Contribution
It is the first formal test, in the authors’ words, of self-critical perfectionism as a transdiagnostic risk factor for depressive and eating disorder symptoms, and it adds need frustration as the process in between. It applies Nolen-Hoeksema and Watkins’s (2011) conditions: a transdiagnostic factor predicts both symptom kinds and reduces their association when modelled as their common predictor. The test is run between persons (Study
- and on within-person change (Study 2).
Key insight
The transdiagnostic question is posed as a comorbidity question: does the candidate process absorb the association between two kinds of symptom? Need frustration does, much more than perfectionism does. Perfectionism “could only explain a small part of their co-occurrence” (p. 11); with need frustration in the model, depression and eating pathology “were not (Study 1) or were only minimally (Study 2) related to each other” (p. 11).
Assumptions
- Measures (p. 5). Self-critical perfectionism: the Frost MPS’s Concern over Mistakes (9 items) and Doubts about Actions (4). Need frustration: “the frustration items from the Basic Psychological Need Satisfaction and Need frustration Scale”, 12 items, 4 per need. Depressive symptoms: the Children’s Depression Inventory, 27 items. Eating pathology: the EDI-II’s Drive for Thinness, Bulimia and Body Dissatisfaction, aggregated.
- Latent variables from parcels (three per construct in Study 1, two in Study 2), built by the item-to-construct balance technique (note, p. 13).
- Covariates: age, gender, age- and gender-adjusted BMI, in all models; in Study 2 also the level of each variable.
- Study 2’s design (pp. 7–8): two waves, six months apart; latent change models in which change scores are correlated or regressed on each other. Both “predictors” and “outcomes” are changes over the same interval.
- Non-clinical samples. Caucasian, middle-class Belgian secondary-school students; Study 2 academic track only.
Key results
Study 1, cross-sectional (pp. 4–7)
- N = 284 in the text (p. 4), “N = 248” in the abstract; 58% girls, mean age 14.1.
- Measurement model: CFI .99, RMSEA .031, SRMR .03, χ²/df = 61.43/48.
- Depression and eating pathology correlate r = .50 (latent).
- Perfectionism → eating β = .34, → depression β = .45; the symptom correlation falls to .42.
- With frustration as full mediator (CFI .98, RMSEA .04, χ²/df = 114.48/75), partial mediation does not fit better, Δχ²(2) = 3.08, ns. Indirect effects via frustration: eating β = .28 [.18, .39], depression β = .55 [.41, .69]. The residual symptom correlation is .18, p > .05.
Study 2, two waves (pp. 7–11)
- N = 608 at T1, 477 at both; 59% girls, mean age 16.19. Older students dropped out more (p. 8); study variables did not predict attrition.
- Only depressive symptoms changed in mean (−.09, p < .001; Table 1).
- Change correlations (Table 2): frustration change with depression change .55, with eating change .38; perfectionism change with frustration change .31.
- Correlated change of the two symptoms β = .43; with perfectionism change as predictor .39; with frustration change as mediator, partial mediation fits better than full (Δχ²(2) = 17.34, p < .001; CFI .97, RMSEA .06, SRMR .06), and the correlated change falls to .24, p < .01. Indirect effects via frustration change: depression β = .26 [.17, .36], eating β = .12 [.06, .19].
The authors’ reading (pp. 11–13)
- “Need frustration represents a transdiagnostic vulnerability process that helps to explain why self-critical perfectionism relates to depressive symptoms and eating disorder symptoms” (abstract).
- Limitations they state: “Due to the correlational nature of these findings, future experimental research is needed to examine whether increases in need frustration actually cause increases” (p. 12); generalization to clinical samples and adults untested; externalizing problems untested.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Need frustration statistically accounts for most of the association between depressive and eating disorder symptoms in adolescents | moderate | latent models in two samples; Study 1 residual r = .18 (ns) from .50; Study 2 correlated change .24 from .43 |
| C2 | Need frustration mediates self-critical perfectionism’s link to both symptom kinds | moderate | indirect effects with CIs in both studies; full in Study 1, partial for eating in Study 2 |
| C3 | Need frustration is a transdiagnostic process | weak | two internalizing symptom domains in community adolescents; mediation in cross-sectional and same-interval change data |
| C4 | Need frustration precedes the symptoms | not claimed | the design cannot show it: Study 2’s mediator and outcomes are changes over the same six months; the authors call the findings correlational |
Concepts
- Transdiagnostic risk factor — a variable that predicts several kinds of disorder and reduces their association when modelled as a common predictor (Nolen-Hoeksema & Watkins 2011).
- Transdiagnostic process — a mediator that explains why such a factor predicts several disorders and further reduces their association.
- Correlated change — covariation of two variables’ latent change scores between two waves; “within-person change” in the paper’s sense.
Connections
- Boone et al. 2014, the longitudinal precedent: self-critical perfectionism predicted increases in need frustration, which predicted increases in binge eating.
- Bartholomew, Ntoumanis, Ryan, Bosch and Thøgersen-Ntoumani 2011 (PSPB), on need frustration with depression and disordered eating in athletes; a different paper from LIT-1954.
- Chen et al. 2015: the BPNSFS, whose frustration items are the measure.
Bearing on the record
- THEORY-052 and LIT-561. THEORY-052 quotes LIT-561 p. 12: Campbell et al. (2018a) found that need frustration “not only predicted adolescents’ depressive symptoms and eating pathology, but also largely accounted for their co-occurrence and co-evolution across time”. Read against the paper, “accounted for their co-occurrence and co-evolution” is right: the symptom correlation and the symptoms’ correlated change both shrink with frustration in the model. “Predicted” is right only statistically. There is no lagged prediction: the longitudinal study has two waves, and its frustration and symptom variables are changes over the same six months. THEORY-052’s note of 2026-10-28 calls it “the one longitudinal study cited”; it is longitudinal in having two waves, and gives no evidence of temporal order. On the promote_when: it is not prospective prediction of onset; it covers two internalizing domains, “more than one kind”, but not externalizing; and the frustration measure is the BPNSFS frustration scale, which Murphy et al. (LIT-1978, NOTE-1530) argue is a keyed low-satisfaction score (THEORY-1311), with items (“I feel like a failure because of the mistakes I make”) close in content to both the perfectionism predictor and depressive symptoms. So it does not meet the promote_when, and THEORY-052 stays where it is.
- The comorbidity framing is the strongest form of the transdiagnostic claim in the record: not “frustration correlates with each disorder” but “frustration absorbs the association between disorders”. It could have failed (a residual symptom correlation near .50 would have), and in Study 1 it did not. That is worth stating in THEORY-052 alongside the measurement objections.
- THEORY filed from this reading: THEORY-1322 (Active), the comorbidity result and its design.
- No instruction for machine-learning practice.
Limitations
- No temporal order (C4). Mediation by same-interval change scores cannot say whether frustration changes first.
- Shared method and content. All four constructs are adolescent self-reports; the BPNSFS frustration items include failure and pressure content close to the Frost subscales and to the CDI.
- Community adolescents, one country, internalizing symptoms only.
- The sample size differs between the abstract (248) and the text (284) for Study 1.
- Not seen: the Supporting Information’s descriptives and correlations, and the coefficients drawn only in Figures 1–3.
Open questions
- Does frustration at one wave predict symptom change to the next, with three or more waves and random intercepts?
- Does the comorbidity result hold with a frustration measure that has thwarting content (the PNTS) or balanced keying?
- Does it extend to externalizing problems, which the authors name as the next test (p. 12)?