Written and maintained by CASRAI Editorial Board
Published
On this page
The PSWQ (Penn State Worry Questionnaire) is a self-report measure of trait worry. Its standard 16-item form uses responses from 1 to 5, reverses five items and produces a total from 16 to 80. For research, identify the exact form before scoring: abbreviated questionnaires, cohort-specific coding and the past-week adaptation have different ranges.
Definition: Penn State Worry Questionnaire
Meyer, Miller, Metzger and Borkovec developed the PSWQ in 1990 to measure a person’s tendency to worry. Their original validation report describes a 16-item instrument selected through factor analysis, with evidence of internal consistency, stability over time and relationships with other psychological measures.
The construct matters when choosing a research outcome. The PSWQ addresses characteristics of worrying rather than providing an inventory of particular worry topics. Stöber’s research distinguishes this approach from the Worry Domains Questionnaire, which asks about specific areas of concern. A study of how persistently people worry therefore asks a different measurement question from a study of what they worry about.
Keep that distinction in the protocol and data dictionary. Name the instrument, language, item set and scoring convention alongside the outcome variable. A column labelled only “worry” conceals choices that another researcher needs to reproduce the score.
PSWQ scoring formula and reverse-scored items
For the full 16-item form coded 1–5, reverse items 1, 3, 8, 10 and 11. Keep the other eleven responses unchanged, then sum all sixteen scored items. The University of Cambridge’s CamCOPS implementation documents these positions and the resulting 16–80 range.
R = {1, 3, 8, 10, 11}; sᵢ = 6 − xᵢ for i in R, otherwise sᵢ = xᵢ; T = Σᵢ₌₁¹⁶ sᵢ.
Here, i is the item number in the original questionnaire; xᵢ is the recorded response before reversal; R is the set of reverse-scored positions; sᵢ is the scored response; Σ means add the sixteen values; and T is the final total. The constant 6 comes from the response endpoints: 1 + 5 = 6. Thus reversal maps 1 to 5, 2 to 4, 3 to 3, 4 to 2 and 5 to 1.
The allowable range follows directly: sixteen scored responses of 1 sum to 16, and sixteen scored responses of 5 sum to 80. Those endpoints concern scored responses, not a raw response sheet with identical numbers throughout. Apply reversal once, and preserve the original responses separately from the derived variables.
Worked example: a complete PSWQ score
This is a synthetic scoring example, not an observed participant record or a diagnostic case. Suppose the raw responses, in original item order 1–16, are:
2, 4, 1, 3, 4, 5, 3, 2, 4, 1, 2, 4, 3, 5, 4, 3.
The five reverse-item responses are 2, 1, 2, 1 and 2. Transforming them gives 4, 5, 4, 5 and 4, whose subtotal is 22. The eleven unchanged items contribute 4 + 3 + 4 + 5 + 3 + 4 + 4 + 3 + 5 + 4 + 3 = 42. Therefore T = 22 + 42 = 64.
A second calculation checks the arithmetic. The raw response sum is 50; the raw reverse-item subtotal is 8. Replacing each of five responses x with 6 − x changes the sum by 30 − 2 × 8. Hence T = 50 + 30 − 16 = 64, agreeing with the item-by-item calculation.
A useful check record should store both results and require agreement. It should also confirm that the input contains exactly sixteen valid responses. A plausible-looking total alone cannot reveal a missed reversal: in this example, simply adding the raw responses would give 50.
Interpreting PSWQ scores in research
Within the same correctly scored form, larger totals indicate more worry. Interpret a total with its instrument version and study population; the possible range by itself does not establish diagnostic categories. The table gives reading guidance rather than clinical severity bands.
| Result or finding | Research interpretation | Basis |
|---|---|---|
| Full-form total of 64 | A value on the 16–80 scale; report the form and coding before making comparisons. | CamCOPS scoring; synthetic example above. |
| Value outside 16–80 | Incompatible with sixteen complete, correctly scored 1–5 responses; inspect the inputs and form. | Derived from the full-form endpoints. |
| Eight-item total of 32 | Potentially valid on TILDA’s 8–40 scale; it is not a full-form total. | TILDA scale documentation. |
| A published sample mean | A description of that sample, not automatically a boundary separating categories. | Stöber’s sample-specific descriptive statistics. |
For analysis, retain the continuous score and document any grouping rule separately. If a protocol uses a threshold, identify the validation study, population and purpose supporting it. Do not turn a score’s position between its mathematical endpoints into a diagnosis.
Psychometric properties: reliability and validity
Evidence for PSWQ measurement should be attached to a sample and design. In Stöber’s 1998 study of 148 nonclinical students, the PSWQ had Cronbach’s α = .89 and a four-week test–retest correlation of .87. Its mean was 43.78, with SD = 10.09; observed scores ranged from 24 to 71.
That study also obtained ratings from three acquainted peers. The correlation between self-ratings and averaged peer ratings was .55. This adds a different perspective from repeating the same self-report: temporal stability and agreement across reporters address different questions. See the explanations of Cronbach’s alpha and convergent and discriminant validity when planning the evidence to report.
Factor structure also requires attention. Hopko and colleagues’ 2003 study in older adults with generalized anxiety disorder found poor fit for established one- and two-factor models. Their modified model removed eight items and showed improved fit with reported reliability and validity evidence. That result supports specifying the tested version and population rather than assuming that one published model applies unchanged everywhere.
PSWQ versions and response coding
The data table below compiles four documented implementations or adaptations. Its ranges are either explicitly documented or calculated from the cited item counts and endpoints. The rows describe distinct measurement specifications; they are not a conversion table for interchangeable outcomes.
TILDA documents an eight-item form with responses 1–5, summed to 8–40, without negatively worded items. Its variable numbering belongs to that short form. Consequently, applying the full questionnaire’s five reverse-item positions to those eight columns would mis-score the data.
The ADNI-D procedures manual instead documents sixteen items coded 0–4, retaining the five reverse-item positions. That coding yields a possible sum of 0–64, with reversal calculated as 4 − x. For an otherwise identical complete response set, changing every response category from 0–4 to 1–5 increases the scored total by sixteen: T₁–₅ = T₀–₄ + 16. This is a coding identity, not evidence that different adaptations measure identically.
The PSWQ-Past Week studied by Stöber and Bittencourt changes more than the reference period. It uses fifteen past-tense items, removes the original lifetime-worry item and uses responses 0–6. The resulting possible range is 0–90. Their study involved 28 nonclinical high worriers and reported average α = .91. Replacing the instructions on a standard questionnaire with “past week” does not recreate that validated adaptation.
Assumptions and common scoring mistakes
The calculation assumes that the response columns correspond to the stated item order and coding. Before deriving scores, compare the export against the actual administered form. In particular, establish whether an assessment platform exported raw responses or already reversed item scores.
- Double reversal: reversing an already scored item restores its raw direction. Keep variable names explicit, such as raw_item01 and scored_item01.
- Silent partial totals: the examples below require complete records. Do not treat a sum of available responses as though it were the sixteen-item total.
- Missing-value codes: TILDA’s official processing removes negative nonresponse and error codes before summation. Such codes must not enter an arithmetic total.
- Mixing versions: a shared acronym or overlapping items does not make the full, abbreviated and past-week scores directly comparable.
The complete-record rule used here is an explicit analysis choice, not a claim that a universal PSWQ prorating rule exists. Specify missing-data handling before analysis and report how many records it excludes. Also inspect the score distribution; the guide to floor and ceiling effects explains why a bounded measurement scale deserves attention at both endpoints.
How to report PSWQ results: APA-style example
Report the version, response coding, reversed positions, scoring rule and missing-data decision in the methods. In results, give the sample size and descriptive statistics for the analyzed records. Reliability estimates from another publication should be distinguished from estimates calculated in your own sample.
Illustrative methods wording: “Trait worry was measured using the 16-item Penn State Worry Questionnaire (Meyer et al., 1990). Responses ranged from 1 to 5. Items 1, 3, 8, 10 and 11 were reverse-scored before summation, producing possible totals of 16–80. Totals were calculated only for records with all sixteen valid responses.”
For a sourced description of prior evidence, write: “In Stöber’s (1998) nonclinical student sample (N = 148), PSWQ scores averaged 43.78 (SD = 10.09), with α = .89 and four-week test–retest r = .87.” These are that study’s results, not invented results for a new project. Replace the descriptive statistics with your own measured values when reporting your research.
How to calculate PSWQ scores in Excel, R and Python
These examples implement the full 16-item, 1–5 form only. They require complete input and preserve the raw responses. Test them with the synthetic record above, which must return 64, before applying them to an export.
Excel
Put items 1–16 in A2:P2, in order. Apply whole-number data validation from 1 to 5. The formula requires sixteen numeric cells, then uses the independently checked sum identity:
=IF(COUNT(A2:P2)=16,SUM(A2:P2)+30-2*SUM(A2,C2,H2,J2,K2),NA())
A blank or nonnumeric response produces #N/A. The formula relies on the stated input validation for response bounds; it does not independently enforce integer categories. Keep the validation step when transferring the formula to another workbook.
R
x = c(2,4,1,3,4,5,3,2,4,1,2,4,3,5,4,3)
stopifnot(length(x) == 16, !anyNA(x), all(x %in% 1:5))
scored = x
rev = c(1,3,8,10,11)
scored[rev] = 6 - scored[rev]
sum(scored) # 64
R uses positions starting at 1, matching the questionnaire numbers. The checks stop execution for a missing response, an invalid category or the wrong number of items.
Python
x = [2,4,1,3,4,5,3,2,4,1,2,4,3,5,4,3]
if len(x) != 16 or any(type(v) is not int or v not in range(1,6) for v in x):
raise ValueError("Expected 16 integer responses from 1 to 5")
reverse = {1,3,8,10,11}
scored = [6-v if i in reverse else v for i,v in enumerate(x,1)]
total = sum(scored) # 64
Starting enumerate at 1 avoids an off-by-one reversal error. For either language, place the calculation after a documented mapping from dataset columns to questionnaire positions; alphabetical sorting of item names is not a substitute for that mapping.
Questionnaire access and reuse
CamCOPS provides an implementation and a dedicated PSWQ documentation page. Its documentation states that the questionnaire is in the public domain and records explicit permission from Thomas Borkovec to Rudolf Cardinal dated 2015-07-23. That is a documented statement about the instrument’s use there, rather than an inferred licence from finding a downloadable questionnaire.
For a reproducible study, retain the exact form administered and its accompanying provenance or permission documentation. Record separately any translation, adaptation or platform terms. The references below distinguish the original development report, cohort specifications and software scoring implementation so that the measurement you describe can be matched to the measurement participants completed.
Sources
- Meyer et al. (1990), Development and validation of the Penn State Worry Questionnaire, authors’ abstract. Checked 2026-11-14.
- University of Cambridge, CamCOPS PSWQ documentation and scoring implementation. Checked 2026-11-14.
- Stöber (1998), Reliability and validity of two widely-used worry questionnaires, author manuscript, Table 1. Checked 2026-11-14.
- Hopko et al. (2003), Assessing worry in older adults, authors’ abstract. Checked 2026-11-14.
- The Irish Longitudinal Study on Ageing, PSWQ scale documentation and scoring code. Checked 2026-11-14.
- ADNI-D, Procedures Manual, February 15, 2017, printed page 151. Checked 2026-11-14.
- Stöber and Bittencourt (1998), Weekly assessment of worry, author manuscript, methods and Table 2. Checked 2026-11-14.
Data: PSWQ forms and coding specifications
Original compilation of the full-form CamCOPS implementation, ADNI-D response coding, TILDA short form and the published PSWQ-Past Week adaptation. Counts and response endpoints come from the four listed primary sources. Possible total endpoints are calculated as item count multiplied by the scored response endpoints; they are not diagnostic thresholds.
| Documented form or implementation | Item count (items) | Response minimum (points) | Response maximum (points) | Possible total minimum (points) | Possible total maximum (points) | Scoring specification and source |
|---|---|---|---|---|---|---|
| Full 16-item PSWQ (CamCOPS) | 16 | 1 | 5 | 16 | 80 | Reverse original items 1, 3, 8, 10, 11 as 6 minus response. Source: CamCOPS implementation. |
| ADNI-D 16-item coding | 16 | 0 | 4 | 0 | 64 | Reverse original items 1, 3, 8, 10, 11 as 4 minus response. Derived total range. Source: ADNI-D manual, printed p. 151. |
| TILDA eight-item short form | 8 | 1 | 5 | 8 | 40 | No negatively worded items; sum the eight responses. Source: TILDA PSWQ documentation. |
| PSWQ-Past Week | 15 | 0 | 6 | 0 | 90 | Original item 12 dropped; reverse negatively worded items as 6 minus response. Derived total range. Source: Stöber and Bittencourt (1998), methods and Table 2. |
Sources
- University of Cambridge CamCOPS PSWQ scoring implementation (accessed 2026-11-14)
- ADNI-D Procedures Manual 2017-02-15, printed page 151 (accessed 2026-11-14)
- TILDA PSWQ scale documentation (accessed 2026-11-14)
- Stöber and Bittencourt (1998), Weekly assessment of worry, methods and Table 2 (accessed 2026-11-14)
Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.







