Getting to Know You

Trust and Trust Development

Currently, the most widely accepted definition of trust is that of Rousseau, Sitkin, Burt, and Camerer in which trust is described as “a psychological state comprising the intention to accept vulnerability based upon positive expectations of the intentions or behavior of another” (1998: 395). This definition combines two important perspectives that have been the bases of previous conceptualizations of trust, a willingness to be vulnerable (e.g., Mayer et al., 1995) and positive expectations of others (e.g., Lewicki & Bunker, 1996). In recent years, organizational scholars have increasingly focused their attention on trust as an intention to behave consisting of two dimensions: reliance on others and disclosure of information to others (Gillespie, 2003). In contrast to a traditional behavioral view of trust as a cooperative action (e.g., Deutsch, 1958), Gillespie’s model defines trust as an intention to engage in behavior that increases vulnerability. In comparison to previous distinctions between trust and trust behavior (e.g., Mayer et al.), Gillespie’s conceptualization is closer to trust as it represents a willingness to take a risk rather than actual risk-taking behavior. The inclusion of more than one dimension builds on previous theoretical work that emphasizes the contextual nature of trust (e.g., Hardin, 1996) and allows researchers to distinguish between two trust behaviors that are common to workplace relationships. The conceptualization of trust as reliance and disclosure intentions has received considerable support in the literature (e.g., Dietz & Den Hartog, 2006; McEvily & Tortoriello, 2011) and offers a means of determining what an individual might trust another party to do and to what extent. Gillespie’s conceptualization of an intention to engage in trust behavior builds on the Rousseau et al. definition of trust as a psychological state and provides a method of operationally separating trust from its antecedents. The relatively new operationalization of trust as reliance and disclosure increases the need for research to examine how trust intentions develop and how they are related to other common trust variables.

Newcomer Trust Development

Socialization is the process through which individuals adapt to a new role or job in an organization (Chao, O’Leary-Kelly, Wolf, Klein, & Gardner, 1994). When individuals join a new organization, they strive to reduce feelings of uncertainty by familiarizing themselves with their new task and social environment (Simosi, 2010). Existing research illustrates that the initial socialization of newcomers lasts approximately 2 to 3 months (G. Chen, 2005), during which time employees adapt to their new positions, are integrated into the organization, and build relationships with their colleagues (D. Chan & Schmitt, 2000). The importance of developing effective working relationships with colleagues has been a key theme in socialization research, and organizational theorists are increasingly recognizing the importance of newcomer perceptions of relationships at work as a key mechanism for socialization success (Allen & Rhoades Shanock, 2013). The psychological contract literature has also highlighted the importance of this period of socialization in informing new recruits’ perceptions of what they can expect from their organization and their colleagues (De Vos, Buyens, & Schalk, 2003). From the moment that they join an organization, new hires must interact and cooperate with their new peers and other organizational members. Before these interactions can take place, it is necessary for individuals to make a judgment regarding their willingness to be vulnerable to their colleagues.

Li’s (2012) outline of the contexts in which trust is most critical describes situations in which uncertainty and vulnerability are high, where unmet expectations represent a significant risk, and where a level of interdependence is expected. Socialization represents a period of an employee’s working life that is likely to incorporate all of these characteristics. In socialization programs where large groups of newcomers join an organization together, levels of social uncertainty and risk may be exacerbated by the general lack of work and organizational experience in the group and the pressure to stand out from the crowd and make a positive impression. Uncertainty reduction theory (Berger & Calabrese, 1975) suggests that this will motivate individuals to form positive work relationships. In this formal socialization context, the organization strives to provide a forum for these positive interactions with plenty of opportunities for shared activity. Decisions to rely on colleagues or share information with them will be made continually throughout the life span of working relationships, but the period of socialization offers a unique context from which to gain an improved understanding of trust development.

The prevailing initial trust theories (McKnight et al., 1998; Meyerson et al., 1996) suggest that newcomer trust may not have a zero baseline. The socialization literature agrees that early newcomer perceptions develop rapidly (Saks & Ashforth, 1997) and are strong predictors of attitudes and outcomes at a later date (Bauer & Green, 1994). In the socialization context, organizations strive to provide newcomers with a positive socialization experience and a positive forum for newcomers to interact with each other, thus encouraging trusting behaviors and the development of stronger peer relationships (Homans, 1950). Development of trust in a relationship is often proposed to be dependent on trustor experiences during a history of interactions with the trustee built up over a period of time. If trust begins at a nonzero baseline, it is expected that the positive experience of a trusting relationship with coworkers will motivate employees to interact with their new organizational environment with the intention of building social resources, thus reinforcing levels of trust in colleagues (Fredrickson, 2001). As individuals perceive themselves engaging in trust behaviors, their positive expectations of coworkers will be strengthened (Bem, 1972) and their trust levels will increase in a self-reinforcing spiral. This potential trust trajectory is in line with the proposed feedback loop in Mayer and colleagues’ (1995) model of interpersonal trust and Zand’s (1972) argument that trust is a self-reinforcing phenomenon where trust leads to more trust. Theoretically then it can be expected that trust development during socialization should develop in an upward trajectory.

Longitudinal field research on coworkers has tended to employ a time-lagged design (e.g., Colquitt et al., 2011) rather than a truly longitudinal repeated measure of variables across time periods. Furthermore, previous longitudinal research investigating trust has focused predominantly on trust at a team level of analysis. Reciprocal trust between teams has been demonstrated as a longitudinal process where team trust is continually revised on the basis of up-to-date trustworthiness perceptions following interaction between the teams (Serva, Fuller, & Mayer, 2005). Further empirical evidence from team research also suggests that trust tends to increase over time (Wilson, Straus, & McEvily, 2006), that trust grows more complex and multidimensional as relationships develop (Webber, 2008), and that early trusting beliefs have a significant impact on beliefs 2 months later (Crisp & Jarvenpaa, 2013). In general, team-level results are in line with the proposition that the mechanism through which trust develops is positive exposure to the other party. During their first months in an organization, continued exposure to the common learning experiences and positive social support typical of institutionalized socialization gradually decreases newcomer uncertainty in predicting the behavior of their peers and increases feelings of social comfort (Kim, Cable, & Kim, 2005). This social environment provides the ideal forum for coworkers to engage in positive interaction. Therefore, the following is hypothesized:

Individual Differences in Newcomer Trust Development

In addition to improving our understanding of general patterns of trust development between new colleagues, insight into the individual differences that exist in how trust behaviors develop is vital to achieving a more nuanced picture of relationship processes. Trust has often been conceptualized by psychologists as a dispositional characteristic or personality trait of the trustor that is not specific to any referent or context. While this approach has been challenged in recent years, the Mayer et al. (1995) positioning of propensity to trust as a separate but influential variable in trust decisions has revived the popularity of the construct. Although trust as a trait variable is likely to have a distal influence on trust processes in general, propensity to trust is theoretically positioned as an antecedent to trust and an influence on the impact of trust cues (Mayer et al.). Mayer and colleagues define propensity to trust as a “general willingness to trust others” (1995: 715) and suggest that propensity to trust may provide trustors with the ability to make a leap of faith. An individual’s propensity to trust provides them with a generalized expectation about the reliability of others (Rotter, 1971) that is proposed to act as a filter through which the actions of other people are interpreted (Govier, 1994).

Colquitt et al. (2007) reported the correlation between propensity to trust and trust as significant but small, suggesting that the relationship may be context dependent. For example, it has been suggested that propensity to trust is more critical to trust formation in ambiguous situations (Gill, Boies, Finegan, & McNally, 2005) and those that lack personal trust cues (Grant & Sumanth, 2009). Furthermore, as communication frequency in a dyad increases, propensity to trust becomes less important (Becerra & Gupta, 2003). Accordingly, propensity to trust can be considered more central to the formation of trust if the trustor does not have sufficient information about the situation or trustee to form an accurate expectation or belief. This makes propensity to trust uniquely relevant to new trust relationships and in particular newcomers in a socialization context. Previous research tended to employ laboratory or time-lagged cross-sectional designs, making it impossible to track the impact of propensity to trust over time within a relationship. We propose that in a socialization context, where no preceding relationship history exists between newcomers, propensity to trust will affect initial trust levels as little information is available to the trustor about coworker traits or behavior. As this new information becomes available to the new recruits, propensity to trust should decrease in significance as a trust source. Accordingly, we hypothesize the following:

Cues for Trust Development

A central question in trust research has been which personal, relational, or situational factors combine to allow trustors to make a decision to engage in trusting behavior. Recent debate about the relative importance of particular trust cues (e.g., Dietz, 2011) echoes previous calls (Schoorman, Mayer, & Davis, 2007) for researchers to establish the time frame in which particular predictors of trust exert their influence. A wide array of micro- and macrocues have been proposed as antecedents to trust. For instance, the importance of the three-factor model of trustworthiness (competence, benevolence, and integrity; Mayer et al., 1995) is widely accepted, as is the importance of repeated positive interaction over time (Lewicki & Bunker, 1996). While this information is likely to play a role in trust development during socialization, other trust cues may be more specific to the context of new work relationships.

In their discussion of the formation of swift trust, Meyerson et al. (1996) posit the importance of the reputation associated with occupational roles as an important cue when interacting with unknown trustees. Social identity theorists suggest that newcomers form a stereotype of prototypical group members that guides their expectations for the behavior of individuals associated with that prototype (Hogg, 2000). Recent theoretical work has supported the notion that groups (e.g., members of a particular occupation) can carry certain trust-related attributes, such as benevolence and competence, as part of their stereotype, affecting the way that strangers feel and act towards members of that group (Cuddy, Glick, & Beninger, 2011). Trust predicated on the role occupied by the trustee is based on our perception of the barriers to obtaining that role, the adequacy of the education and training needed to fulfill that role, and the social mechanisms that govern adherence to role typical behavior (Kramer & Lewicki, 2010). Empirically, role-based trust has been shown to be related to individual trust in supervisors at the early stages of a relationship (Shamir & Lapidot, 2003).

In addition, McKnight et al. (1998) argue that initial formation of trusting beliefs is informed by cognitive and institutional cues, including the categorization of the other party and perception of safeguards inherent in organizational structures. The construct of rule-based trust refers to an employee’s sense that the organizational system supports trust between coworkers through the empowerment of trust behaviors and the constraint of untrustworthy acts (Mollering, 2012). The simultaneous empowerment and constraint of behavior is facilitated by the existence of injunctive norms that signal to newcomers the behavior expected of good employees and descriptive norms that communicate the typical behavior in a specific context (Kramer & Lewicki, 2010). The effect of this perception may be even more salient when organizational membership is shared, therefore fostering perceptions of similarity between coworkers (Zucker, 1986).

During socialization, individuals form a social identity and define themselves and coworkers as members of an in-group or out-group (Ashforth & Mael, 1989). As newcomers develop a shared sense of identity, they form normative schema to guide their perception of the behavior and values of other newcomers as in-group members (Livingstone, Haslam, Postmes, & Jetten, 2011). Brewer (1996) argues that the depersonalization inherent in this process provides a basis for depersonalized trust in in-group members and negates the need for personal knowledge in assessing the risk of initial interactions. In line with this, it has been proposed that feelings of identification play a key role in initial trust building (Williams, 2001).

Although a variety of trust cues are available to an individual, it may not be possible for trustors to simultaneously attend to all of this available information. Using the principle of bounded rationality, Bijlsma and Koopman (2003) argue that the entire array of complex antecedents of trust is unlikely to be considered by every trustor in every trust decision. Indeed, it is more probable, given realistic information processing and time constraints, that individuals choose to attend to a finite number of cues at any one time. Unfortunately, little empirical research exists to provide insight into the relative importance of different trust cues. In fact, much of the trust literature has failed to separate common trust antecedents from trust decisions, using cues such as trustworthiness as a proxy for measuring trust.

One theoretical perspective that offers insight into when certain cues might be important was put forward by Kramer and Lewicki (2010). Drawing on previous work, they propose that the concept of presumptive trust describes positive expectations of others founded upon factors such as perceptions of the rules embedded in a shared organizational environment (rule-based trust), perceptions of the other’s role (role-based trust), and identification with the trustee. This form of trust is thought to be important when the trustor has little information about a trustee. In the context of socialization, presumptive trust cues are likely to be important early in new recruits’ relationships with their coworkers. G. Chen and Klimoski (2003) describe the first few days of socialization as an anticipatory phase where expectations of coworkers are formed quickly on the basis of anticipated behavior. At this stage, individuals have access to information about the job title and organizational membership of their colleague but little evidence about any personal characteristics on which they could base a trust decision. In light of the lack of personal information, cooperative behavior relies on the presumption of trust (Kramer & Lewicki), freeing up cognitive resources to allow the new recruits to perform effectively in their new environment (Mayer & Gavin, 2005). In essence, presumptive trust during socialization draws on newcomers’ schema of their new organization and the delineation of roles and groups within it. The social categorization processes underlying socialization suggest that the desire to reduce uncertainty in the new work environment should drive the use of heuristics to guide initial interactions and allow for the rapid building of coworker relationships.

This research draws on the theoretical work of Kramer and Lewicki (2010) and Mayer et al. (1995) to distinguish between presumptive trust cues and personal trust cues. Presumptive trust cues refer to social and environmental information, including role-based trust, rule-based trust, and identification. Personal trust cues describe those attributes of the trustee described by the trustworthiness dimensions of competence, benevolence, and integrity. As a relationship develops and more personal trust cues are available, reliance on presumptive trust is thought to diminish. McKnight and Chervany suggest that the initial trust phase ends once “parties gain verifiable information by first hand interactional or transactional experience” (2006: 29). In this way, presumptive trust cues should function in a similar manner to trustor dispositional factors, such as propensity to trust—important at the start but with a diminishing impact over time. What is less clear from the literature is how the shift from presumptive to personal cues takes place and when. Although evidence from longitudinal trust research is limited, it has been demonstrated that initial trust is predicted by unique antecedents to those that affect more mature relationships (e.g., Jarvenpaa, Knoll, & Leidner, 1998). This study proposes that as newcomers are socialized into their new organization and as they gain access to more personal information and insights about their newcomer peers, their willingness to engage in trusting behavior with these colleagues will be based increasingly on personal rather than presumptive trust cues.

Participants

Participants in this study were new employees in the Irish practice of a Big 4 international accountancy and consultancy firm. All participants are knowledge workers, a sector of the workforce that is both growing (G. Chen & Klimoski, 2003) and central to the economy (Kessels, 2004). The average age of the sample was 22.26 years (SD = 1.23), and participants were 54.9% female. All participants (N = 193) had completed their bachelor’s degree with 36.3% completing master’s-level courses. New recruits all began work for the organization on the same day and were assigned to working groups within an hour of joining the organization.

Research Design

This study was designed to collect data longitudinally over four waves of data collection. Previous longitudinal trust studies have examined trust using between two and four data points and with time lags ranging from 1 week to 6 months (e.g., Colquitt et al., 2011; Jarvenpaa et al., 1998; Serva et al., 2005; Shamir & Lapidot, 2003; Webber, 2008; Wilson et al., 2006). In this study, all data were collected within the first 3 months of participants’ employment with the organization. This time period was selected as it (a) represents a theoretically acceptable time frame for the initial socialization of new employees (G. Chen, 2005) and (b) allows new recruits sufficient time to interact with their coworkers and develop more stable working relationships where trust behaviors would have time and opportunity to occur. Time intervals were spaced unevenly over the 3 months with data collected in Week 1, Week 4, Week 10, and Week 12. Unequal intervals were logistically necessary because of a series of work experience sessions organized for new recruits that meant that they were unavailable to the researcher at certain times. Although this design limits the availability of information about the middle period of socialization, it also represents an advantage of the study as longitudinal research that collects data in waves that are spread away from the midpoint of the data collection period offer greater reliability and precision in their measurement of change (Singer & Willett, 2003).

The socialization context in the organization can be described as a collective, formal, fixed, and sequential learning environment (Van Maanen & Schein, 1979) where newcomers attend scheduled group socialization activities, are clearly identifiable as new recruits, and are provided with a detailed timetable of their expected socialization experience. Within the sample, participants were assigned to a smaller working group by the organization. The smaller working groups consisted of up to 30 trainee accountants and were formed as a result of logistical reasons concerning the structuring of the socialization period and the training facilities available. These smaller working groups formed the basis of the referent in this study. At the end of the socialization period, newcomers were assigned to new project teams with more experienced workers within the organization. Because of the nature of their work, it is expected that the structure of these project teams would change regularly on the basis of client needs as is typical of the multiplex organizational structure of large modern professional service firms (Greenwood, Morris, Fairclough, & Boussebaa, 2010). Understanding the development of trust in collective groups is important because of the increasingly team-based structure of modern organizations (Fulmer & Gelfand, 2012). In professional service firms in particular, groups and project teams are a typical feature of work, and it is through group cooperation that professional service firms generate knowledge and solutions for their clients. Accordingly, in this context the coworker group was considered the most appropriate referent for the study of trust development. The use of a multiple coworker referent for trust is in line with previous research in the area (e.g., Colquitt et al., 2011; Schaubroeck, Peng, & Hannah, 2013). Additionally, as data collection began on the 1st day of employment, it was not possible to identify stable coworker dyads to act as trust referents.

Participants were recruited to the study during induction training on their 1st day of work. Participation in the study was voluntary and unincentivized, and the objectives and longitudinal nature of the research was explained fully at Time 1. Out of 198 potential respondents, 98.5% chose to participate. Questionnaires were completed by participants in the presence of the first author during working hours at training sessions, and participants were allowed to withdraw from the study at any time. Attrition of participants is a common issue in longitudinal research; however, attrition rates in this study were low. Following Ployhart and Vandenberg’s (2010) advice that a minimum of three data points is required to demonstrate nonlinear growth over time, participants who responded on less than three occasions were excluded from the study, resulting in a final sample of 193 (and an overall response rate of 97.5%). During hypothesis testing, missing data were estimated at random using a full information maximum likelihood method (Enders & Bandalos, 2001) in order to avoid the introduction of bias (Ployhart & Vandenberg).

Measurement

All variables were assessed at each of the four time points with the exception of propensity to trust, which was measured only at Time 1. The order of scales within the survey was randomized between time points. Participants were instructed to answer questions using coworkers in their socialization group as the referent. All items were assessed using a Likert-type response scale ranging from 1 to 7.

Confirmatory Factor Analysis

Confirmatory factor analysis was performed to determine the best model fit for the data; to compare alternative models, the chi-square difference test was used (Bentler & Bonett, 1980). First, we examined the factor structure of Gillespie’s (2003) Behavioral Trust Inventory. Analysis revealed that a two-factor structure where reliance and disclosure are treated as separate dimensions represents a significantly better fit for the data, χ²(32) = 61.21, p < .05, CFI = .98, TLI = .97, RMSEA = .07, than a one-factor model, χ²(33) = 238.08, p < .001, CFI = .86, TLI = .81, RMSEA = .18. Next we compared a nine-factor target model, where each of the subfactors of trust and trust cues represent distinct factors, to alternative models where some or all of these factors are collapsed. Results in Table 1 show that the target model achieved superior fit to each of the alternative models. Similar results were found across alternative waves of data collection. Because of the self-report nature of our measures, common method variance was assessed using the single method factor procedure (Podsakoff, MacKenzie, & Podsakoff, 2012), which indicated that this was not a problematic source of bias in these data. Table 2 displays the descriptive statistics for all variables.

OPEN IN VIEWER

Table 1

Goodness-of-Fit Indices and Difference Tests for Confirmatory Factor Analysis Model Comparisons

Model	χ2	df	CFI	TLI	RMSEA	Δχ2	Δdf
1. 9 Factor (Target Model)	1,846.43***	1180	.92	.91	.05	—	—
2. 8 Factor (Trust Intentions collapsed)	2,040.45***	1188	.90	.89	.06	194.02***	8
3. 6 Factor (Trust Intentions & Trustworthiness collapsed)	2,435.53***	1201	.85	.84	.07	395.08***	13
4. 3 Factor (Trust & Trust Cues collapsed)	3,536.39***	1213	.72	.70	.10	1,100.86***	12
5. 1 Factor (All scales collapsed)	4,089.15***	1216	.65	.63	.11	552.76***	3

Note: CFI = comparative fit index; TLI = Tucker Lewis index; RMSEA = root mean square error of approximation.

***

p < .001.

OPEN IN VIEWER

Table 2

Correlations, Means, and Standard Deviations

Measure	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21	22	23	24	25	26	27	28	29	30	31	32	33	M	SD
1. T1Reliance	(.85)																																	4.85	0.88
2. T1Disclosure	.38***	(.85)																																4.11	1.05
3. T1Competence	.46***	.34***	(.93)																															5.27	0.9
4. T1Benevolence	.32***	.45***	.64***	(.85)																														4.46	0.88
5. T1Integrity	.42***	.38***	.74***	.72***	(.84)																													4.86	0.82
6. T1Group Identification	.30***	.25***	.42***	.38***	.39***	(.75)																												5.11	0.95
7. T1Role based	.31***	.30***	.34***	.47***	.45***	.26***	(.89)																											4.91	0.78
8. T1Rule based	.20*	.25***	.31***	.25***	.31***	.32***	.38***	(.92)																										5.83	0.91
9. T2Reliance	.39***	.26***	.38*8*	.29***	.27***	.26***	.30***	.24***	(.86)																									5.26	0.76
10. T2Disclosure	.06	.36***	.14	.20*	.14	.36***	.18*	.12	.37***	(.85)																								4.61	0.96
11. T2Competence	.32***	.22*	.48***	.35***	.35***	.22**	.31***	.24***	.62***	.35***	(.89)																							5.46	0.73
12. T2Benevolence	.18*	.24***	.29***	.47***	.36***	.24***	.34***	.18*	.45***	.46***	.57***	(.85)																						4.66	0.85
13. T2Integrity	.25***	.24***	.40***	.42***	.44***	.24***	.32***	.27***	.38***	.31***	.64***	.66***	(.79)																					5.03	0.71
14. T2Group Identification	.21*	.15*	.27***	.20*	.21*	.15*	.12	.20*	.31***	.13	.38***	.37***	.47***	(.77)																				5.14	0.91
15. T2Role based	.23**	.20*	.37***	.44***	.34***	.20*	.39***	.31***	.36***	.20*	.55***	.62***	.66***	.44***	(.94)																			5.09	0.77
16. T2Rule based	.14	.25***	.35***	.34***	.30***	.25***	.29***	.44***	.32***	.21*	.45***	.43***	.55***	.36***	.57***	(.95)																		5.74	0.88
17. T3Reliance	.31***	.25***	.18*	.13	.16*	.25***	.24***	.16*	.52***	.32***	.46***	.40***	.42***	.18*	.34***	.20*	(.91)																	5.29	0.84
18. T3Disclosure	−.01	.20*	−.02	.10	−.01	.20*	.12	.10	.21*	.45***	.30***	.36***	.25***	.03	.23**	.10	.48***	(.89)																4.66	1.04
19. T3Competence	.19*	.07	.31***	.20*	.21*	.07	.30***	.18*	.36***	.31***	.59***	.39***	.49***	.28***	.40***	.34***	.57***	.42***	(.92)															5.49	0.75
20. T3Benevolence	.02	.08	.11	.33***	.20*	.08	.28***	.09	.21*	.30***	.32***	.51***	.39***	.10	.30***	.22**	.42***	.56***	.60***	(.86)														4.72	0.88
21. T3Integrity	.09	.08	.13	.16*	.22**	.08	.27***	.10	.21*	.24***	.33***	.38***	.50***	.10	.35***	.26***	.56***	.48***	.68***	.64***	(.79)													5.04	0.71
22. T3Group Identification	.08	.12	.17*	.16*	.15*	.12	.27***	.10	.22**	.23**	.30***	.31***	.35***	.45***	.32***	.15*	.37***	.36***	.49***	.47***	.46***	(.83)												5.12	0.92
23. T3Role based	.10	.13	.19*	.28***	.22**	.13	.43***	.27***	.27***	.20*	.35***	.38***	.42***	.11	.53***	.42***	.40***	.41***	.53***	.56***	.63***	.47***	(.94)											5.02	0.76
24. T3Rule based	.23**	.33***	.19*	.15*	.21*	.33***	.39***	.34***	.32***	.24***	.28***	.33***	.38***	.25***	.37***	.53***	.36***	.21*	.37***	.29***	.43***	.36***	.59***	(.96)										5.53	0.99
25. T4Reliance	.15*	.19*	.13	.14	.14	.19*	.28***	.17*	.42***	.27***	.40***	.36***	.36***	.17*	.31***	.32***	.66***	.41***	.55***	.48***	.56***	.35***	.50***	.47***	(.92)									5.49	0.77
26. T4Disclosure	−.01	.32***	.09	.15*	.07	.32***	.15*	.16*	.23**	.45***	.26***	.38***	.26***	.15*	.19*	.23**	.38***	.58***	.37***	.40***	.36***	.28***	.41***	.38***	.64***	(.92)								4.97	1.08
27. T4Competence	.13	.11	.33***	.21*	.24***	.11	.24***	.18*	.34***	.28***	.53***	.35***	.39***	.23**	.30***	.30***	.43***	.31***	.62***	.45***	.46***	.36***	.44***	.33***	.67***	.51***	(.95)							5.65	0.78
28. T4Benevolence	.01	.12	.11	.22**	.10	.12	.25***	.07	.22**	.24***	.29***	.43***	.36***	.19*	.29***	.27***	.34***	.37***	.45***	.59***	.47***	.33***	.51***	.35***	.59***	.62***	.71***	(.92)						5.09	0.97
29. T4Integrity	.09	.14	.21*	.26***	.25***	.14	.30***	.26***	.20*	.22**	.33***	.35***	.44***	.17*	.38***	.42***	.38***	.27***	.49***	.48***	.54***	.37***	.60***	.49***	.59***	.46***	.69***	.71***	(.80)					5.11	0.83
30. T4Group Identification	.12	.15*	.16*	.18*	.20*	.15*	.28***	.18*	.25***	.26***	.31***	.36***	.38***	.46***	.33***	.20*	.43***	.33***	.47***	.40***	.42***	.62***	.45***	.40***	.60***	.45***	.57***	.50***	.57***	(.88)				5.31	1.05
31. T4Role based	.16*	.17*	.29***	.33***	.27***	.17*	.40***	.27***	.29***	.24***	.36***	.44***	.47***	.25***	.52***	.45***	.39***	.35***	.48***	.47***	.51***	.41***	.71***	.54***	.54***	.46***	.52***	.62***	.70***	.56***	(.96)			5.19	0.91
32. T4Rule based	.16*	.23**	.26***	.16*	.21*	.23**	.23**	.34***	.34***	.15*	.28***	.30***	.30***	.20*	.37***	.41***	.34***	.25***	.31***	.30***	.37***	.28***	49***	.60***	.48***	.39***	.36***	.38***	.50***	.34***	.51***	(.93)		5.32	1.01
33. Propensity to Trust	.19*	.17*	.27***	.24***	.27***	.04	.43***	.31***	.21*	.09	.29***	.21*	.23**	.13	.28***	.15*	.20*	.12	.24***	.25***	.17*	.14	.17*	.08	.20*	.15*	.16*	.19*	.14*	.17*	.08	.18*	(.71)	4.51	0.67

Note: N = 193. Coefficient alpha reliability estimates are shown in parentheses.

*

p < .05.

**

p < .01.

***

p < .001.

MEI

We examined the MEI, the stability of the conceptual framework participants use to respond to survey items, before proceeding with longitudinal analysis. Reliance and disclosure were tested using three steps, with increasingly stringent constraints, following the guidelines set out by Vandenberg and Lance (2000). The likelihood ratio test (Bollen, 1989) and changes of .01 or greater in CFI (Cheung & Rensvold, 2002) were used to compare models. Results of the analyses are presented in Table 3.

OPEN IN VIEWER

Table 3

Measurement Invariance Tests

Model	χ2	df	CFI	TLI	RMSEA	Δχ2	Δdf
1. Configural Invariance	1,188.92***	572	.95	.94	.07
2. Metric Invariance	1,233.83***	596	.95	.94	.07	44.91**	24
3. Scalar Invariance	1,390.36***	620	.94	.93	.08	156.53***	24
4. Partial Scalar Invariance	1,267.96***	611	.95	.94	.07	34.13**	15

Note: CFI = comparative fit index; TLI = Tucker Lewis index; RMSEA = root mean square error of approximation.

**

p < .01.

***

p < .001.

In the first step, we assessed configural invariance to check if the two-factor structure generalizes over time. Acceptable model fit, χ²(572) = 1,188.92, p < .001, CFI = .95, TLI = .94, RMSEA = .07, indicates that two factors exist in every data wave and that the same items load onto reliance and disclosure at each time point. The second step involves a test of metric invariance where item factor loadings are constrained to be equal across measurement waves. Goodness-of-fit indices, χ²(596) = 1,233.83, p < .001, CFI = .95, TLI = .94, RMSEA = .07, were compared to the configural model; no change was observed in CFI, and the decrease in model fit indicated by the chi-square difference was not significant at the p < .001 level. Thus metric invariance was established.

Finally, we assessed scalar invariance by constraining the item intercepts to be equal across data waves. A significant chi-square difference and change in CFI, χ²(620) = 1,390.36, p < .001, CFI = .94, TLI = .93, RMSEA = .08, indicate the existence of some level of noninvariance of item intercepts. Following the recommendations of Byrne, Shavelson, and Muthén (1989), we investigated further in an attempt to achieve partial scalar invariance. We examined the tau parameters of the metric invariance model and constructed 95% confidence intervals to identify items that functioned differentially across data waves and to minimize the risk of false detection (Yoon & Millsap, 2007). Using these confidence intervals, we identified three items (two disclosure and one reliance) that showed higher levels of variance over time and removed the constraints on these items one by one. The items that displayed the highest levels of variance were “Rely on your group members to represent your work accurately to others,” “Discuss how you honestly feel about your work, even negative feelings and frustration,” and “Discuss work-related problems or difficulties that could potentially be used to disadvantage you.” Allowing the intercepts to vary on these three items resulted in a model that showed no difference in CFI and a chi-square change that is not significant at the p < .001 level—χ²(611) = 1,267.96, CFI = .95, TLI = .94, RMSEA = .07. As our partial scalar invariance model meets previously recommended criteria (Byrne et al.; Cheung & Lau, 2012), we concluded that the concepts of reliance and disclosure have sufficiently stable properties over time and were suitable for longitudinal analysis.

Hypothesis Testing

We used a latent growth modeling approach to test our hypotheses. First, longitudinal change in the two subfactors of trust, reliance and disclosure, were modeled using univariate latent growth modeling. In line with accepted practices in the management literature, four latent growth models were fit to each variable to assess the structure of the factor residuals and to determine whether change in trust is linear. Linear change was modeled by fixing the change factor loadings in the model equal to 0, 1, 3, and 3.66, in line with recommendations of Lance, Meade, and Williamson (2000). The uneven increases in factor loadings are calculated and rescaled to reflect the unevenly spaced measurement occasions used in data collection where an increase of 1 represents an interval of 3 weeks. Nonlinear or optimal change was modeled by fixing the first two change factor loadings (to 0 and 1 as before) and leaving the second two free to be estimated by MPlus. For both linear and optimal models, we compared nested models where the residual variances were freely estimated (heteroscedastic) to those where the residuals were constrained to be equal (homoscedastic; Willett & Sayer, 1994). Table 4 displays the results of the eight univariate models. Results indicated that the optimal change function significantly improved model fit in comparison to the linear model for both reliance and disclosure. Constraining the residual variances resulted in a poorer model fit for reliance. Consequently, an optimal heteroscedastic model, χ²(3) = 10.41, p < .05, CFI = .96, TLI = .93, RMSEA = .11, was accepted as the most accurate representation of change in reliance over time. When homoscedastic residuals were added to the optimal disclosure model, a small but nonsignificant increase in the chi-square index was observed. As constraining the residual variances provides a more parsimonious model structure, an optimal homoscedastic model, χ²(6) = 14.76, p < .05, CFI = .95, TLI = .94, RMSEA = .09, of disclosure was retained.

OPEN IN VIEWER

Table 4

Goodness-of-Fit Indices and Difference Tests for Univariate Latent Growth Model Comparisons

Model	χ2	df	CFI	TLI	RMSEA	Δχ2	Δdf
1. Reliance Linear Heteroscedastic	36.13***	5	.84	.81	.18
2. Reliance Linear Homoscedastic	60.12***	8	.74	.80	.18
Model 1 vs. Model 2						23.99***	3
3. Reliance Optimal Heteroscedastic a	10.41*	3	.96	.93	.11
Model 1 vs. Model 3						25.72***	2
4. Reliance Optimal Homoscedastic	19.04**	6	.93	.93	.11
Model 3 vs. Model 4						8.63*	3
5. Disclosure Linear Heteroscedastic	29.00***	5	.85	.82	.16
6. Disclosure Linear Homoscedastic	38.07***	8	.81	.86	.14
Model 5 vs. Model 6						9.07*	3
7. Disclosure Optimal Heteroscedastic	10.20*	3	.95	.91	.11
Model 5 vs. Model 7						18.8***	2
8. Disclosure Optimal Homoscedastic a	14.76*	6	.95	.94	.09
Model 7 vs. Model 8						4.56	3

Note: CFI = comparative fit index; TLI = Tucker Lewis index; RMSEA = root mean square error of approximation.

a

Retained model.

*

p < .05.

**

p < .01.

***

p < .001.

Figure 1 displays the mean latent growth curves for reliance and disclosure. Results support Hypothesis 1, indicating that levels of reliance and disclosure increase over time. The slope factor mean for reliance and disclosure was positive and statistically significant, indicating that participants’ intention to rely on and disclose information to their coworkers increased over time. The results also indicate that the rate of growth was faster during the initial period as participants first began to gather information about their new colleagues. In addition to the increase in trust levels proposed by Hypothesis 1, our results offer a number of interesting findings with regards to the pattern of trust development in new relationships. A statistically significant factor variance estimate for both initial status and slope of reliance and disclosure indicates that some individuals report higher intentions to engage in reliance and disclosure behaviors from the first day of their new job. A significant factor variance estimate also exists for the slope of both trust intentions, suggesting that some newcomers’ rates of reliance and disclosure increased at faster rates than others. Furthermore, there is a negative and statistically significant relationship between the initial status and slope of both variables, suggesting that individuals with higher initial levels of trust intentions showed less change over time than those that began with lower levels. These parameters are summarized in Table 5.

OPEN IN VIEWER

Figure 1

Mean Latent Growth Curves for Reliance and Disclosure

OPEN IN VIEWER

Table 5

Univariate Latent Growth Model Parameter Estimates

	Reliance		Disclosure
	Unstandardized	Standardized	Unstandardized	Standardized
SL Mean	0.36***	0.72***	0.49***	1.0***
IS Mean	4.56***	6.85***	4.11***	5.76***
SL Variance	0.24**	—	0.24**	—
IS Variance	0.50***	—	0.51***	—
IS with SL	−0.23**	−0.64***	−0.15*	−0.44***

Note: SL = slope; IS = initial status.

*

p < .05.

**

p < .01.

***

p < .001.

Hypothesis 2 proposed that propensity to trust is related to initial status of trust intentions but not to changes in reliance and disclosure over time. This was tested by regressing the initial status and slope of reliance and disclosure on propensity to trust. Augmenting the model with propensity to trust resulted in good model fit—reliance: χ²(5) = 10.49, ns, CFI = .97, TLI = .95, RMSEA = .08; disclosure: χ²(8) = 15.95, p < .05, CFI = .95, TLI = .94, RMSEA = .07. Results of the model indicate that propensity to trust is significantly related to initial status in reliance (β = 0.23, p < .05) and disclosure (β = 0.21, p < .05). Propensity to trust is unrelated to the rate of change in either variable. Overall, these results provide support for Hypothesis 2.

To test Hypotheses 3a and 3b, eight augmented latent growth models (four reliance, four disclosure) were created. These models allowed us to estimate the structural relationships between each of the trust cues and levels of reliance and disclosure at different points in the latent growth curve. Biesanz, Deeb-Sossa, Papadakis, Bollen, and Curran (2004) recommend that the origin or zero point for time in a latent growth model should be coded at the point in the model where the researcher would like to examine effects and relationships. In our analysis, we generated four augmented models for each variable (reliance and disclosure) and changed the coding of time in each model to reflect the stage of socialization we were interested in examining. For instance, to assess the structural relationships of trust cues with trust at Time 3, the first two change factor loadings were freely estimated while the second two were set at 0 and 1. Next, the six trust cues were regressed onto the intercept of the latent growth models to test which cues were significant. In each model, propensity to trust was included as a control variable to account for the influence of this individual difference on the perception and use of trust cues. Results of the analyses are presented in Table 6. Hypotheses 3a and 3b, which proposed that the impact of presumptive trust cues would decrease over time while the impact of personal trust cues increased, are not supported by the data. However, although the hypothesized presumptive-personal shift is not apparent, it is evident from the data that different trust cues predict intentions to engage in reliance and disclosure trust behaviors. The changing significance of the covariate predictors in the latent growth models also suggests a pattern may exist in the importance of trust cues over time. Potential interpretations of these results are addressed in the discussion section.

OPEN IN VIEWER

Table 6

Influence of Trust Cues on Reliance and Disclosure at Each Time Point

Trust Cue	Time 1	Time 2	Time 3	Time 4
Reliance
Competence	0.45***	0.54***	0.38***	0.38***
Benevolence	−0.13	0.12	0.03	0.11
Integrity	0.13	0.05	0.19*	−0.01
Group Identification	0.19*	0.07	0.01	0.26***
Role-Based Trust	0.15	0.02	−0.01	0.06
Rule-Based Trust	−0.02	−0.02	0.17***	0.20***
Disclosure
Competence	0.01	0.19	0.14	0.15
Benevolence	0.43***	0.49***	0.37***	0.54***
Integrity	0.13	0.04	0.04	−0.22
Group Identification	0.05	−0.07	0.07	0.20*
Role-Based Trust	0.07	−0.16	0.06	0.13
Rule-Based Trust	0.11	0.09	0.16	0.08

Note: The table presents standardized values for each measure.

*

p < .05.

***

p < .001.

Limitations and Future Research

Several limitations to the research must be noted. First, our study was conducted using a sample of newcomers who were members of one organization in a particular industry. Although organizational newcomers allow the perfect opportunity to track trust development from Day 1, further research is needed to investigate whether our findings can be generalized to other contexts. This sample of newcomers is relatively homogenous in age and previous work experience. This homogeneity may have influenced certain variables, in particular factors such as propensity to trust, which may vary with respect to previous positive or negative experiences in the workplace. Future research might investigate whether recent graduates demonstrate different trust development patterns to those who have built up more considerable work experience. The impact of different forms of socialization on the trust development process presents another avenue for future research. Participants in this sample experienced a socialization period characterized by institutionalized socialization tactics (Jones, 1986), which may have encouraged them to form a stronger social bond with each other, often linked to more positive organizational outcomes, including acceptance of organizational norms (Cable & Parsons, 2001) and positive work attitudes (Mignerey, Rubin, & Gorden, 1995). In comparison, new organizational members who are faced with a more variable, informal socialization environment may experience higher levels of uncertainty, lower group cohesion, and less positive interaction with their fellow newcomers as a result of a lack of collective socialization activities (Ashforth & Saks, 1996). These factors are in turn likely to negatively affect the trust development process. In addition, the referent used in this study was “other trainees in my group” where some groups had up to 30 members. Future research should investigate longitudinal trust in a more specific referent, for instance, with one particular coworker or within a smaller team. Moreover, by studying the process within specific dyads, future research could improve our understanding of the extent to which trust development trajectories are symmetrical. Patterns of trust development in the later stages of this study may also be influenced by the prospect of being reassigned to an alternative group in the near future. This relatively transient structuring of groups is not unusual in the context of professional service work where teams are regularly assembled and disassembled to meet changing client needs (Greenwood et al., 2010). Nevertheless, future research is needed to determine the impact of expectations regarding relationship length on trust intentions and behavior.

Second, our measures were completed by self-report, increasing the risk of common method bias in our results. However, our longitudinal design and the reordering of questions in the survey did allow us to limit some aspects of this bias (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003). Additionally, results from the single method factor procedure (Podsakoff et al., 2012) indicate that common method bias is not a major concern in this data set. While these steps do not entirely remove the potential for our results to be influenced by the self-report nature of our data, they go some way towards reducing the possibility of bias from a limitation that is typical of studies in this field. Perceptions of others and intentions to behave in a certain manner are within-person variables that are arguably measurable only through self-report (G. Chan, 2009).

Third, our tests of MEI uncovered a lack of scalar invariance for 3 of the 10 items in the Behavioral Trust Inventory (Gillespie, 2003). In practice, some level of noninvariance is common (Cheung & Lau, 2012; Vandenberg & Lance, 2000). However, testing for MEI is vital in helping researchers to interpret results and offers potentially meaningful information regarding the differential functioning of survey items across time. The noninvariance of certain items from other popular trust scales across cross-cultural groups (Wasti, Tan, Brower, & Önder, 2007) and time (Schaubroeck et al., 2013) has already been demonstrated. Our results appear to support the idea that employee understanding of the concept of trust is something that may change over time. Further research is needed to examine why certain items are less stable across time than others and whether the concept of trust is something that adapts and changes continually during an individual’s working life or if this change is specific to key periods (e.g., joining a new organization).

Finally, this study does not link trust to performance of newcomers during their socialization period. The focus of this study was to provide a deeper understanding of the trust development process rather than the impact of behaviors such as reliance and disclosure on individual or group performance measures. Considerable previous effort has been made in the literature to firmly establish the link between trust and performance (e.g., Colquitt et al., 2007); however, little research has been conducted with a longitudinal design. Further insight could be gained by collecting longitudinal data on performance as trust develops in a new relationship.

Practical Implications

There are several practical implications for management from this research. For example, if initial periods of socialization are most important to changes in trust, management may consider employing methods that consolidate this initial trust baseline and support the development of reliance and disclosure behavior. Structured social interaction associated with social events might be planned to provide the opportunity for newcomers to spend time with each other and facilitate this relational bonding. The fragility of trust and its vulnerability in the face of trust-undermining events (Kramer, 1999) suggests that the activities planned for newcomers in the socialization period are likely to have an important impact on whether initial trust is consolidated and built on or quickly destroyed.

Our results also point to the relative importance of cues that newcomers attend to when making trust decisions. In fact, we have provided evidence that information attended to in the first hours on a job is crucial. This would imply that managers need to pay particular attention to 1st-day socialization processes and optimize the opportunity for positive and supportive interaction on tasks that require collaboration and encourage trust consolidation. Team-building tasks and exercises are also known to influence trust and motivational development (Dirks, 1999). As evidence of coworker competence and benevolence continue to feed trust decisions, it may be useful for managers to craft specific work tasks that would showcase these characteristics. Consideration should be given to allowing time and opportunity for newcomer relationships to evolve from personal common ground to communal common ground (Priem & Nystrom, 2014), associated with a more resilient form of trust. Finally, while peak trust development occurs early in socialization, our research indicates that development is ongoing. This suggests that management should continue to exert effort in designing tasks and interactive spaces for newcomers to collaborate and build deeper relational ties.