested to provide an improvement in model fit according to AIC, the CFA model variant was suggested to provide an improvement in model fit according to BIC. Thus, the model fit was ambivalent for which model variant ought to be favoured and, in turn, preliminarily precluded the presence of items inadvertently addressing multiple adjacent non-target construct spaces as a potential primary source of construct multidimensionality. Furthermore, although the factor correlations among the ESEM (|r| = .73 to .82) and CFA model (|r| = .80 to .87) variants differed, the differences were, descriptively, not substantial.
The pattern of findings suggested that, though the model fit to the data and latent factor correlations of the CFA model variant differed from those of the corresponding ESEM variant, the extent of the differences was, descriptively, not vast—thereby replicating and confirming the chosen exploratory model of Study 1. Indeed, of the construct psychometric multidimensionality represented in the candidate measurement structure under consideration (i.e., correlated three-factor model), the presence of items inadvertently addressing multiple adjacent nontarget construct spaces could be deemed a minimal source—which, in turn, spoke to the tenability of retaining the CFA model variant (i.e., modelling fully discriminatory indicator variables) for further testing. Therefore, using the CFA model variant, we proceeded to test the presence of an overarching construct of trust as an alternate source of construct multidimensionality.
Construct multidimensionality due to an overarching construct
Although both the correlated three-factor model and the bifactor model variant provided a tenable degree of fit to the data, the bifactor model showcased superior fit to the data according to the indicators of model fit that includes a correction for parsimony and penalty for over-parameterization. The full array of loadings and intercepts of the bifactor model are outlined in Table 7.
The bifactor model showcased that the overarching general factor of trust was well-defined by the presence of strong, significant target loadings across the full array of items (|λ| = .57 to .86, Median = .70). More specifically, however, the faith items (|λ| = .57 to .86, Median = .78) demonstrated elevated target loadings on the overarching general factor in comparison to the predictability (|λ| = .57 to .73, Median = .61) and dependability items (|λ| = .59 to .76, Median = .62). Over and above the overarching general factor of trust, the specific factors of predictability (|λ| = .42 to .53, Median = .44) and dependability (|λ| = .26 to .47, Median = .40) were also well-defined by virtue of appreciable target loadings. This suggests that the predictability and dependability items not only reflect the general content substance shared across the full array of items, but also tap into relevant specificity and address additional information beyond the general factor. The specific factor of faith, in contrast, was showcased to be defined by virtue of a compilation of appreciable as well as nonsignificant target loadings (|λ| = .05 to .30, Median = .09). This suggests that some faith items seem to reflect both the generality of the content substance shared across the full array of items as well as an appreciable degree of relevant specificity beyond the general factor, whereas the remaining faith items seem only to reflect the former.
Taken together, these findings speak to the appropriateness of an overarching concept and bifactor structure underlying responses to the compilation of item contents. We therefore chose the bifactor model variant as the final measurement model of trust and proceeded to our tests of invariance.
(in)Variability of the measurement model
Model-level (in)Variability
Summaries of our multigroup CFA models are presented in Table 6, whereas summaries of our model-level invariance tests are presented in Table 8. We submitted the chosen measurement model of trust (i.e., the bifactor model) to a suite of invariance tests on individuals involved in newly-formed romantic relationships versus individuals involved in long-term romantic relationships (nNewly-formed = 387 vs. nLong-term = 460). The loadings and intercepts of the configural invariance model are outlined in Table 9. This model had acceptable fit across the two independent samples (CFI = .98) according to cut-offs. More importantly, permutation tests using both CFI and RMSEA supported configural invariance (ΔCFI = .97, p = .441; ΔRMSEA = .053, p = .421). In comparison to the configural invariance model, the metric invariance model was found to fit significantly worse (Δχ2[30] = 67.35, p < .001). Finally, in comparison to the metric invariance model, the scalar invariance model was found to fit significantly worse (Δχ2[13] = 84.75, p < .001). Across the two modelled samples, the pattern of results suggests that the theoretical structure of the model was invariant configurally, yet the model seemed to be demonstrating non-invariability at the level of the loading and intercept parameters. Furthermore, a sensitivity analysis indicated that the imbalance between the modelled sample sizes was unlikely to have any practical impact on the outcomes of our model-level invariance tests (see Footnote 4).
Parameter-level (in)Variability
Due to the number of identified sources of non-invariance, we outline the patterns of loading and intercept non-invariability that we believe to be most relevant. We sought to establish partial invariance by relaxing the minimum number of equality constraints necessary to achieve a fit comparable to the baseline model (i.e., a nonsignificant Δχ2 test) (see ). Parameters producing the largest resulting decreases in χ2 were prioritized for relaxation (see Table 10 as well as the Supplementary Materials).
Partial metric invariance model
We fitted a candidate partial metric invariance model by relaxing the general factor loadings from items 16 and 18 as well as relaxing the faith factor loadings from items 14 and 15. This partial metric invariance model did not fit significantly worse than the configural invariance model (Δχ2[26] = 33.73, p = .142) and was, therefore, retained. Items 16 and 18 comprised loadings onto the general factor that were revealed to be stronger for individuals involved in long-term relationships than for individuals involved in newly-formed relationships. Such non-invariability might suggest the contents expressed in items 16 (“I can rely on my partner to react in a positive way when I expose my weaknesses to him/her”) and 18 (“When I share my problems with my partner, I know he/she will respond in a loving way even before I say anything”) resonate more with conceptualizations of trust within relationships that carry denser shared history, while taking on relatively lesser importance in how individuals in early-stage relationships may conceptualize the construct.
Among the relaxed faith factor loadings, the loading for item 14 was revealed to be stronger for individuals in newly-formed relationships, whereas the loading for item 15 was revealed to be stronger for those in long-term relationships. Despite these group differences, neither sample supported the emergence of a well-defined residualized factor from the modeled compilation of faith-based indicators (see Table 9). Thus, the parameter-level non-invariance identified within the faith factor should be interpreted with caution, as poorly defined specific factors can lack substantive meaning.
Partial scalar invariance model
The scalar invariance model constrained all intercepts of the retained partial metric invariance model to equality across individuals in newly-formed relationships versus those in long-term relationships. The scalar invariance model fit significantly worse than the partial metric invariance model (Δχ2[13] = 79.44, p < .001), indicating the presence of non-invariant intercepts. Following the diagnosis of intercept parameters responsible for the omnibus detection of scalar non-invariance (see Table 10), we fitted a partial scalar invariance model by freeing the intercepts of items 2, 7, 12, 18, 20, and 26. This partial scalar invariance model did not fit significantly worse than the partial metric invariance model (Δχ2[7] = 11.18, p = .131) and was, therefore, retained (i.e., no further modifications were examined). Those in newly-formed relationships exhibited higher indicator intercepts on all freed items compared to those in long-term relationships. Among the six items freed in the partial scalar invariance model, four comprised faith-based content (e.g., “I am willing to let my partner make decisions for me” [Item 26]), while the remaining two comprised predictability-based (“My partner behaves in a very consistent manner”) and dependability-based (“I am certain that my partner would not cheat on me, even if the opportunity arose and there was no chance that he/she would get caught”) content.
Discussion
In Study 2, we found support for the appropriateness of a general construct underlying responses to our revised version of assessment. Predictability and dependability item contents appeared to address the general construct as well as relevant specificity associated with their respective specialized constructs, whereas most faith items appeared to only address the general construct. Importantly, the model- and parameter-level tests of invariance garnered evidence supporting the notion that individuals in newly-formed relationships and individuals in long-term relationships may conceptualize trust for the romantic partner in non-equivalent ways. Although there was evidence for configural invariance, the evidence for metric and scalar non-invariability across individuals in newly-formed versus long-term relationships may reflect interpretive differences of the item contents and, in turn, potential divergences in the construction of trust. We discuss the implications of these results in the General Discussion.
General discussion
A goal of this research was to enhance our understanding of the nature of trust in romantic relationships by assessing the measurement as well as invariance of the construct between large samples of individuals in newly-formed compared to long-term romantic relationships. Some important implications of our research and results are that (a) we replicated the factor structure of trust as conceptualized and assessed by but with the important caveat that the predictability factor is better represented by positively-worded items than the negatively-worded items suggested by the original analyses; (b) the development of trust does not appear to begin at relatively lower levels in newly-formed relationships compared to relationships that have existed for a longer time; and, (c) according to the measurement invariance analyses there (i) is a great deal of consistency in the construct of trust between newly-formed and long-term relationships, but also (ii) some important differences that potentially suggest how trust might be conceptualized and/or measured differently over time. In the following sections, we discuss these implications for the field’s understanding of the nature of trust, including additional future research needed to test the possibilities suggested by our findings.
Measurement of trust
theory-based measure of trust is immensely popular, largely owing to its reputation for offering a greater depth of insight into the theoretical entity compared to other comparable measurement instruments (e.g., ; ). However, the quality of the data and methods employed in fall short of modern standards, and research building on their foundation with further (and ongoing) investigation has been absent. With large samples of individuals involved in romantic relationships, our psychometric investigation yielded a resulting measurement instrument and model that broadly dovetails with conceptualization, as well as garnered psychometrically informed insights on how to better model trust’s elements.
In line with , our results pointed to a multidimensional, tripartite conceptualization of trust reflecting a demarcation of predictability-, dependability-, and faith-based item contents. Our resulting measurement instrument, however, assesses these constructs through a compilation of items slightly different from those selected by . Nevertheless, such differences in the item compilations proved necessary for reclaiming a sound measurement structure that upholds the hierarchical nature of trust (i.e., an overarching, general construct of trust) and, in turn, an overall composite score of trust (see ; )—representing an improvement that should benefit future research using this modified scale.
Furthermore, representing trust’s multidimensionality and hierarchical organization under a bifactor model (i.e., a coexistence of a general construct and specific subconstructs) proved fruitful in terms of empirical adequacy and for garnering insight into the properties of the individual test items and substance of the constructs being assessed by those items. We found support for a general construct of trust that was defined most strongly via faith-based items. Additionally, our findings suggested that whereas predictability and dependability-based items can address meaningful residual specificity (i.e., information that is not conveyed by the general construct), the majority of faith-based items offered little meaningful residual specificity and instead reflected pure indicators of the general construct. An important step for future investigations on the measurement front of trust research would be to uncover item contents that can effectively address not only trust but also substantive faith specificity. Future forays into the content substance of the faith that romantically-attached individuals hold for their partner may particularly benefit from the use of qualitative methods (e.g., interviews, text analysis, etc.) and item response theory techniques (e.g., ).
Development of trust
It is generally assumed that trust grows over time as experiences accumulate. Such accounts may imply that the trust felt toward a partner should start at a relatively low amount from the initial periods of a relationship and then grow through the gathering of experiences with the partner. Our results, however, help to provide psychometric context and further clarity for prior descriptive data documenting people’s trust being at a heightened state regardless of whether they are involved in a new or established relationship.
Our results suggest that when in romantic relationships, people are trusting of their partners as a default, with feelings of trust not requiring time and experiences with a partner to develop from lower to higher levels. Several possible implications stem from this finding. For instance, only for certain individuals may trust in a romantic partner develop (e.g., grow and/or decline) because of relationship-specific experiences. The interplay of attachment orientation and trust development in particular might offer a promising avenue of inquiry (see ). Furthermore, although people may report high levels of trust for their partner across all stages of the relationship, the nature of how they express their trust to their partner behaviourally may vary in important ways. Future investigations into the nature of trust’s development may benefit from the inclusion of measurement instruments that target the behaviours of trust (e.g., ).
Measurement invariance and construction of trust
Construction convergences
Those in newly-formed relationships and those in long-term relationships were revealed to possess a broadly invariant structure of trust. The trust (i.e., general factor) of both those involved in newly-formed relationships and those involved in long-term relationships was most strongly defined by faith items rather than predictability or dependability items. Yet both samples seemed to possess strongly defined specific constructs of only predictability and dependability, while faith remained a weakly defined specific construct. Such patterns of sameness may speak to the (un)importance of the roles faith and evidence play in people’s trust throughout the relationship’s timeline.
Regardless of the stage of relationship development, faith may reflect the most important contents by which romantically-attached individuals construct their conceptualization of trust, whereas evidence may take on a more ancillary function. For instance, though evidence may not serve as a necessary precondition for the development of high trust, evidence may instead serve to calibrate trust (to a fitting amount). If trusting romantic partners at a relatively high level is the default, then evidence should only assume an appreciable role in trust’s development when perceived as contradictory to such certainty. As individuals confront diagnostic situations and issues of maturing interdependency, past experiences concerning trust may be susceptible to interpretation and/or dismissal (; ; ). While trust should suffer a breakage at the hands of a major betrayal (e.g., discovering a partner’s infidelity), a breakage may also result from a series of minor negative experiences that are trivial in isolation but coalesce over time into a bigger problem that can no longer be overlooked. Future research should endeavour to tease apart the roles of faith and evidence in trust’s development and junctures by which evidence may precipitate a breakage in one’s trust.
Construction divergences
Some of the non-invariance that emerged highlight potential areas of the trust construct that may differ as a function of relationship maturation. For the majority of the items responsible for scalar non-invariance, individuals in newly-formed relationships had substantially higher indicator intercepts compared to those in long-term relationships. Specifically, when latent factors are at baseline (i.e., zero), the expected values of item responses were higher for individuals in newly-formed relationships across a broader range of item contents than for those in long-term relationships. This non-invariability may speak to the audaciousness of trust in new relationships. Any uncertainties of the initial periods of the partnership are susceptible to being dampened by early romantic feelings and thus left unchallenged (; ; ). Having such confidence in the partner and relationship’s future may in part function to provide opportunity for the relationship to survive the uncertainty inherent in the early stages of development.
In contrast to those in newly-formed relationships, statements of faith like “I can rely on my partner to react in a positive way when I expose my weaknesses to him/her” and “When I share my problems with my partner, I know he/she will respond in a loving way even before I say anything” were revealed to define the general factor substantially more for those involved in long-term relationships. Certainty in being met with positivity and love from the partner within positions of vulnerability might be especially central to the meaning of trust for those involved in long-term relationships, while taking relatively lesser psychological significance in the trust of those involved in newly-formed relationships. Whereas uncertainty concerning a newfound partner can be overshadowed by early romantic love, people may later become more aware of the perils of disappointment and rejection as the relationship deepens. Thus, as romantically-attached individuals progress deeper into the relationship, confidence in the partner’s goodwill and responsiveness in the context of vulnerability may gain increasingly more importance and become a prominent theme in people’s trust.
Recommendations and precautions
Based on the results of our investigations we therefore offer the following recommendations and precautions for future work involving trust within romantic relationships. Given that researchers typically use the scale to calculate an overall score of trust for their analyses, we strongly recommend using our modified compilation of item contents (see Table 5). The creation and use of an overall composite score presume the presence of a generalized factor underlying the diversity of items involved—a measurement structure in need of formal support (; ). Via our psychometric efforts, we identified a compilation of items that demonstrated a coherent generalized factor, offering users an assessment that can reliably yield an unambiguous overall score of trust.
To the extent researchers are interested in assessing the specific theorized content domains of trust independently, we also recommend using our modified item compilation—namely the subsets of item contents contained therein (see Table 5). Our identified subsets of predictability- and dependability-based item contents have a capacity for tapping into meaningful residual specificity beyond generalized trust, providing downstream users with reliable (short) assessments of perceptions of a partner’s predictability and dependability. However, caution should be exercised in using our resulting subset of faith-based items to create a stand-alone index of faith. Unlike predictability and dependability, our findings indicated that the resulting subset of faith-based item contents broadly reflects the generalized construct and addresses little-to-no meaningful faith specificity. We recommend researchers perform a bifactor analysis to examine the extent to which their chosen compilation of items can, under the unique conditions of their data, reflect a coherent generalized construct (i.e., trust) as well as reflect meaningful residualized (sub)constructs (e.g., predictability, dependability, and faith). Doing so can garner useful insights into the appropriateness of a generalized factor of trust as an underlying structure of the user’s data and in turn offer them demonstrable support for the generation and use of composite scores for overall trust as well as for predictability, dependability, and/or faith.
Furthermore, although tests of measurement invariance showcased that our modified compilation of items broadly performed in a similar manner across individuals involved in different stages of relationship maturation (i.e., newly-formed vs. long-term relationships), some items were found to perform differently (i.e., metric and scalar non-invariance) across these groups. We therefore recommend researchers continue to examine the nature of the measurement invariance of items across individuals involved in various relationship stages, as well as across other grouping variables that may drive noninvariant assessments (and conceptualizations) of trust (e.g., culture, sexual orientation, relationship structure, etc.). Such forays into the psychometric (non)invariability of items—as well as the item contents of other instantiations of trust—will prove useful not only in uncovering a generalizable assessment of trust, but also for garnering deeper psychological insights into the (dis)similarities in people’s conceptualization of the intra-psychic entity (see ; ).
Limitations and final considerations
The findings from our research should be interpreted with some limitations in mind. Our studies were conducted with samples comprised predominantly of white, heterosexual individuals—potentially limiting the generalizability of our findings. Information on specific socio-demographic variables—such as disability status, household size, and income—was not collected in this research. Furthermore, our research contrasted trust across two samples of individuals involved in relationships of vastly different durations, yet the nature of trust’s development in relationships may require intensive longitudinal designs to capture fully.
Overall, the research we presented derived from a process of evaluating core assumptions of theory and measurement of trust and comparing them to existing evidence. Whereas trust scale has been employed heavily in subsequent research, no research has systematically evaluated the measurement model of the instrument nor the theoretical speculations regarding the growth of trust in relationships. Our research attempted to redress these important gaps—providing consistent and strong support for a modified version of original trust scale as well as shining some light on the nature of trust in the early and later stages of relationship development. We hope our research efforts can spur a research direction toward treating the concept of trust in romantic relationships as an important and interesting analytic outcome and topic of study in its own right.
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) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the Social Sciences and Humanities Research Council of Canada (SSHRC) (435-2016-0625).
Open research statement As part of IARR's encouragement of open research practices, the author(s) have provided the following information: This research was pre-registered. The aspects of the research that were pre-registered were the research ideas, data analysis plan, study design, and sampling plan. The registration was submitted to the Open Science Framework. The data used in the research are available. The data can be obtained at: https://osf.io/bvfe8/ or by emailing: [email protected]. The materials used in the research are available. The materials can be obtained at: https://osf.io/6gsrh/ or by emailing: [email protected].
Notes
1. The three-factor EFA solution, although evidenced to possess only somewhat reasonable fit to the data, was also closely examined in tandem with the four-factor EFA solution. See Supplementary Materials.
2. Our chosen criteria for distinguishing newly-formed relationships—romantically involved for 6 months or less—versus long-term relationships—romantically involved for 5 years or more—were loosely based on theories and stage models of relationship development (e.g., ; ; ). Theoretical models of relationship development typically center on the qualitative experiences separating earlier versus later stages of relationship maturation rather than the duration, given that relationships can progress at drastically different speeds. That is, time is thought not necessarily to be the primary mechanism of relationship development but instead, a proxy that can reflect and provide insight into the growth of a relationship. Our chosen criteria were therefore designed as pragmatic guidelines for recruiting participants involved in different stages of their relationship, rather than to claim specific turning points or developmental milestones of a romantic relationship.
3. A three-factor model’s first-order correlated-factor variant and second-order factor variant (i.e., the same three first-order factors defining a single second-order factor) are mathematically equivalent (). Therefore, of the two, we chose only to specify and interpret the correlated three-factor model variant.
4. Given the unbalanced sample sizes in the modeled groups of individuals involved in newly-formed relationships (n = 387) versus individuals involved in long-term relationships (n = 460), we performed a sensitivity analysis via implementing an approach similar to that described in . Over 100 iterations, we examined measurement invariance (i.e., configural, metric, and scalar invariance) between the (original) sample of n = 387 individuals involved in newly-formed relationships and a randomly drawn subsample of n = 387 respondents from the (original) sample of individuals involved in long-term relationships. Our tests of measurement invariance across all 100 iterations echoed the findings from the original study. Supplementary files—full analysis strategy, scripts, and tabularized and visualized output—are available in our Open Science Framework repository.
References
- Altman I., Taylor D. A. (1973). Social penetration: The development of interpersonal relationships. Holt, Rinehart & Winston.
- Boon S. D., Holmes J. G. (1991). The dynamics of interpersonal trust: Resolving uncertainty in the face of risk. In Hinde J. A., Groebel J. (Eds.), Cooperation and prosocial behavior (pp. 190–211). Cambridge University Press.
- Campbell L., Camanto O. J., Stanton S. C. E. (2025). The nature of trust in romantic relationships. In Rotenberg K. J., Petrocchi S., Levante A., Lecciso F. (Eds.), Handbook of trust and social psychology (pp. 93–108). Edward Elgar Publishing. https://doi.org/10.4337/9781803929415.00017
- Campbell L., Stanton S. C. (2019). Adult attachment and trust in romantic relationships. Current Opinion in Psychology, 25, 148–151. https://doi.org/10.1016/j.copsyc.2018.08.004
- Chen F. F. (2008). What happens if we compare chopsticks with forks? The impact of making inappropriate comparisons in cross-cultural research. Journal of Personality and Social Psychology, 95(5), 1005–1018. https://doi.org/10.1037/a0013193
- Cheung G. W., Lau R. S. (2012). A direct comparison approach for testing measurement invariance. Organizational Research Methods, 15(2), 167–198. https://doi.org/10.1177/1094428111421987
- Cheung G. W., Rensvold R. B. (1999). Testing factorial invariance across groups: A reconceptualization and proposed new method. Journal of Management, 25(1), 1–27. https://doi.org/10.1177/014920639902500101
- De Ayala R. J. (2009). The theory and practice of item response theory. Guilford Press.
- Enders C., Bandalos D. (2001). The relative performance of full information maximum likelihood estimation for missing data in structural equation models. Structural Equation Modeling: A Multidisciplinary Journal, 8(3), 430–457. https://doi.org/10.1207/S15328007SEM0803_5
- Fisher H. E. (1998). Lust, attraction, and attachment in mammalian reproduction. Human Nature, 9(1), 23–52. https://doi.org/10.1007/s12110-998-1010-5
- Fletcher G. J. O., Kerr P. S. G. (2010). Through the eyes of love: Reality and illusion in intimate relationships. Psychological Bulletin, 136(4), 627–658. https://doi.org/10.1037/a0019792
- Fletcher G. J. O., Simpson J. A., Thomas G. (2000). Ideals, perceptions, and evaluations in early relationship development. Journal of Personality and Social Psychology, 79(6), 933–940. https://doi.org/10.1037/0022-3514.79.6.933
- Hatfield E., Schmitz E., Cornelius J., Rapson R. L. (1988). Passionate love: How early does it begin? Journal of Psychology & Human Sexuality, 1(1), 35–51. https://doi.org/10.1300/J056v01n01_04
- Hazan C., Shaver P. (1987). Romantic love conceptualized as an attachment process. Journal of Personality and Social Psychology, 52(3), 511–524. https://doi.org/10.1037/0022-3514.52.3.511
- Holmes J. G., Rempel J. K. (1989). Trust in close relationships. In Hendrick C. (Ed.), Close relationships: Review of personality and social psychology (10, pp. 187–220). Sage Publications.
- Holzinger K. J., Swineford F. (1937). The Bi-factor method. Psychometrika, 2(1), 41–54. https://doi.org/10.1007/BF02287965
- Hu L., 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
- Johnson-George C., Swap W. C. (1982). Measurement of specific interpersonal trust: Construction and validation of a scale to assess trust in a specific other. Journal of Personality and Social Psychology, 43(6), 1306–1317. https://doi.org/10.1037/0022-3514.43.6.1306
- Jorgensen T. D., Kite B. A., Chen P.-Y., Short S. D. (2018). Permutation randomization methods for testing measurement equivalence and detecting differential item functioning in multiple-group confirmatory factor analysis. Psychological Methods, 23(4), 708–728. https://doi.org/10.1037/met0000152
- Jorgensen T. D., Pornprasertmanit S., Schoemann A. M., Rosseel Y., Miller P., Quick C., Garnier-Villarreal M., Selig J., Boulton A., Preacher K., Coffman D., Rhemtulla M., Robitzsch A., Enders C., Arslan R., Clinton B., Panko P., Merkle E., Chesnut S., Johnson A. R. (2022). semTools: Useful tools for structural equation modeling (0.5-6). https://cran.r-project.org/web/packages/semTools/index.html
- Kelley H. H., Holmes J. G., Kerr N. L., Reis H. T., Rusbult C. E., Van Lange P. A. M. (2001). An atlas of interpersonal situations (1st ed.). Cambridge University Press. https://doi.org/10.1017/CBO9780511499845
- Khojasteh J., Lo W.-J. (2015). Investigating the sensitivity of goodness-of-fit indices to detect measurement invariance in a bifactor model. Structural Equation Modeling: A Multidisciplinary Journal, 22(4), 531–541. https://doi.org/10.1080/10705511.2014.937791
- Kleinert T., Schiller B., Fischbacher U., Grigutsch L.-A., Koranyi N., Rothermund K., Heinrichs M. (2020). The trust game for couples (TGC): A new standardized paradigm to assess trust in romantic relationships. PLoS One, 15(3), Article e0230776. https://doi.org/10.1371/journal.pone.0230776
- Knapp M. L. (1978). Social intercourse: From greeting to goodbye. Allyn & Bacon.
- Larzelere R. E., Huston T. L. (1980). The dyadic trust scale: Toward understanding interpersonal trust in close relationships. Journal of Marriage and Family, 42(3), 595. https://doi.org/10.2307/351903
- Little R. J. A. (1988). A test of missing completely at random for multivariate data with missing values. Journal of the American Statistical Association, 83(404), 1198–1202. https://doi.org/10.1080/01621459.1988.10478722
- Little T. D. (2013). Longitudinal structural equation modeling. Guilford Press.
- Luchies L. B., Wieselquist J., Rusbult C. E., Kumashiro M., Eastwick P. W., Coolsen M. K., Finkel E. J. (2013). Trust and biased memory of transgressions in romantic relationships. Journal of Personality and Social Psychology, 104(4), 673–694. https://doi.org/10.1037/a0031054
- Messick S. (1995). Validity of psychological assessment: Validation of inferences from persons’ responses and performances as scientific inquiry into score meaning. American Psychologist, 50(9), 741–749. https://doi.org/10.1037/0003-066X.50.9.741
- Mikulincer M. (1998). Attachment working models and the sense of trust: An exploration of interaction goals and affect regulation. Journal of Personality and Social Psychology, 74(5), 1209–1224. https://doi.org/10.1037/0022-3514.74.5.1209
- Mikulincer M., Shaver P. R. (2007). Attachment in adulthood: Structure, dynamics, and change. Guilford Publications.
- Morin A. J. S., Arens A. K., Marsh H. W. (2016). A bifactor exploratory structural equation modeling framework for the identification of distinct sources of construct-relevant psychometric multidimensionality. Structural Equation Modeling: A Multidisciplinary Journal, 23(1), 116–139. https://doi.org/10.1080/10705511.2014.961800
- Murray S. L., Holmes J. G. (2010). Trust as motivational gatekeeper in adult romantic relationships. In Horowitz L. M., Strack S. (Eds.), Handbook of interpersonal psychology (1st ed., pp. 193–207). Wiley. https://doi.org/10.1002/9781118001868.ch12
- Murstein B. I. (1970). Stimulus. Value. Role: A theory of marital choice. Journal of Marriage and Family, 32(3), 465–481. https://doi.org/10.2307/350113
- Norona J. C., Welsh D. P., Olmstead S. B., Bliton C. F. (2017). The symbolic nature of trust in heterosexual adolescent romantic relationships. Archives of Sexual Behavior, 46(6), 1673–1684. https://doi.org/10.1007/s10508-017-0971-z
- Putnick D. L., Bornstein M. H. (2016). Measurement invariance conventions and reporting: The state of the art and future directions for psychological research. Developmental Review: Developmental Review, 41, 71–90. https://doi.org/10.1016/j.dr.2016.06.004
- R Core Team. (2023). R: A language and environment for statistical computing. R Foundation for Statistical Computing. https://www.R-project.org/
- Reis H. T. (2007). Steps toward the ripening of relationship science. Personal Relationships, 14(1), 1–23. https://doi.org/10.1111/j.1475-6811.2006.00139.x
- Reis H. T. (2012). Perceived partner responsiveness as an organizing theme for the study of relationships and well-being. In Campbell L., Loving T. J. (Eds.), Interdisciplinary research on close relationships: The case for integration (pp. 27–52). American Psychological Association. https://doi.org/10.1037/13486-002
- Rempel J. K., Holmes J. G., Zanna M. P. (1985). Trust in close relationships. Journal of Personality and Social Psychology, 49(1), 95–112. https://doi.org/10.1037/0022-3514.49.1.95
- Revelle W. (2024). psych: Procedures for psychological, psychometric, and personality research (2.4.1). https://cran.r-project.org/web/packages/psych/index.html
- Rosseel Y. (2012). lavaan: An R package for structural equation modeling. Journal of Statistical Software, 48(2). https://doi.org/10.18637/jss.v048.i02
- Sakaluk J. K. (2019). Expanding statistical frontiers in sexual science: Taxometric, invariance, and equivalence testing. The Journal of Sex Research, 56(4–5), 475–510. https://doi.org/10.1080/00224499.2019.1568377
- Sakaluk J. K., Fisher A. N. (2019). Measurement memo I: Updated practices in psychological measurement for sexual scientists. The Canadian Journal of Human Sexuality, 28(2), 84–92. https://doi.org/10.3138/cjhs.2019-0018
- Sakaluk J. K., Quinn-Nilas C., Fisher A. N., Leshner C. E., Huber E., Wood J. R. (2021). Sameness and difference in psychological research on consensually non-monogamous relationships: The need for invariance and equivalence testing. Archives of Sexual Behavior, 50(4), 1341–1365. https://doi.org/10.1007/s10508-020-01794-9
- Schlomer G. L., Bauman S., Card N. A. (2010). Best practices for missing data management in counseling psychology. Journal of Counseling Psychology, 57(1), 1–10. https://doi.org/10.1037/a0018082
- Simpson J. A. (2007). Psychological foundations of trust. Current Directions in Psychological Science, 16(5), 264–268. https://doi.org/10.1111/j.1467-8721.2007.00517.x
- Yoon M., Lai M. H. (2018). Testing factorial invariance with unbalanced samples. Structural Equation Modeling: A Multidisciplinary Journal, 25(2), 201–213. https://doi.org/10.1080/10705511.2017.1387859
- Zumbo B. D. (2023). A dialectic on validity: Explanation-focused and the many ways of being human. International Journal of Assessment Tools in Education, 10(Special Issue), 1–96. https://doi.org/10.21449/ijate.1406304