Contribution
It asks whether need frustration predicts depressive symptoms once negative affect, which overlaps both, is in the same model, and whether it comes before them or with them. Two student samples, one an ordinary academic year and one the first six weeks of the pandemic, give the same answer: of the three variables, need frustration was the one whose earlier score predicted the others’ later scores.
Key insight
The study tests incremental validity over time against the nearest rival, general negative affect, rather than against low satisfaction. In both samples the frustration → later depression and frustration → later negative affect paths are significant, and the negative affect → later depression path is not (Tables 2 and 4). The authors read this as “need frustration is an antecedent of depressive symptoms over time” (abstract).
Assumptions
- Measures. Negative affect: five adjectives such as “unhappy” and “frustrated”, cited to Emmons (1992), 7-point. Need frustration: “the 9-item Balanced Measure of Psychological Needs scale (BMPN)”, three items per need (“I feel lonely”; “I experienced some kind of failure or was unable to do well at something”; “I feel pressured in my life”), averaged into one score. Depressive symptoms: the CESD-R 10, averaged.
- The BMPN items are its dissatisfaction half. The BMPN has 18 items, satisfaction and dissatisfaction (Kermavnar et al., Table 3); the paper uses nine negatively worded ones and calls them frustration. Satisfaction was not modelled.
- Model. Fully cross-lagged path models in Mplus, observed variables, FIML; standardized coefficients (STDYX). Within-wave covariances estimated. No random intercepts in the reported model: footnote 3 says a latent SEM and a random-intercept model were tried, and are on OSF.
- Samples. Study 1: a Canadian university, Mage 20.44, 85% female; 80% retained across three waves (T1 379, T2 333, T3 304). Study 2: the same university, 83% female, two waves in April and late May 2020; 242 and 200 combined family and peer reports.
Key results
Study 1, the academic year (Table 2)
- Autoregressive paths all significant (e.g. depression .46 then .26; frustration .43 then .32).
- Need frustration → later depression: T1→T2 β = .13 [.03, .25], p = .01; T2→T3 β = .17 [.03, .31], p = .02.
- Need frustration → later negative affect: T1→T2 .18 [.06, .29], p = .002; T2→T3 .16 [.02, .31], p = .02.
- Negative affect → later depression: −.01, p = .85; .10, p = .25.
- Depression → later need frustration: T1→T2 .20 [.05, .34], p = .007; T2→T3 .01, p = .88. Depression → later negative affect: .19, p = .007; .01, p = .97.
- Negative affect → later need frustration: T1→T2 −.07, p = .27; T2→T3 .17 [.01, .34], p = .04.
- Fit: χ²(9) = 36.77, p < .001, RMSEA = .09 [.06, .12], CFI = .979, SRMR = .048, called “good fit”.
- Within-wave correlations of frustration with depression .56, .67, .62 (Table 1).
Study 2, the pandemic (Table 4)
- Need frustration → later depression .18 [.04, .32], p = .010; → later negative affect .17 [.03, .30], p = .018.
- Negative affect → later depression .07, p = .287; depression → later frustration −.01, p = .987; negative affect → later frustration .09, p = .287. “The only significant cross-lagged predictor was need frustration.”
- The model is just-identified, so no fit is reported.
- Others’ reports (Table 5). Predicting May 2020 other-reported depression (“is someone who is depressed”, “can be moody”; α = .62, .67), controlling for the September 2019 report: baseline .62, p < .001; self-reported frustration .14 [−.01, .29], p = .074; self-reported depression .08, p = .397; negative affect .03, p = .743. R² = .486.
The authors’ conclusions (Discussion)
- Need frustration is “an antecedent and concomitant of depressive symptoms”, and the research “further implicates need frustration as a transdiagnostic risk factor, as it was both a predictor of negative affect and depressive symptoms longitudinally”.
- The others’ reports “remove the problem of shared method variance that plagues self-report studies in this area”.
- Limitations they state: personality not controlled (Nishimura & Suzuki 2016 found it attenuates need–well-being links); one affect measure; samples underpowered for a full latent SEM; “cannot be used as evidence for causation”.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Earlier need frustration predicts later depressive symptoms, controlling for earlier depression and negative affect, in students | moderate | three significant lagged paths over two samples, β .13–.18; self-report; observed-variable CLPM without random intercepts |
| C2 | Negative affect does not predict later depressive symptoms once need frustration is in the model | moderate | non-significant in all three lags; one five-adjective affect measure |
| C3 | Need frustration precedes depression rather than following it | weak | the reverse path was significant at the first lag of Study 1 (β = .20, larger than the forward .13) and not elsewhere; two-wave Study 2 cannot separate them |
| C4 | Need frustration predicts depression as seen by others | weak | β = .14, p = .074, two short items with α .62 and .67 |
| C5 | Need frustration is a transdiagnostic risk factor | weak | the outcomes are depressive symptoms and negative affect, one disorder and a mood state; asserted in the discussion |
Concepts
- Need frustration (as measured here) — the mean of the BMPN’s nine negatively worded items, described as “the active thwarting of one’s needs” in the introduction.
- Antecedent / concomitant — a variable that predicts another’s later value net of its own earlier value, versus one correlated with it at the same time.
- Fully cross-lagged path model — every variable at each wave regressed on every variable at the previous wave.
Connections
- Campbell et al. 2018 (LIT-1964) is cited as showing need frustration “mediates the relationship between negative environments and depressive symptoms”, and among works using it as a mediator. Its sample was self-critical perfectionism, not environments.
- Cordeiro et al. 2016 and Vermote et al. 2021 are named as the only prior multi-wave studies of need frustration and ill-being.
- Sheldon and Hilpert 2012: the BMPN.
Bearing on the record
- THEORY-052. It is prospective evidence the record lacked, and it meets part of the promote_when (“prospective studies in which frustration … predicts”) and fails the rest. One kind of disorder, not more: the outcomes are depressive symptoms and negative affect, so the authors’ “transdiagnostic” is unsupported by their own design. The measure is a self-report of the kind the promote_when excludes: “I feel lonely” is a frustration item here, and loneliness is a depressive symptom on the CES-D 10 (that the CESD-R 10 used has a loneliness item is from the instrument, not this text). The outcome is symptom level, not first onset. And the reverse path, depression → later frustration, was as large as the forward one at the first lag of Study 1, which is the “vicious cycle” THEORY-052 already notes from LIT-561. It does strengthen the claim that frustration predicts depression beyond negative affect, which the theory’s “What this does not say” does not yet record.
- LIT-1978 (Murphy et al.). Levine’s frustration score is the BMPN’s negative half with no satisfaction half modelled, and the rival it beats is also a negatively valenced self-report. So the keying objection does not bite in its BPNSFS form (no satisfaction comparison is made), but the result is “a negatively worded need report predicts a negatively worded symptom report six weeks to four months later”, and does not show that what predicts is frustration rather than low satisfaction.
- CLAIM-368 separates frustration from distress as constructs in an organization. This paper separates their measures in individuals: need frustration and negative affect correlate .40–.67 within waves, yet their lagged roles differ. That is a parallel, at the individual and felt level, for the claim’s “distinct from distress”; it is not evidence for a claim about organizations, and the claim does not need it.
- QUESTION-369. Frustration predicting depression is the well-being route to need-hood again: a symptom report, not a functional deficit.
- THEORY filed from this reading: THEORY-1349 (Active), the reported lagged paths and what they do not show.
- No instruction for machine-learning practice.
Limitations
- No random intercepts in the reported models. A traditional CLPM mixes stable between-person differences into its lagged paths; the random-intercept model the authors tried is only on OSF. The lagged coefficients are therefore not within-person effects.
- Students, mostly women, at one Canadian university; Study 2 at a single six-week window of the pandemic.
- Fit. Study 1’s RMSEA of .09 is at the poor end of the usual range, though called good; Study 2’s model is saturated.
- Study 2’s Table 3 repeats Study 1’s descriptives. Its means and SDs for T1 and T2 depression, negative affect and frustration (2.06 (0.51), 3.50 (1.08), 3.64 (1.10), 2.29 (0.61), 3.95 (1.28), 4.02 (1.25)) are identical to Table 1’s for Study 1’s first two waves, while the correlations differ. One of the two is a copying error; the paper does not say which. Study 2’s sample is also described as “this year-long online study” though its waves were six weeks apart (it was part of a longer project).
- The affect measure. Five adjectives, among them “frustrated”, cited to Emmons (1992); one negative-affect item names frustration, so the two predictors share a word.
- Others’ reports are two Big Five items averaged over family and peers, α .62 and .67, and the frustration effect on them is not significant at .05.
Open questions
- Do the forward paths survive a random-intercept cross-lagged model? The OSF supplement may say.
- Does frustration predict depression beyond low satisfaction, with keying balanced? The design leaves satisfaction out.
- Does the same score predict a second kind of disorder, or first onset?