Patterns of Thriving and Struggling: Teacher Professional Identity and Psychological Need Profile Membership and Movement Among Pre-Service Teachers

Abstract

This longitudinal study examined how pre-service teachers’ psychological need experiences are configured and how teacher professional identity is associated with profile membership and movement across these configurations. Drawing on self-determination theory (SDT), data were collected from Kuwaiti pre-service teachers across three waves over one academic year. Latent transition analysis was conducted with 317 complete-case participants using autonomy, competence, and relatedness satisfaction and frustration as indicators. A four-profile solution was retained: Flourishing, Moderately Thriving, Need-Frustrated, and Struggling. The label Moderately Thriving is comparative: this profile showed high need satisfaction in absolute terms but somewhat greater need frustration than Flourishing. The decoded profile assignments showed moderate descriptive stability across adjacent waves, with stability rates of 54.6% and 62.1%, though the Struggling profile showed markedly higher persistence than the others. Wave 1 teacher professional identity was associated with profile membership at both Wave 1 and Wave 3. In the dimension-level models, Teacher Self-Efficacy and Commitment to Teaching each showed adjusted associations across profiles and waves, whereas Professional Orientation was associated with profile membership only at Wave 1. Initial professional identity was not significantly associated with improvement or deterioration relative to stability. However, change in professional identity from Wave 1 to Wave 3 was associated with the direction of profile movement: increases were associated with lower odds of deterioration and higher odds of improvement. Because identity change and profile movement were assessed over the same period, the direction of this association cannot be established. These findings complicate a simple protective view of professional identity and suggest that supporting future teachers may require attention to both identity development and need-supportive programme conditions.

1. Introduction

Teacher education is a period of professional preparation, but it is also a period in which students’ developing sense of themselves as future teachers is tested. Teacher professional identity develops through experience, reflection, and participation in educational settings rather than remaining fixed throughout preparation (Beijaard et al., 2004; Izadinia, 2013). Pre-service teachers must develop teaching knowledge and capability while responding to coursework, assessment, practicum demands, and relationships within their programme. Their development therefore involves both teacher professional identity and the extent to which their psychological needs for autonomy, competence, and relatedness are satisfied or frustrated (Ryan & Deci, 2000, 2020).

Existing research has connected need-supportive teacher education conditions with professional identity development (Wong & Liu, 2024a, 2025). Person-centred research also shows that need satisfaction and need frustration may combine into distinct profiles and that membership in these profiles can change over time (Warburton et al., 2020; Krijgsman et al., 2025). Less is known, however, about how these configurations develop among pre-service teachers across an academic year and how professional identity is associated with profile membership and movement. This matters because need satisfaction and need frustration are related but distinct experiences (B. Chen et al., 2015). Students may feel supported in one need while experiencing frustration in another, and such mixed configurations can be concealed by average scores.

Teacher professional identity may be relevant because it includes beliefs about teaching capability, commitment to teaching, and orientation towards the profession (Hanna et al., 2019; Wong & Liu, 2024b). However, these dimensions should not be assumed to constitute a single protective resource. A student may feel capable of teaching or strongly committed to the profession while still encountering programme conditions that frustrate autonomy, competence, or relatedness. Professional identity may therefore be associated with psychological need profile membership without necessarily protecting students against movement towards less adaptive configurations.

The present three-wave study uses latent transition analysis to identify psychological need profiles among pre-service teachers and describe stability and movement between them (Ryoo et al., 2018). It then examines concurrent and prospective associations between teacher professional identity and profile membership, followed by whether initial identity and change in identity are associated with the direction of profile movement. Because the study is observational, these analyses identify associations rather than causal or developmental mechanisms.

2. Literature Review

Teacher education is more than a period in which students learn methods, curriculum design, and classroom management. It is also a period in which they begin to decide whether teaching is a profession they can realistically enter and sustain. That judgement is shaped by coursework and assessment and also shaped by practicum experience and by the ways students gradually come to recognise, or sometimes question, themselves as future teachers. For this reason, teacher professional identity is central to understanding pre-service teacher development. It refers to how student teachers connect their beliefs about teaching with their sense of capability, their commitment to teaching, and their imagined future professional role (Beijaard et al., 2004; Izadinia, 2013; Wong & Liu, 2024b).

Research on teacher professional identity has generally treated identity as developmental rather than fixed. Beijaard et al. (2004) argued that teacher identity is formed through experience, interpretation, and participation in professional contexts. Izadinia’s (2013) review reached a similar conclusion for student teachers, showing that identity development is shaped by reflection, mentoring, and school experience. It is also shaped by the tensions that appear when students move from imagining teaching to practising it. A student may enter teacher education with a fairly idealised view of the profession, then encounter a practicum classroom where managing time, responding to learners, and meeting institutional expectations make teaching feel less straightforward. More recent reviews confirm that teacher identity research has expanded considerably, although the field remains theoretically and methodologically diverse (Rushton et al., 2023; Yan, 2024). This diversity is useful because teacher identity is a broad construct. At the same time, it makes it more difficult to explain how particular identity resources are related to changing psychological experiences across teacher education.

Quantitative research has helped make some aspects of teacher professional identity more measurable. Hanna et al. (2019) showed that teacher identity measures often include self-efficacy, motivation, commitment, and views of teaching. Wong and Liu (2024b) developed this line of work further by validating a Teacher Professional Identity Scale for student teachers. Their measure focuses on teacher self-efficacy, commitment to teaching, and professional orientation. These dimensions are especially relevant to the present study because they connect identity with the practical demands of becoming a teacher. Teacher self-efficacy concerns whether students believe they can handle teaching tasks. Commitment to teaching concerns whether they want to remain connected to the profession. Professional orientation reflects how they understand teaching as a professional role.

These identity resources may support students during teacher education. A student with stronger teacher self-efficacy may interpret difficulty as part of learning rather than as evidence that they are unsuited to teaching. A student with stronger commitment may also be more willing to continue when teacher education becomes demanding. However, professional identity should not be treated as automatically protective. Recent evidence on teacher self-efficacy and quit intentions supports this caution by showing that efficacy beliefs may carry different implications under need-supportive and need-thwarting conditions (Alardhi & Ameen, 2026b). Identity can also create tension. Students may strongly want to become teachers and still feel constrained by institutional demands, assessment expectations, or practicum conditions. Pillen et al. (2013) showed that beginning teachers often experience professional identity tensions when there is a conflict between what they want and what is possible in practice. C. Chen et al. (2023) also showed that pre-service teachers’ professional identity can follow different developmental patterns over time. Some students become more secure, while others become less certain as professional demands become more visible. This suggests that identity needs to be studied in relation to changing psychological conditions, rather than treated as a purely positive individual trait.

Self-determination theory (SDT) offers a useful way to examine these psychological conditions. SDT proposes that autonomy, competence, and relatedness are basic psychological needs (Ryan & Deci, 2000, 2020). Autonomy refers to acting with a sense of personal endorsement. Competence refers to feeling effective. Relatedness refers to feeling connected to others. These needs are especially relevant in teacher education because student teachers occupy a mixed position. They are still learners, but they are also preparing for professional responsibility. They need guidance and feedback, but they also need some sense of agency. They need to feel capable, while also accepting that teaching is difficult and often uncertain. They also need support from supervisors, peers, and school communities, particularly when practicum experiences make the profession feel more immediate and less abstract.

A key development in SDT research is the distinction between need satisfaction and need frustration. B. Chen et al. (2015) showed that these are related but separate constructs. This distinction matters because low satisfaction does not necessarily mean active frustration. A student may feel only moderately competent because teaching is still new. That is different from feeling incapable, undermined, or repeatedly exposed to situations that make teaching feel unmanageable. A student may also experience some autonomy while still feeling pressure from assessment or institutional expectations. Treating satisfaction and frustration as opposite ends of a single scale can therefore hide important psychological differences. This is particularly relevant in teacher education, where mixed experiences are common and where the same programme may support students in one respect while placing pressure on them in another.

Recent SDT work in teacher education supports this point. Granero-Gallegos et al. (2024) showed that educator-created empowering and disempowering climates were linked to pre-service teachers’ teaching intention through motivational pathways. Wong and Liu (2024a) also examined teacher professional identity from an SDT perspective and showed that need satisfaction helped explain identity development. Their later longitudinal work suggested that teacher professional identity may grow during coursework but decline during teaching practice when need-supportive conditions are weaker (Wong & Liu, 2025). These studies are valuable because they connect teacher education conditions with motivation and identity. Still, they mainly explain average relationships between variables. They do not show whether different groups of students experience different patterns of need satisfaction and need frustration over time.

Person-centred research addresses this limitation by examining configurations rather than single average scores. In a variable-centred model, autonomy satisfaction, competence satisfaction, and relatedness satisfaction are usually examined as separate predictors or outcomes. This approach is useful for testing general associations, yet it cannot easily show whether some students experience high relatedness and competence alongside constrained autonomy. It also cannot show whether satisfaction and frustration coexist within the same student. Warburton et al. (2020) showed that need satisfaction and need frustration can combine in distinct profiles. Krijgsman et al. (2025) extended this work by examining the temporal stability of need satisfaction and frustration profiles. Their findings are important because they suggest that profile membership can remain stable for some learners and shift for others. Related longitudinal research with Kuwaiti pre-service teachers identified distinct profiles of teaching self-efficacy, task value, perceived cost, and transformative experience, with evidence of both stability and movement across an academic term (Ameen & Alardhi, 2026). Although those profiles involved different constructs, the findings provide contextually relevant support for examining motivational heterogeneity through longitudinal person-centred analysis. Alardhi and Ameen (2026a) offer a useful point of comparison from the Kuwaiti teacher education context. Their findings suggest that teachers’ need experiences are better understood as distinct satisfaction-frustration profiles rather than as simple differences in overall need support. Some profiles were relatively straightforward, while others were more mixed, with teachers reporting relatively high need satisfaction alongside elevated frustration.

This person-centred logic is highly relevant to pre-service teachers. Teacher education brings together coursework, practicum, assessment, and changing expectations about the profession. These experiences are unlikely to affect all students in the same way. Some students may feel supported and capable across time. Others may feel socially connected but restricted in autonomy. Others may continue to feel committed to teaching while also experiencing frustration in the conditions through which they are learning to teach. Such patterns are difficult to capture through mean-level change alone. Latent transition analysis is therefore appropriate because it identifies profiles and estimates movement between them across waves. It allows the study to examine whether students remain in the same psychological need state, move toward a more adaptive profile, or shift toward a more frustrated one (Hickendorff et al., 2018; Ryoo et al., 2018).

The remaining issue concerns the role of teacher professional identity in this process. Existing SDT-informed teacher identity research has mainly examined whether need-supportive conditions predict identity development (Wong & Liu, 2024a, 2025). This direction is important, but the reverse direction also deserves attention. Professional identity is treated here as a multidimensional set of identity resources rather than a single protective disposition. Its dimensions may be related to psychological need experiences through different pathways. Pre-service teachers do not enter teacher education as psychologically neutral. They bring different beliefs about their teaching capability, different levels of commitment, and different understandings of the teaching role. Teacher self-efficacy reflects students’ judgements about whether they can manage teaching tasks. Such beliefs may shape whether difficulties are interpreted as manageable challenges or as evidence of limited capability. They may also influence whether students persist long enough to experience progress and useful feedback, which could be relevant to competence satisfaction (Bandura, 1997). Commitment to teaching may connect demanding coursework or practicum activity to a valued professional direction. When this engagement is willingly endorsed, it may help students retain a sense of purpose. However, commitment can also coexist with pressure when students feel that an important professional identity is threatened. Professional orientation may give programme requirements professional meaning, but endorsing teaching as a valued occupation does not necessarily provide autonomy, competence, or relatedness within the programme itself. These proposed pathways are theoretically plausible, but they were not measured as mediating processes in the present study. They are used here only to explain why the three identity dimensions may show different associations with psychological need experiences. Professional identity may be associated with how students experience teacher education, but the direction of influence cannot be established from the present observational design.

The present study addresses this gap by bringing teacher professional identity research together with SDT-based profile research. It examines psychological need satisfaction and frustration across three waves of teacher education. Using latent transition analysis, it identifies distinct need profiles and examines movement between these profiles over time. It then examines whether Wave 1 teacher professional identity is associated with profile membership at the same wave and prospectively associated with profile membership at Wave 3. It also considers whether initial professional identity and change in professional identity across the study period are associated with the direction of profile movement. In this way, the study considers professional identity not simply as an outcome of need-supportive conditions, but also as an individual resource that may develop alongside changing psychological need experiences. Because change in professional identity and profile movement were measured across the same period, their association does not establish which changed first or whether one caused the other. The study is guided by the following research questions:

• (RQ1) What distinct psychological need satisfaction and frustration profiles exist among pre-service teachers, and how do these profiles differ in their configurations of need satisfaction and frustration?
• (RQ2) How do pre-service teachers transition between psychological need profiles across waves, and what is the overall degree of profile stability?
• (RQ3) How is teacher professional identity at Wave 1 associated with psychological need profile membership at Wave 1?
• (RQ4) Is teacher professional identity at Wave 1 prospectively associated with psychological need profile membership at Wave 3?
• (RQ5) Are initial professional identity and change in professional identity from Wave 1 to Wave 3 associated with the direction of profile movement over the same period?

3. Method

Participants and Procedure

Participants were pre-service teachers enrolled in the College of Education at Kuwait University, recruited through course instructors who distributed the survey link to their students. Eligibility required participants to be at least 18 years of age, and all provided informed consent prior to participation. Data were collected across three waves over the course of one academic year, with approximately one semester between each wave, yielding N = 496 unique participants across the study (n = 433 at Wave 1, n = 421 at Wave 2, n = 373 at Wave 3). The latent transition analysis was conducted on the 317 participants who provided complete data across all three waves. The survey was administered anonymously online at each wave. Participants were drawn from different years of a multi-year teacher education programme, so the three waves did not represent the same programme stage for every participant. Information linking individual survey completion to practicum placements or major assessments was not collected systematically.

Demographic characteristics at Wave 1 are presented in

Table 1. Demographic Characteristics of Participants at Wave 1 (N = 433).

Characteristic	n	%
Age (years): M = 21.10 (SD = 4.55), range = 18–49
Gender
Female	346	79.9
Male	87	20.1
Nationality
Kuwaiti	385	88.9
Non-Kuwaiti	41	9.5
Prefer not to say	7	1.6
Marital Status
Single	359	82.9
Married	51	11.8
Divorced	8	1.8
Prefer not to say	15	3.5
Teaching Specialisation
Arabic Language	122	28.2
English Language	80	18.5
Science	85	19.6
Social Studies	63	14.5
Islamic Education	21	4.8
Mathematics	14	3.2
Early Childhood	12	2.8
Prefer not to say	36	8.3
Year of Study
First year	68	15.7
Second year	199	46.0
Third year	119	27.5
Fourth year	39	9.0
Fifth year or above	8	1.8
GPA
3.50–4.00	218	50.3
3.00–3.49	136	31.4
2.50–2.99	65	15.0
2.00–2.49	7	1.6
Below 2.00	7	1.6

Note. Demographics are reported for the Wave 1 sample. Percentages may not sum to 100 due to rounding or missing data.

. The sample was predominantly female (79.9%) and Kuwaiti (88.9%). The majority of participants were in their second or third year of study (46.0% and 27.5%, respectively) and reported a GPA of 3.00 or above (81.7%). Arabic Language and Science were the most common teaching specialisations.

4. Ethical Considerations

Ethical approval for this study was obtained from the Head of Department at Educational Psychology at Kuwait University on 8 October 2025. As a formal institutional review board had not yet been established at the time of data collection, no ethics approval reference number was assigned. Nevertheless, all study procedures were conducted in strict accordance with relevant institutional guidelines and applicable legal and regulatory requirements. Informed consent was obtained from all participants prior to their involvement in the study, via an online consent statement presented before the questionnaire, whereby participants confirmed their voluntary agreement to participate after reading the study information. No personally identifying information was collected at any stage. Participant privacy and confidentiality were safeguarded through the anonymisation of collected data and secure, password-protected storage, with access restricted exclusively to members of the research team.

5. Measures

5.1. Basic Psychological Need Satisfaction and Frustration Scale

Psychological need satisfaction and frustration were assessed using the Basic Psychological Need Satisfaction and Frustration Scale (BPNSNF; B. Chen et al., 2015), which measures satisfaction and frustration across three basic psychological needs, autonomy, relatedness, and competence, yielding six subscales: Autonomy Satisfaction, Autonomy Frustration, Relatedness Satisfaction, Relatedness Frustration, Competence Satisfaction, and Competence Frustration. Items were rated on a five-point Likert scale ranging from 1 (completely untrue) to 5 (completely true). One item from the Relatedness Frustration subscale (item 22) was excluded following confirmatory factor analysis. Model fit improved with this item removed (full model: CFI = 0.926, RMSEA = 0.041; reduced model: CFI = 0.934, RMSEA = 0.040). Item 22 addressed the experience of relationships as superficial. After its removal, the three retained items continued to represent relational exclusion, interpersonal distance, and the perception of being disliked. However, removing the item reduced the breadth of relational frustration captured by the subscale. Relatedness Frustration also showed a pronounced floor effect: 131 of 317 participants (41.3%) reported the identical minimum score across all three waves. Internal consistency estimates across waves are presented in

Table 2. Internal Consistency (Cronbach’s α) Across Three Waves.

Scale/Subscale	Wave 1	Wave 2	Wave 3
Basic Psychological Need Satisfaction and Frustration Scale
Autonomy Satisfaction	0.527	0.684	0.752
Autonomy Frustration	0.645	0.776	0.846
Relatedness Satisfaction	0.658	0.746	0.799
Relatedness Frustration	0.613	0.743	0.784
Competence Satisfaction	0.746	0.811	0.842
Competence Frustration	0.663	0.790	0.804
Teacher Professional Identity Scale
Teacher Self-Efficacy	0.902	0.925	0.946
Commitment to Teaching	0.879	0.894	0.905
Professional Orientation	0.926	0.927	0.938

Note. BPNSNF = Basic Psychological Need Satisfaction and Frustration Scale; TPI = Teacher Professional Identity Scale (Wong & Liu, 2024b). BPNSNF item 22 was excluded from Relatedness Frustration. Commitment to Teaching item 2 was excluded based on a negative standardised loading (β = −0.394).

. Autonomy Satisfaction reliability was below the conventional threshold at Wave 1 (α = 0.527, ω = 0.601), improving to acceptable levels at Waves 2 and 3. Because this subscale contributed to the six-indicator profile solution, its weak Wave 1 reliability may have understated true between-person variability in Autonomy Satisfaction and attenuated profile separation on this indicator at the beginning of the study. This limitation should be considered when interpreting Wave 1 profile assignments and estimated movement involving the first wave.

5.2. Teacher Professional Identity Scale

Teacher professional identity was assessed using the Teacher Professional Identity Scale (TPI; Wong & Liu, 2024b), comprising three subscales: Teacher Self-Efficacy (9 items), Commitment to Teaching (4 items), and Professional Orientation (5 items). One item from the Commitment to Teaching subscale was excluded based on a negative standardised factor loading (β = −0.394) in the full four-item model, indicating it was directionally inconsistent with the remaining items. The Commitment to Teaching scores and the α coefficients reported in Table 2 were calculated from the retained three-item version. Items were rated on a five-point scale. A composite TPI score was computed as the mean of the three subscale means. Internal consistency was acceptable to excellent across all waves and subscales (see Table 2).

Longitudinal measurement invariance was evaluated for each BPNSNF subscale. Metric invariance held for all six subscales (all ΔCFI ≤ 0.003, p > 0.05). Scalar invariance held fully only for Competence Satisfaction; partial invariance was achieved for Autonomy Satisfaction, Autonomy Frustration, and Relatedness Satisfaction (one to two item intercepts freed per subscale), but could not be established for Relatedness Frustration or Competence Frustration, so wave-to-wave mean comparisons on these two subscales should be interpreted with caution.

6. Analytic Strategy

A latent transition analysis (LTA) was conducted using the depmixS4 package (Visser & Speekenbrink, 2010) in R (Version 4.5.0), with the six BPNSNF subscale means as continuous indicators. LTA is a longitudinal extension of latent profile analysis that simultaneously estimates discrete profiles at each wave and the probabilities of transitioning between profiles across waves (Ryoo et al., 2018). Prior to the LTA, longitudinal measurement invariance (configural, metric, scalar) was tested for each BPNSNF subscale using repeated-measures CFA, evaluated against F. F. Chen’s (2007) criteria (ΔCFI ≤ 0.01, ΔRMSEA ≤ 0.015). The LTA itself proceeded in three stages.

The six continuous indicators were specified as conditionally independent Gaussian responses. Indicator variances were freely estimated across the latent profiles but constrained to remain equal across waves within each profile. Within-profile indicator covariances were fixed to zero. This specification reflects the local independence assumption that associations among the six indicators are accounted for by latent profile membership.

Competing models with two to nine profiles were first explored using up to 400 independently seeded random starts per model. This initial search was diagnostic. It repeatedly produced numerically degenerate solutions in which a profile-specific response standard deviation approached zero, particularly for Relatedness Frustration because of its pronounced floor effect and many exact ties. These solutions were not used for the final model comparison because they were obtained before the admissibility safeguards were applied and did not provide a dependable alternative substantive solution.

To reduce this numerical instability, random jitter was applied to tied observed scores before model estimation. This step was used to break exact ties rather than to represent additional substantive variation in participants’ responses. Solutions in which any profile-specific response standard deviation fell below 0.01 were excluded. The model comparison was then rerun under these common safeguards. Because the guarded search showed substantially greater replication stability, the final reported comparison used 30 starts for the two-profile solution and 75 starts for each remaining solution. The model-comparison table reports only this final guarded search. Model selection considered BIC, replication of the best admissible log-likelihood, minimum profile size, and theoretical interpretability rather than BIC alone. Because jitter modified the observed values for estimation, the pronounced floor effect and its implications are explicitly acknowledged when interpreting the profiles.

The selected solution was refined with 50 additional independently seeded random starts. A heterogeneous alternative allowing separate transition matrices per interval was compared against the homogeneous model via a likelihood ratio test before adopting the homogeneous specification. Viterbi decoding was used to assign participants to their most likely profile at each wave. Classification quality was evaluated using average posterior probabilities (AvePP), with values above 0.70 considered indicative of acceptable classification.

Profile-membership associations were estimated using a maximum-likelihood three-step procedure (Vermunt, 2010) that accounted for classification error while preserving the previously estimated four-profile measurement model. Separate wave-specific classification-error matrices were calculated from the posterior probabilities obtained from the LTA. Each matrix represented the probability of assignment to an observed profile conditional on membership in each latent profile. These matrices were then held fixed while multinomial logistic regressions estimated associations between the predictors and latent profile membership. The corrected likelihood summed over the possible true profiles for each participant, thereby retaining all 317 participants without treating Viterbi-assigned membership as error-free. Flourishing was the reference profile.

Model A used the overall Teacher Professional Identity composite, whereas Model B entered Teacher Self-Efficacy, Commitment to Teaching, and Professional Orientation simultaneously. Wave 1 models examined concurrent profile membership, while Wave 3 models used Wave 1 identity scores to examine prospective associations. Standard errors were obtained from the inverse observed-information matrix.

Because the three Teacher Professional Identity dimensions were moderately to strongly correlated, supplementary collinearity and suppression diagnostics were conducted using the Wave 1 predictor scores from the participant-level complete-case sample (n = 317). For each dimension, tolerance was calculated as 1 − R² from an auxiliary regression of that dimension on the other two dimensions, and the variance inflation factor was calculated as 1/tolerance. To examine the unexpected positive adjusted coefficients for Professional Orientation, post hoc nested maximum-likelihood three-step models estimated its association with profile membership when entered alone, with Teacher Self-Efficacy, with Commitment to Teaching, and with both dimensions. These models used the same fixed Wave 1 classification-error matrix as the primary profile-membership analysis.

Across Waves 1 and 3, the primary Model A and Model B analyses involved 24 coefficient tests: three composite-score contrasts per wave and nine dimension-level coefficients per wave. The nested suppression models were additional post hoc analyses. No formal multiplicity correction was applied. Interpretation therefore emphasised effect sizes, confidence intervals, and consistency across profiles and waves rather than any single p-value.

Trajectory analyses classified participants as Stable, Improving, or Deteriorating based on their movement between ranked profiles from Wave 1 to Wave 3, with profiles ranked using a composite score (summed satisfaction minus summed frustration across the three needs); a frustration-only ranking was computed as a sensitivity check and produced an identical ordering. Change in TPI from Wave 1 to Wave 3 was additionally examined in relation to trajectory class. Multinomial logistic regression was used to examine whether initial TPI and change in TPI were associated with trajectory class. The LTA was estimated on the complete-case sample (n = 317). Attrition analyses compared completers and non-completers on observed Wave 1 variables to identify possible patterns of selective attrition. These comparisons cannot establish that the data were missing at random. As a robustness check, the LTA was re-estimated on the full sample (N = 496) using the available-response likelihood handling implemented in depmixS4. Missing response values were given a neutral contribution to the likelihood, allowing participants with incomplete responses to be retained without imputing replacement scores. This analysis examined whether the complete-case restriction materially altered the profile solution.

As an additional check on the robustness of the retained profile structure across waves, cross-sectional latent profile analyses were conducted independently at each wave using the complete-case sample (n = 317). Each analysis applied the same safeguards against numerical degeneracy described above, including random jitter, a minimum variance threshold, and replication tracking across up to 50 starts for solutions containing two to seven profiles. These analyses examined whether a comparable four-profile configuration could be recovered from each wave separately without drawing on information from the other waves.

7. Results

7.1. Preliminary Analyses

7.1.1. Measurement Quality

Confirmatory factor analysis supported the six-factor structure of the BPNSNF (CFI = 0.934, TLI = 0.923, RMSEA = 0.040, SRMR = 0.045), with all standardised loadings significant at p < 0.001 (range: 0.326–0.737). The three-factor structure of the TPI demonstrated acceptable fit (CFI = 0.952, TLI = 0.944, RMSEA = 0.069, SRMR = 0.042); although the RMSEA marginally exceeded the conventional 0.06 threshold, the CFI and SRMR values indicated good fit, consistent with the known sensitivity of RMSEA in models with limited degrees of freedom (Kenny et al., 2015). Internal consistency estimates are presented in Table 2.

7.1.2. Attrition

Completers (n = 317) and non-completers (n = 116) were compared on all Wave 1 scale scores using independent-samples t-tests. Completers reported significantly higher Autonomy Satisfaction, Relatedness Satisfaction, and Competence Satisfaction, and lower Competence Frustration, relative to non-completers, though effect sizes were small (Cohen’s d range: 0.21–0.37). Autonomy Frustration, Relatedness Frustration, and all TPI subscales did not differ significantly between groups. When all Wave 1 variables were entered simultaneously into a logistic regression predicting completer status, no individual predictor reached significance, indicating that differences were modest and non-systematic. Findings should be interpreted with appropriate caution given the potential for selective attrition on need satisfaction.

7.1.3. Descriptive Statistics

Across waves, need satisfaction subscales were moderate to high (means ranging from 3.98 to 4.37), while need frustration was low to moderate (means ranging from 1.34 to 2.69). TPI subscale means were consistently above the scale midpoint across all three waves. Satisfaction and frustration subscales were negatively intercorrelated as expected, and TPI subscales were strongly positively intercorrelated. To retain all observations available at each measurement occasion, the descriptive statistics and correlations were calculated using the wave-specific samples (n = 433, 421, and 373). The LTA and subsequent transition and predictor analyses used the 317 participants with complete observations across all three waves. Descriptive statistics and intercorrelations are presented in

Table 3. Intercorrelations Among Study Variables by Wave.

Variable	1	2	3	4	5	6	7	8	9	10
Wave 1
1. Autonomy Satisfaction	—
2. Autonomy Frustration	−0.26	—
3. Relatedness Satisfaction	0.52	−0.18	—
4. Relatedness Frustration	−0.22	0.25	−0.36	—
5. Competence Satisfaction	0.56	−0.21	0.48	−0.19	—
6. Competence Frustration	−0.42	0.36	−0.28	0.42	−0.47	—
7. Teacher Self-Efficacy	0.34	−0.09	0.33	−0.25	0.39	−0.28	—
8. Commitment to Teaching	0.40	−0.20	0.36	−0.30	0.30	−0.31	0.56	—
9. Professional Orientation	0.25	−0.04	0.23	−0.28	0.23	−0.20	0.62	0.70	—
10. Teacher Professional Identity (Composite)	0.38	−0.13	0.35	−0.32	0.35	−0.30	0.82	0.88	0.89	—
Wave 2
1. Autonomy Satisfaction	—
2. Autonomy Frustration	−0.42	—
3. Relatedness Satisfaction	0.60	−0.23	—
4. Relatedness Frustration	−0.25	0.34	−0.42	—
5. Competence Satisfaction	0.69	−0.32	0.60	−0.30	—
6. Competence Frustration	−0.40	0.55	−0.41	0.55	−0.49	—
7. Teacher Self-Efficacy	0.43	−0.22	0.43	−0.35	0.53	−0.42	—
8. Commitment to Teaching	0.46	−0.23	0.42	−0.37	0.43	−0.39	0.69	—
9. Professional Orientation	0.34	−0.14	0.41	−0.39	0.38	−0.33	0.70	0.69	—
10. Teacher Professional Identity (Composite)	0.46	−0.22	0.47	−0.42	0.50	−0.43	0.89	0.90	0.89	—
Wave 3
1. Autonomy Satisfaction	—
2. Autonomy Frustration	−0.40	—
3. Relatedness Satisfaction	0.63	−0.22	—
4. Relatedness Frustration	−0.34	0.38	−0.53	—
5. Competence Satisfaction	0.75	−0.32	0.66	−0.41	—
6. Competence Frustration	−0.44	0.57	−0.47	0.63	−0.56	—
7. Teacher Self-Efficacy	0.45	−0.21	0.49	−0.43	0.55	−0.47	—
8. Commitment to Teaching	0.43	−0.25	0.47	−0.42	0.50	−0.47	0.76	—
9. Professional Orientation	0.38	−0.17	0.45	−0.45	0.47	−0.41	0.79	0.81	—
10. Teacher Professional Identity (Composite)	0.46	−0.23	0.51	−0.47	0.55	−0.49	0.91	0.93	0.93	—

Note. N = 433 (Wave 1), 421 (Wave 2), 373 (Wave 3). Values are Pearson correlation coefficients. Variable numbers correspond to the numbered list in the left-hand column and are repeated as column headers to conserve space.

.

• RQ1: What distinct psychological need satisfaction and frustration profiles exist among pre-service teachers, and how do these profiles differ in their configurations of need satisfaction and frustration?

Model fit indices for the two- to nine-profile solutions are presented in

Table 4. Latent Transition Analysis Model Fit Statistics for Two- to Nine-Profile Solutions.

Profiles	Log-Likelihood	AIC	BIC	Entropy	Replication (Best LL/Starts)
2	−4733.06	9520.12	9651.27	0.950	30/30
3	−3679.81	7447.62	7661.35	0.961	3/75
4	−3409.19	6944.37	7250.39	0.946	35/75
5	−2926.69	6021.37	6429.40	0.937	1/75
6	−2810.80	5835.59	6355.35	0.932	1/75
7	−2646.84	5557.68	6198.87	0.933	1/75
8	−2424.09	5166.19	5938.53	0.932	1/75
9	−2408.97	5193.94	6107.15	0.930	1/75

Note. AIC = Akaike Information Criterion; BIC = Bayesian Information Criterion. Lower AIC and BIC values indicate better fit. Entropy ranges from 0 to 1, with values closer to 1 indicating cleaner classification. Replication indicates the number of independent random starts (out of the total attempted) that converged on the reported best log-likelihood. The four-profile solution (bolded) was selected based on the combination of Bayesian Information Criterion, replication stability, minimum class size, and theoretical interpretability; although Bayesian Information Criterion continued to decrease through eight profiles, solutions with seven or more profiles produced classes as small as eighteen observations (1.9% of the pooled observations) and were replicated by only one of seventy-five independent starts, consistent with over-extraction.

. BIC decreased through eight profiles (BIC = 5938.53) before increasing at nine (BIC = 6107.15), confirming that the search extended beyond the initially retained solution. Although the five- and six-profile solutions had lower BIC values than the four-profile solution, the best log-likelihood for each was replicated in only one of 75 starts. These solutions were therefore not considered sufficiently stable to support substantive interpretation. The seven- and eight-profile solutions showed the same limited replication and also produced profiles containing as few as 18 observations (1.9% of the pooled observations), raising further concern about over-extraction (Nylund et al., 2007). In contrast, the four-profile solution was replicated by 35 of 75 starts and had a minimum profile size of 128 observations (13.5%). It was therefore retained because it offered the strongest balance of statistical stability, adequate profile size, classification quality, and substantive interpretability, despite not having the lowest BIC.

The final four-profile solution was estimated with 50 additional random starts (log-likelihood = −3409.19; BIC = 7250.39). Classification quality was high, with average posterior probabilities ranging from 0.922 to 0.989. The two most similar profiles showed near-total classification separation (own-profile probabilities of 0.922 and 0.981, cross-profile probabilities near zero), supporting their empirical separation within the retained solution.

Four profiles reflected distinct configurations of need satisfaction and frustration. Flourishing (n = 202) showed the highest satisfaction and lowest frustration across all needs. At the reported precision, this profile reached the scale ceiling for Relatedness Satisfaction (5.00) and the scale floor for Relatedness Frustration (1.00) at every wave. Flourishing was therefore partly characterised by boundary-level relatedness responses, and its distinction from Moderately Thriving should be interpreted from the full six-indicator configuration rather than from the relatedness indicators alone. Moderately Thriving (n = 387) showed solidly positive satisfaction with somewhat elevated Autonomy Frustration. Need-Frustrated (n = 128) showed satisfaction comparable to Moderately Thriving but meaningfully elevated Relatedness Frustration. Struggling (n = 234) showed the lowest satisfaction and highest frustration across all needs. The label Moderately Thriving is used relative to the Flourishing profile and reflects the coexistence of strong need satisfaction with somewhat greater need frustration. It does not suggest that participants in this profile reported only moderate raw levels of need satisfaction. In the pooled estimates, Moderately Thriving showed higher Autonomy Frustration than Flourishing (2.58 versus 1.88) and higher Competence Frustration (1.62 versus 1.24), despite comparatively high satisfaction scores. The label therefore describes a positive but less optimal configuration relative to Flourishing rather than low functioning in absolute terms. Profile means and prevalence estimates by wave are presented in

Table 5. Profile Indicator Means and Prevalence by Wave.

Profile	N (%)	Autonomy Satisfaction	Autonomy Frustration	Relatedness Satisfaction	Relatedness Frustration	Competence Satisfaction	Competence Frustration
Wave 1
Flourishing	70 (22.1%)	4.70	2.00	5.00	1.00	5.00	1.27
Moderately Thriving	158 (49.8%)	4.11	2.53	4.50	1.00	4.35	1.54
Need-Frustrated	36 (11.4%)	4.07	2.81	4.40	2.02	4.57	1.64
Struggling	53 (16.7%)	3.10	3.52	3.56	2.26	3.28	3.07
Wave 2
Flourishing	63 (19.9%)	4.79	1.88	5.00	1.00	5.00	1.19
Moderately Thriving	121 (38.2%)	4.19	2.58	4.52	1.00	4.34	1.69
Need-Frustrated	52 (16.4%)	4.19	2.52	4.40	1.87	4.49	1.87
Struggling	81 (25.6%)	3.26	3.47	3.55	2.40	3.44	3.10
Wave 3
Flourishing	69 (21.8%)	4.86	1.77	5.00	1.00	5.00	1.25
Moderately Thriving	108 (34.1%)	4.22	2.66	4.63	1.00	4.49	1.65
Need-Frustrated	40 (12.6%)	4.19	2.65	4.42	2.07	4.42	1.84
Struggling	100 (31.5%)	3.38	3.38	3.54	2.69	3.40	3.26
Overall (Pooled)
Flourishing	202	4.78	1.88	5.00	1.00	5.00	1.24
Moderately Thriving	387	4.16	2.58	4.54	1.00	4.39	1.62
Need-Frustrated	128	4.16	2.64	4.41	1.97	4.49	1.79
Struggling	234	3.27	3.44	3.55	2.49	3.38	3.16

Note. Values reflect Basic Psychological Need Satisfaction and Frustration Scale subscale means on a 1–5 scale. Profile percentages are calculated within each wave. The Overall (Pooled) row reports means and total sample sizes across all three waves combined.

.

Within the longitudinal LTA assignments, all four profiles were represented at each wave, although their prevalence changed over time. Moderately Thriving declined from 49.8% at Wave 1 to 34.1% at Wave 3, whereas Struggling increased from 16.7% to 31.5%. A supplementary analysis examining whether a comparable four-profile configuration could be recovered independently at each wave is reported below (see Robustness Check: Wave-Specific Profile Structure).

• RQ2: How do pre-service teachers transition between psychological need profiles across waves, and what is the overall degree of profile stability?

Overall profile stability was 54.6% from Wave 1 to Wave 2 and 62.1% from Wave 2 to Wave 3. A heterogeneous transition model allowing separate matrices per interval did not improve fit over the homogeneous model, χ²(32) = 10.17, p = 0.999, ΔBIC = 154.41 favouring homogeneity. Because several transition probabilities were at or near the boundary of the parameter space, the usual chi-square reference distribution may not be exact. The homogeneous model was therefore retained primarily because of its substantially lower BIC and greater parsimony rather than on the p-value alone. Transition probabilities between profiles at both intervals are presented in

Table 6. Empirical Profile Transition Probabilities Across Measurement Intervals.

Origin Profile	Flourishing	Moderately Thriving	Need-Frustrated	Struggling
Wave 1 to Wave 2
Flourishing	0.500	0.286	0.129	0.086
Moderately Thriving	0.158	0.506	0.190	0.146
Need-Frustrated	0.083	0.361	0.361	0.194
Struggling	0.000	0.151	0.000	0.849
Wave 2 to Wave 3
Flourishing	0.683	0.143	0.063	0.111
Moderately Thriving	0.165	0.545	0.132	0.157
Need-Frustrated	0.096	0.404	0.385	0.115
Struggling	0.012	0.148	0.000	0.840

Note. Values represent row-conditional probabilities (the probability of transitioning to each destination profile given membership in the origin profile at the earlier wave). Diagonal values (bolded) represent profile stability. The two interval-specific matrices are descriptive summaries calculated from participants’ Viterbi-decoded profile assignments; they are not separate model-estimated transition matrices. A heterogeneous transition model allowing separate matrices for each interval did not improve fit over a single shared (homogeneous) transition structure, χ²(32) = 10.17, p = 0.999.

.

The Struggling profile was by far the most stable (84.9% and 84.0% retention). The descriptive retention rate for Flourishing was higher in the second interval than in the first (50.0% to 68.3%). Need-Frustrated was least stable (36.1% and 38.5%), most often moving toward Moderately Thriving. Direct movement from Struggling to Flourishing was essentially absent.

• RQ3: How is teacher professional identity at Wave 1 associated with psychological need profile membership at Wave 1?

Maximum-likelihood three-step multinomial regressions examined associations between Wave 1 TPI scores and Wave 1 profile membership, with Flourishing as the reference profile. Classification uncertainty was incorporated using the fixed Wave 1 classification-error matrix. Year of study was not included as a covariate because the small upper-year groups and the structural zero in the Need-Frustrated profile were likely to produce unstable estimates. The distribution of profile membership by year of study is therefore reported descriptively in

Table 7. Year of Study by Psychological Need Profile Membership at Wave 1.

Year of Study	N (Complete-Case Sample)	Flourishing	Moderately Thriving	Need-Frustrated	Struggling
First Year	53	14 (26.4%)	29 (54.7%)	3 (5.7%)	7 (13.2%)
Second Year	159	38 (23.9%)	74 (46.5%)	23 (14.5%)	24 (15.1%)
Third Year	78	15 (19.2%)	38 (48.7%)	10 (12.8%)	15 (19.2%)
Fourth Year	22	2 (9.1%)	15 (68.2%)	0 (0.0%)	5 (22.7%)
Fifth Year or Above	5	1 (20.0%)	2 (40.0%)	0 (0.0%)	2 (40.0%)

Note. Values are based on the 317 participants with complete data across all three waves. Percentages are calculated within year of study. No participants in the fourth year or fifth year and above were classified into the Need-Frustrated profile at Wave 1. This structural zero, combined with the small numbers of participants in the upper-year groups, was likely to produce unstable covariate estimates. Year of study was therefore not included in the profile-membership regression models. Fourth-year (n = 22) and fifth-year-or-above (n = 5) cells are small and should be interpreted with caution.

. The regression results are presented in

Table 8. Maximum-Likelihood Three-Step Multinomial Regression Estimates for Associations Between Teacher Professional Identity and Wave 1 and Wave 3 Profile Membership (Reference = Flourishing).

Comparison vs. Flourishing	Model A: TPI Composite	Model B: Teacher Self-Efficacy	Model B: Commitment to Teaching	Model B: Professional Orientation
Panel A. Wave 1 Profile Membership (Wave 1 TPI Predictors)
Moderately Thriving	OR = 0.67 95% CI [0.38, 1.19] p = 0.174	OR = 0.39 95% CI [0.22, 0.71] p = 0.002	OR = 0.59 95% CI [0.34, 1.02] p = 0.059	OR = 2.61 95% CI [1.42, 4.80] p = 0.002
Need-Frustrated	OR = 0.49 95% CI [0.25, 0.98] p = 0.044	OR = 0.41 95% CI [0.19, 0.90] p = 0.025	OR = 0.41 95% CI [0.20, 0.82] p = 0.012	OR = 2.78 95% CI [1.26, 6.17] p = 0.012
Struggling	OR = 0.22 95% CI [0.12, 0.40] p < 0.001	OR = 0.16 95% CI [0.08, 0.35] p < 0.001	OR = 0.25 95% CI [0.13, 0.48] p < 0.001	OR = 4.07 95% CI [1.95, 8.51] p < 0.001
Panel B. Wave 3 Profile Membership (Wave 1 TPI Predictors)
Moderately Thriving	OR = 0.33 95% CI [0.16, 0.68] p = 0.003	OR = 0.57 95% CI [0.32, 1.01] p = 0.054	OR = 0.46 95% CI [0.26, 0.83] p = 0.010	OR = 1.53 95% CI [0.79, 2.96] p = 0.209
Need-Frustrated	OR = 0.44 95% CI [0.18, 1.06] p = 0.067	OR = 0.45 95% CI [0.22, 0.93] p = 0.030	OR = 0.65 95% CI [0.30, 1.40] p = 0.276	OR = 1.71 95% CI [0.69, 4.26] p = 0.248
Struggling	OR = 0.15 95% CI [0.07, 0.31] p < 0.001	OR = 0.39 95% CI [0.21, 0.70] p = 0.002	OR = 0.34 95% CI [0.19, 0.62] p < 0.001	OR = 1.35 95% CI [0.70, 2.61] p = 0.368

Note. TPI = Teacher Professional Identity; OR = odds ratio; CI = confidence interval. All confidence intervals are 95% confidence intervals. Model A uses the overall TPI composite. Model B enters Teacher Self-Efficacy, Commitment to Teaching, and Professional Orientation simultaneously. Classification error was incorporated using fixed wave-specific classification-error matrices derived from the LTA posterior probabilities. All models included 317 participants. Across the two waves, 24 coefficient tests were conducted without a multiplicity correction. Interpretation should therefore emphasise effect sizes, confidence intervals, and consistency across profiles and waves.

.

In Model A, the TPI composite was associated with lower odds of Struggling membership, OR = 0.22, 95% CI [0.12, 0.40], p < 0.001, and Need-Frustrated membership, OR = 0.49, 95% CI [0.25, 0.98], p = 0.044, relative to Flourishing. Its association with Moderately Thriving membership was not statistically significant, OR = 0.67, 95% CI [0.38, 1.19], p = 0.174.

In Model B, Teacher Self-Efficacy was associated with lower odds of Moderately Thriving, OR = 0.39, 95% CI [0.22, 0.71], p = 0.002; Need-Frustrated, OR = 0.41, 95% CI [0.19, 0.90], p = 0.025; and Struggling membership, OR = 0.16, 95% CI [0.08, 0.35], p < 0.001. Commitment to Teaching was associated with lower odds of Need-Frustrated, OR = 0.41, 95% CI [0.20, 0.82], p = 0.012, and Struggling membership, OR = 0.25, 95% CI [0.13, 0.48], p < 0.001. Its association with Moderately Thriving membership did not meet the conventional significance threshold, OR = 0.59, 95% CI [0.34, 1.02], p = 0.059. Professional Orientation showed positive adjusted associations with all three non-Flourishing profiles, with ORs ranging from 2.61 to 4.07.

Collinearity diagnostics did not indicate problematic multicollinearity. VIFs were 1.80 for Teacher Self-Efficacy, 2.22 for Commitment to Teaching, and 2.28 for Professional Orientation; corresponding tolerance values were 0.557, 0.451, and 0.439. Post hoc nested maximum-likelihood three-step models nevertheless showed a suppression pattern. When entered alone, Professional Orientation was associated with lower odds of Struggling membership relative to Flourishing, OR = 0.57, 95% CI [0.39, 0.83], p = 0.004, and was not significantly associated with Moderately Thriving or Need-Frustrated membership. In the full model, its coefficients became positive for Moderately Thriving, OR = 2.61, 95% CI [1.42, 4.80], p = 0.002; Need-Frustrated, OR = 2.78, 95% CI [1.26, 6.17], p = 0.012; and Struggling membership, OR = 4.07, 95% CI [1.95, 8.51], p < 0.001. This sign reversal is consistent with statistical suppression. Nevertheless, the positive adjusted coefficients remain conditional associations and should not be interpreted as evidence that valuing teaching is harmful.

• RQ4: Is teacher professional identity at Wave 1 prospectively associated with psychological need profile membership at Wave 3?

The maximum-likelihood three-step multinomial regressions were repeated with Wave 3 profile membership as the outcome and Wave 1 TPI scores as predictors, with Flourishing as the reference profile (Table 8). In Model A, the Wave 1 TPI composite was associated with lower odds of Moderately Thriving, OR = 0.33, 95% CI [0.16, 0.68], p = 0.003, and Struggling membership, OR = 0.15, 95% CI [0.07, 0.31], p < 0.001, at Wave 3 relative to Flourishing. Its association with Need-Frustrated membership did not meet the conventional significance threshold, OR = 0.44, 95% CI [0.18, 1.06], p = 0.067.

In Model B, Teacher Self-Efficacy was associated with lower odds of Need-Frustrated, OR = 0.45, 95% CI [0.22, 0.93], p = 0.030, and Struggling membership, OR = 0.39, 95% CI [0.21, 0.70], p = 0.002. Its association with Moderately Thriving membership did not meet the conventional significance threshold, OR = 0.57, 95% CI [0.32, 1.01], p = 0.054. Commitment to Teaching was associated with lower odds of Moderately Thriving, OR = 0.46, 95% CI [0.26, 0.83], p = 0.010, and Struggling membership, OR = 0.34, 95% CI [0.19, 0.62], p < 0.001, but not Need-Frustrated membership, OR = 0.65, 95% CI [0.30, 1.40], p = 0.276. Professional Orientation was not significantly associated with Wave 3 profile membership, with ORs ranging from 1.35 to 1.71 and p-values ranging from 0.209 to 0.368. These results indicate selective prospective associations rather than a uniform relationship between initial professional identity and later profile membership.

• RQ5: Are initial professional identity and change in professional identity from Wave 1 to Wave 3 associated with the direction of profile movement over the same period?

7.2. Trajectory Analysis

Participants were classified into trajectory groups based on their movement between ranked profiles from Wave 1 to Wave 3. Of the 317 participants, 170 (53.6%) were Stable, 97 (30.6%) Deteriorating, and 50 (15.8%) Improving. Initial TPI at Wave 1 was not significantly associated with deterioration relative to stability (OR = 1.038, 95% CI [0.764, 1.410], p = 0.811) or improvement relative to stability (OR = 1.319, 95% CI [0.849, 2.050], p = 0.218). In contrast, change in TPI from Wave 1 to Wave 3 was associated with trajectory group. Increases in TPI were associated with lower odds of deterioration (OR = 0.482, 95% CI [0.337, 0.689], p < 0.001) and higher odds of improvement (OR = 1.630, 95% CI [1.031, 2.576], p = 0.037) relative to stability. Because change in TPI and profile movement were assessed over the same period, these estimates identify associations and do not establish the temporal or causal direction between the two processes. Deterioration also varied descriptively by year of study (24.4–60.0%), though Year 4 (n = 22) and Year 5+ (n = 5) were too small for reliable interpretation.

7.2.1. Robustness Check: Full-Sample Estimation

As a robustness check against the complete-case restriction, the four-profile solution was re-estimated on the full available sample (N = 496) using the available-response likelihood procedure described above. Resulting profile means were largely consistent with the complete-case solution, with differences of 0.08 or less across all six indicators for three of the four profiles; the Flourishing profile showed somewhat larger differences on Autonomy Frustration (0.13) and Competence Frustration (0.24), though its overall configuration (high satisfaction, low frustration) remained substantively unchanged. The broad profile configuration was therefore similar in the complete-case and full-sample solutions. However, this robustness check does not establish that the missing data were ignorable or rule out bias in the transition and predictor estimates, which remained based on the 317 participants with complete observations across all three waves.

7.2.2. Robustness Check: Wave-Specific Profile Structure

The wave-specific latent profile analyses did not produce a uniformly stable model-selection pattern. BIC did not favour the four-profile solution at any of the three waves. Nevertheless, the four-profile model was reproduced in 36 of 50 starts at Wave 1 and 21 of 50 starts at Wave 2, whereas solutions with larger numbers of profiles at these waves were generally reproduced in only one or two starts. This provides partial support for the recoverability of a four-profile configuration at Waves 1 and 2, although it does not establish that four profiles were unequivocally optimal when each wave was analysed separately.

At Wave 3, no candidate solution showed strong replication. The four-profile solution was reproduced in only 1 of 50 starts. The solution obtained in that start descriptively resembled the longitudinal configuration in both shape and relative size, including a Flourishing-like profile characterised by near-ceiling satisfaction and floor-level frustration (n = 54, 17.0%) and a Struggling-like profile characterised by the lowest satisfaction and highest frustration (n = 109, 34.4%). However, because the four-profile solution was reproduced only once, this similarity cannot be treated as independent confirmation of the same profile structure at Wave 3. The instability may partly reflect the smaller amount of information available in a single-wave analysis, but alternative wave-specific structures cannot be ruled out. Overall, the wave-specific analyses provided partial support at Waves 1 and 2 but remained inconclusive at Wave 3. The retained four-profile solution should therefore be understood primarily as the structure supported by the complete longitudinal data rather than as a structure independently confirmed at every wave.

8. Discussion

This study examined how pre-service teachers’ psychological need experiences were organised across one academic year and whether teacher professional identity was associated with their profile membership and movement between profiles. Four main findings emerged. First, the retained solution identified four model-based profiles: Flourishing, Moderately Thriving, Need-Frustrated, and Struggling. Second, assigned profile membership showed moderate overall stability, although stability differed considerably across profiles. The Struggling profile was particularly persistent, whereas the Need-Frustrated profile showed the greatest movement. Third, Wave 1 teacher professional identity was associated with concurrent profile membership and prospectively associated with profile membership at Wave 3. Teacher Self-Efficacy and Commitment to Teaching both showed adjusted associations across profiles and waves, although the specific profile contrasts differed between the two dimensions. Professional Orientation was associated with profile membership only at Wave 1. Finally, initial professional identity did not reliably distinguish students who improved, deteriorated, or remained stable. Change in professional identity across the same period was associated with profile movement: increases in professional identity were linked to lower odds of deterioration and higher odds of improvement relative to stability. This finding shifts the interpretation away from professional identity as a fixed protective resource at the beginning of the study period and towards an association between changes in professional identity and changes in psychological need experiences. Because professional identity change and profile movement were assessed over the same period, however, the temporal direction of this association cannot be established.

The four-profile solution supports the value of examining psychological need experiences from a person-centred perspective. The Flourishing and Struggling profiles represented the clearest contrasting configurations. Flourishing combined high satisfaction with low frustration across the three needs, whereas Struggling combined comparatively low satisfaction with high frustration. The Flourishing profile was also partly characterised by displayed ceiling-level Relatedness Satisfaction and floor-level Relatedness Frustration. This concentration at the scale boundaries limits how finely its relatedness experience can be distinguished from the Moderately Thriving profile. The two intermediate profiles were more complex. Moderately Thriving participants reported relatively strong need satisfaction alongside some autonomy frustration. Participants assigned to the Need-Frustrated profile reported broadly similar levels of satisfaction, autonomy frustration, and competence frustration, but noticeably higher relatedness frustration. The distinction between these two profiles therefore rested mainly on whether positive need experiences coexisted with a stronger sense of relational exclusion or disconnection. This pattern is consistent with the self-determination theory distinction between need satisfaction and need frustration. Low satisfaction does not necessarily indicate active frustration, and satisfaction in one aspect of experience can coexist with frustration in another (B. Chen et al., 2015; Ryan & Deci, 2020). At the same time, the substantive distinction between these intermediate profiles should be interpreted cautiously because it depended largely on relatedness frustration, for which the measurement analyses identified a pronounced floor effect. The profile names are therefore used as descriptive summaries of model-based patterns rather than as fixed or objectively distinct types of pre-service teachers. The wave-specific analyses further qualify this interpretation. A comparable four-profile configuration was reproducible at Waves 1 and 2, but the standalone Wave 3 models were unstable. Although the four-profile solution obtained at Wave 3 descriptively resembled the longitudinal configuration, its limited replication means that it cannot be treated as independently confirmed. The retained profile structure is therefore supported mainly by the joint longitudinal analysis, with partial rather than uniform corroboration from the separate waves.

The prevalence of the Moderately Thriving profile is noteworthy. It was the largest profile at each wave, although its prevalence declined from 49.8% at Wave 1 to 34.1% at Wave 3. Participants assigned to this profile generally reported positive experiences of autonomy, competence, and relatedness, but their autonomy frustration was higher than that observed in the Flourishing profile. This suggests that broadly positive need experiences can coexist with some sense of pressure or constrained choice. The present study cannot identify which aspects of the programme contributed to this pattern. Nevertheless, the finding directs attention to the conditions through which teacher education is organised. Clear rationales for programme requirements, meaningful opportunities for student input, and some choice in how learning tasks are approached may support autonomy without removing necessary academic or professional expectations. These implications concern programme conditions rather than assumed deficits within the students assigned to this profile. They should also not be interpreted as evidence of a distinct cultural characteristic among Kuwaiti pre-service teachers.

The transition results add a developmental layer to this interpretation. Profile stability was moderate, with 54.6% of participants remaining in the same profile from Wave 1 to Wave 2 and 62.1% from Wave 2 to Wave 3. The Struggling profile was the most stable at both intervals, with retention rates of 84.9% and 84.0%, whereas the descriptive retention rate for Flourishing was higher in the second interval (68.3%) than in the first (50.0%). However, the heterogeneous transition model did not improve fit over the homogeneous model, so this interval difference should not be interpreted as evidence that the underlying transition structure changed over time. At the same time, the relatively low stability of the Need-Frustrated profile and its frequent movement towards Moderately Thriving suggest that assigned profile membership was not fixed for all students. Most observed transitions were between adjacent profiles rather than across large distances in the profile space. This pattern is compatible with gradual movement in need satisfaction and frustration, although it does not establish a particular developmental mechanism. This finding extends previous person-centred SDT research by illustrating both continuity and movement in assigned need profiles across an academic year (Warburton et al., 2020; Krijgsman et al., 2025).

The increase in the Struggling profile from Wave 1 to Wave 3 also deserves careful attention. The proportion of participants assigned to this profile rose from 16.7% at Wave 1 to 31.5% at Wave 3. These percentages were calculated for the same 317 complete-case participants at each wave, so the increase reflects a change in the distribution of their decoded profile assignments rather than a change in the composition of the sample. However, participants entered the study from different programme years. The aggregate pattern may therefore combine different year-of-study trajectories and cannot be interpreted as a common stage-related change across the programme. The small fourth- and fifth-year groups also prevented reliable cohort-specific comparisons. Although the study was not designed to isolate the causes of this shift, the pattern indicates that comparatively low need satisfaction and high need frustration became more prevalent within the sample over the academic year. It does not show that programme exposure caused the change because specific programme experiences were not measured. The finding resonates with work showing that teacher identity and motivation may not develop in a simple upward direction during preparation (C. Chen et al., 2023; Wong & Liu, 2025). Becoming a teacher can strengthen commitment for some students, but it can also make the constraints and emotional demands of the profession more visible. The practical implication is that teacher education programmes should not assume that students who enter with generally positive psychological need experiences will remain psychologically supported throughout the programme. Periodic attention to students’ need satisfaction and need frustration may therefore be useful, particularly around potentially demanding periods such as practicum. However, the present findings do not identify which programme stages or experiences were responsible for the observed changes. Future longitudinal research should align measurement waves with practicum, assessment, and other programme milestones and measure these experiences directly. This would allow changes in psychological need profiles and professional identity to be interpreted in relation to the conditions occurring between waves.

The regression findings provide a differentiated account of teacher professional identity. At Wave 1, the composite TPI score was associated with lower odds of Need-Frustrated and Struggling membership relative to Flourishing, but not with Moderately Thriving membership. At Wave 3, initial TPI was prospectively associated with lower odds of Moderately Thriving and Struggling membership, but not Need-Frustrated membership. The composite findings therefore varied across profiles and waves rather than showing a uniform association with every non-Flourishing profile.

The dimension-level analyses further qualify this pattern. At Wave 1, Teacher Self-Efficacy was associated with all three profile contrasts, whereas Commitment to Teaching was associated with Need-Frustrated and Struggling membership. The association between Commitment to Teaching and Moderately Thriving membership was close to the conventional significance threshold. At Wave 3, Teacher Self-Efficacy was associated with Need-Frustrated and Struggling membership, while its association with Moderately Thriving membership was close to the conventional threshold. Commitment to Teaching was associated with Moderately Thriving and Struggling membership but not Need-Frustrated membership. These findings suggest that capability beliefs and commitment may relate differently to particular configurations of psychological need experiences. However, the study did not measure the processes through which these associations might operate and cannot establish either dimension as a causal antecedent or protective mechanism.

Professional Orientation followed a different pattern. When the three identity dimensions were entered simultaneously, higher Professional Orientation was associated with greater odds of all three non-Flourishing profiles at Wave 1, but none of these associations was evident at Wave 3. The VIF and tolerance values did not indicate problematic multicollinearity, but the nested maximum-likelihood three-step models showed that the direction of the Professional Orientation coefficients changed as the other identity dimensions were introduced. When Professional Orientation was entered alone, it was associated with lower odds of Struggling membership and was not significantly associated with the Moderately Thriving or Need-Frustrated contrasts. The positive coefficients emerged only after Teacher Self-Efficacy and Commitment to Teaching were controlled. This sign reversal is consistent with statistical suppression and indicates that the full-model coefficients represent conditional associations rather than the overall relationship between Professional Orientation and profile membership.

These diagnostics clarify the statistical pattern, but they do not establish its substantive mechanism. The results should not be interpreted as evidence that valuing teaching as a profession is harmful. Moreover, the nested models were conducted post hoc, and the primary profile-membership analyses involved 24 coefficient tests without a multiplicity correction. The suppression pattern therefore requires replication before a stronger substantive interpretation can be offered. Future research should examine Professional Orientation separately and alongside the other identity dimensions across different teacher education contexts.

At the same time, the findings do not justify a simple protective interpretation of professional identity. Initial TPI was associated with selected concurrent and prospective profile contrasts, but it was not significantly associated with whether students’ profiles improved or deteriorated rather than remained stable. This null association does not show that baseline professional identity was irrelevant. It indicates more narrowly that a single Wave 1 composite did not distinguish the subsequent direction of profile movement. Students who began with similar levels of professional identity may have encountered different feedback, relationships, practicum demands, or personal circumstances during the year. These unmeasured experiences could have influenced both their psychological need experiences and their professional identity after baseline. The composite score may also conceal different roles of self-efficacy, commitment, and professional orientation. Initial identity may therefore be associated with profile membership at particular waves without operating as a durable buffer against later movement.

The trajectory analysis makes this point clearly: 53.6% of participants were stable, 30.6% deteriorated, and 15.8% improved. Although initial TPI did not reliably explain these trajectory differences, change in TPI across the study period did. Increases in TPI were associated with lower odds of deterioration (OR = 0.482) and higher odds of improvement (OR = 1.630) relative to stability. Observed change in professional identity was therefore more closely associated with the direction of profile movement than identity measured at Wave 1. However, because TPI change and profile movement were assessed over the same interval, this pattern cannot establish whether identity changed first, whether changing need experiences shaped identity, or whether both processes responded to other educational or personal experiences.

Several limitations should shape the interpretation of these findings. The LTA was conducted on complete cases, and attrition analyses indicated small differences between completers and non-completers in several need satisfaction indicators. Although the multivariable attrition model did not identify systematic predictors of completion, selective attrition cannot be fully ruled out. The full-sample robustness check produced a broadly similar four-profile structure, which provides some reassurance that the complete-case restriction did not substantially alter the profile configurations. However, this check does not remove the possibility of bias in the transition and predictor estimates, which were based on participants with complete data across all three waves. The reliability of Autonomy Satisfaction at Wave 1 was also below the conventional threshold, which means that findings involving autonomy at the first wave should be interpreted with caution. Because Autonomy Satisfaction was one of the six indicators used to estimate the profiles, this measurement weakness may have attenuated individual differences and affected profile assignment and estimated movement involving Wave 1. No indicator-level sensitivity analysis excluding Autonomy Satisfaction was conducted; therefore, the robustness of the four-profile structure to this limitation remains uncertain.

Longitudinal measurement invariance was also incomplete. Metric invariance was supported for all six need dimensions, but full scalar invariance was established only for Competence Satisfaction. Partial scalar invariance was obtained for Autonomy Satisfaction, Autonomy Frustration, and Relatedness Satisfaction, while scalar invariance was not established for Relatedness Frustration or Competence Frustration. Residual-variance invariance, sometimes described as strict invariance, was not tested in the repeated-measures CFA. Although the LTA constrained profile-specific indicator variances to remain equal across waves, this model constraint does not replace a separate test of strict measurement invariance. Consequently, some apparent differences across waves may partly reflect changes in measurement rather than substantive change in students’ experiences.

The LTA also relied on the local independence assumption, under which indicator covariances within profiles were fixed to zero. This is a standard but restrictive specification because residual associations among the satisfaction and frustration indicators beyond those explained by profile membership were not modelled. The profile-membership regressions incorporated classification uncertainty using fixed wave-specific classification-error matrices. Their estimates therefore remain dependent on the adequacy of the retained profile solution and its classification-error structure.

Relatedness Frustration presented an additional measurement concern. A substantial proportion of participants reported the minimum score across all three waves, and one item was removed following the confirmatory factor analysis. The resulting improvement in model fit supported the statistical decision, but removing the item may have narrowed the content represented by this subscale. Accordingly, the retained three-item score should not be assumed to represent the full breadth of relatedness frustration. Profile differences involving this indicator should be interpreted as reflecting the more limited content captured by the retained items. The floor effect also restricts variation and warrants particular caution when interpreting the distinction between Moderately Thriving and Need-Frustrated, which depended largely on relatedness frustration. In addition, the study relied on self-report measures and an observational design. The results therefore show associations between professional identity and need-profile membership, not causal effects. Although Wave 1 professional identity preceded Wave 3 profile membership, this temporal ordering does not rule out unmeasured explanations. Moreover, change in professional identity and profile movement were assessed over the same period, so their direction of influence cannot be determined. The sample was drawn from one national and institutional context, with a predominantly female and Kuwaiti sample, so generalisation to other teacher education systems should be made carefully. Programme structure, practicum timing, and supervision arrangements were not measured systematically; consequently, the study cannot determine whether particular programme experiences account for the observed profile transitions. The gender imbalance also limited the feasibility of adequately powered comparisons of profile formation and transition patterns by gender. Finally, profile ranking for the trajectory analysis was necessarily interpretive, even though it was grounded in overall need satisfaction and frustration patterns. The frustration-only sensitivity check produced the same ordering, but this does not make the ordering objective. Although the maximum-likelihood three-step procedure accounted for classification uncertainty in the profile-membership regressions, the trajectory groups still depended on assigned profile membership and the chosen ordering of the four profiles. The improving, deteriorating, and stable categories should therefore be understood as analytically constructed summaries rather than fixed developmental pathways.

Despite these limitations, the study offers several contributions. Conceptually, it brings teacher professional identity research into closer dialogue with self-determination theory by showing that identity resources are associated with patterned need experiences across time. It also shows that the dimensions of professional identity are not uniformly associated with these experiences. Teacher Self-Efficacy showed the broadest adjusted pattern across the two waves, while Commitment to Teaching was associated with selected profile contrasts at both waves. Professional Orientation showed a Wave 1-specific pattern that did not persist at Wave 3. Methodologically, it illustrates the usefulness of latent transition analysis for identifying configurations of need satisfaction and frustration and describing patterns of stability and movement between them. Practically, the findings suggest that teacher education programmes should look beyond average levels of student motivation. A cohort may report generally positive need experiences while still including students who experience active frustration or a combination of low satisfaction and high frustration. Identifying these profiles can help programmes design more targeted support. However, the model-based profiles should not be used as diagnostic labels or treated as fixed categories of students. Instead, they can draw attention to different aspects of the programme environment that may warrant closer examination.

For students whose experiences resemble the Moderately Thriving pattern, programmes might retain existing sources of competence and relatedness satisfaction while providing clearer rationales for requirements, meaningful opportunities for choice, and greater student voice. The Need-Frustrated pattern points more specifically to the possible value of dependable mentoring relationships, respectful supervisory feedback, and opportunities for genuine peer connection. This implication should remain cautious because the distinction involving relatedness frustration was affected by the measurement limitations noted above. The Struggling pattern may call for a broader review of programme demands and support because it combined comparatively low satisfaction with high frustration across the three needs. Support could therefore include manageable opportunities to develop competence, accessible academic guidance, and attention to whether students feel heard and included. These suggestions follow from self-determination theory and the observed profile configurations, but the present study did not test whether any particular intervention would produce profile improvement.

9. Conclusions

This study showed that pre-service teachers’ psychological need experiences across teacher education are heterogeneous, moderately stable, and capable of changing across an academic year. Four model-based profiles were identified across three waves: Flourishing, Moderately Thriving, Need-Frustrated, and Struggling. The label Moderately Thriving is comparative: participants in this profile reported high need satisfaction in absolute terms but somewhat greater need frustration than those in Flourishing. The Struggling profile was particularly persistent and became more prevalent by Wave 3, whereas Need-Frustrated showed the greatest movement between profiles. The findings suggest that many students do not simply thrive or struggle in a general sense. Instead, they may report relatively positive satisfaction alongside some autonomy frustration or experience stronger relatedness frustration while retaining broadly positive satisfaction in other areas. Others may experience a more consistently difficult pattern marked by comparatively low need satisfaction and high need frustration. These profiles should be understood as descriptive configurations rather than fixed types of pre-service teachers.

Wave 1 teacher professional identity was associated with concurrent profile membership and prospectively associated with selected profile membership at Wave 3. Teacher Self-Efficacy and Commitment to Teaching both showed adjusted associations across profiles and waves, although the profiles they distinguished were not identical. Professional Orientation showed positive adjusted associations with the non-Flourishing profiles at Wave 1 but not at Wave 3. Post hoc nested maximum-likelihood three-step models indicated that these positive coefficients emerged through a sign reversal after Teacher Self-Efficacy and Commitment to Teaching were controlled, a pattern consistent with statistical suppression. These adjusted associations should therefore not be interpreted as evidence that Professional Orientation itself is harmful. Initial TPI did not reliably distinguish students who improved, deteriorated, or remained stable over time. By contrast, change in professional identity across the same period was associated with profile movement. Increases in TPI from Wave 1 to Wave 3 were associated with lower odds of deterioration and higher odds of improvement relative to stability. This distinction is important. Professional identity was associated with students’ profile membership, while change in professional identity across the study period was associated with the direction of profile movement. Because identity change and profile movement were assessed over the same period, this association does not demonstrate that strengthening professional identity caused students to move towards more adaptive profiles. Changing educational conditions, practicum experiences, institutional demands, or other forms of support may have influenced both processes, but these possibilities were not directly examined in the present study.

Overall, the findings argue for a more nuanced view of pre-service teacher development. Professional identity appears to be an important individual resource, but the findings do not establish it as a stable protective factor or as a substitute for need-supportive teacher education. Programmes that wish to support future teachers may need to attend not only to students’ beliefs about teaching, but also to the everyday conditions through which autonomy, competence, and relatedness are either supported or frustrated. From this perspective, the central question is not only which students report a strong professional identity at a given point in teacher education, but also how programme conditions relate to students’ psychological need experiences and their continuing identity development. This interpretation recognises professional identity and institutional conditions as connected but distinct parts of pre-service teacher development. The present findings indicate associations between these processes, while further research is needed to establish how they influence one another over time.