2026 Ringing Census: why the questions are asked as they are
Companion to the implementation design · started 31 October 2026
Prepared by Mark Ainsworth with Claude (Anthropic). Mark, Elva and the committee made the choices; Claude gathered the reasoning from the decision log, the specifications, the meeting notes and Mark’s written discussions, and wrote it up for Mark to review.
Background and brief
Every question in the two census surveys should have a reason a reader can find. The census will be repeated, compared with and argued over, and each choice of wording costs something: comparison with 1988, a few seconds of a volunteer’s time, or a piece of data protection risk. A reader who was not in the meetings needs to see what each question is for, where it came from, and what it gives up.
This page does that, question by question, for the Tower Representative Survey and the Individual Ringer Survey. The implementation design says what the census does as a whole; this page says why each question looks the way it does.
The brief from Mark (6 November 2026):
- Record the reasoning for every question, not just the contested ones.
- Put here only reasoning that is on the record: the decision log, the specifications, meeting notes, and Mark’s written discussions.
- Where the reasoning is not on the record, say so plainly. Likely reasons, and gaps found while writing this page, go to a separate list of questions for Mark and the committee (
design/_questions-to-query.md, not published on this site). Each query has a number such as QM-12. - Much of the committee’s discussion is in recorded Zoom calls and in a long design session between Mark and Elva in late October. Neither transcript was available when this page was written. Entries that depend on them say so.
What the questions are checked against. The wording quoted here is the QuestionPro build as fetched on 5 November 2026 at 23:35 GMT (questionnaire/design/tower_test.md and individual_test.md). The API gives no routing, so routing is described from the specifications and the decision log, not from the build. QuestionPro reuses some question codes, so each entry gives the question’s id where a code is ambiguous. The entries for the questions added on 16 November (Ringer Q55, Tower A-R3, A-R4 and A-R5) are checked against the build fetched that day at 23:09 and 23:17 GMT.
Out of scope. The enumerator check-in form, and the pilot-only feedback questions, except where they bear on the census itself.
How to read an entry
Each question has a short entry with the same parts:
- Asks: the question as built, shortened where long.
- Why: what the answer is used for.
- Origin: who proposed it, and when it took its current form.
- 1988: the nearest 1988 question, and whether a comparison holds.
- Reasoning: how well the reason is recorded.
| Reasoning label | Meaning |
|---|---|
| Recorded | The reason is written down in the decision log, a specification, meeting notes or a written discussion, and the source is named |
| Partly recorded | Where the question came from is recorded, but not why it has its present wording or options |
| Not yet recorded | The question exists but its reason is not on the record. The entry gives a query number instead of a guess |
Status stripes follow the key in the implementation design: Open for a substantive point not settled, Open: minor for a smaller one. An entry with no stripe has no known open point.
Principles behind every question
These rules shaped almost every question. Each entry below restates the one that matters for it.
| Principle | What it means for the questions | Where it is recorded |
|---|---|---|
| The short ask gets answered | The tower return is split into a five-minute Core, which enumerators chase, and an optional Tail, never chased. In 1988 the short, agent-mediated band count reached 94% of sampled towers; the long tower captain form reached 60% | Tower Representative Survey specification v0.2–v0.4, section 2 |
| Keep 1988’s words where a 1988 figure rests on them | The band count, the shared-band categories and the “advance further” question kept 1988’s wording where possible. Attitudes need not match 1988; demographics should | Specification §4.1. Mark, August 2026: “I don’t think aligning to the 1988 survey is nearly so important on this attitudinal things than it is on the demographics” |
| Two ends, one test | Every tie between a ringer and a tower is reported by the tower rep and by the ringer, in the same terms, so the two can be joined | Specification §10, “Cross-instrument invariants”; D-060 to D-062 |
| Record what happened; apply definitions at analysis | Questions ask for counts and facts. Rules such as “who is an active ringer” are applied afterwards, so they can be argued over without re-running the census | Implementation design, “What we count”; analysis/shared-ringers.qmd |
| No names, no special category data | Ringers give no name or contact details. No question asks about health, religion or ethnicity. Under-18s are counted as numbers only | D-039; implementation design §8 |
| Assume low technical confidence | Short pages, plain words, few required answers, a tower list to pick from rather than names to type | Implementation design §7 |
| Full-circle ringing only | Every count means full-circle ringing (and handbells in the individual survey). Chiming, tolling, carillons and clocks are out | D-066 (agreed by Tina, 3 November) |
| Don’t let the census become a beast | A baseline snapshot, not every question for every purpose; deeper topics can be later, smaller surveys | Vicki, email of 21 October 2026, on the ringing-zone questions |
Part 1. Topics that span both surveys
Age bands: university ages get their own band
Questions: Individual Q1 (current age) and Q3 (age when first learned to ring); Tower rep A-B2 (ages of band members).
| Question | Bands used in 2026 | Bands used in 1988 |
|---|---|---|
| Individual Q1, current age | Under 18, 18–21, 22–29, 30–39, then by decade to 80+ | Under 15, 15–19, 20–39, 40–59, 60 or over |
| Individual Q3, age first learned | Under 12, 12–14, 15–17, 18–21, 22–29, 30–39, then by decade to 80+ | Under 15, 15–19, 20–29, 30–39, 40–49, 50 or over |
| Tower rep A-B2, band members | Under 18, 18–21, 22–39, 40–59, 60+ | Under 15, 15–19, 20–39, 40–59, 60 and over, each by sex |
The choice. The bands split at 18 and at 22, so that the usual university years, 18 to 21, form a band of their own.
Why 18. Under-18s are counted and surveyed under tighter rules: no contact, no free text kept, a number only in the tower return. The age question has to separate them cleanly.
Why 18 to 21 on its own (Mark, 31 October 2026). Universities are now a key source of recruitment and activation, at least for the Red zone of method ringing. And far more people go to university than at either earlier survey. The age participation index, the share of young people entering higher education, was 14.3 per cent in Great Britain in 1972 and 15.1 per cent in 1988–89: almost unchanged between the two surveys. For the cohort measured in 2020/21, 47 per cent of young people in England had entered higher education by age 25. The 2026 figure is measured differently, but the change since both earlier surveys is roughly threefold, and it all happened after 1988. So the years people tend to spend at university are worth separating from the years after it, in current age, in age when learned, and in the ages of a tower’s band.
Why the tower rep has fewer bands. The captain is estimating the ages of the whole band from memory. Five bands is about as many as that estimate can carry, and small cells at small towers raise the risk of identifying under-18s. Every tower-rep boundary (18, 22, 40, 60) is also an individual-survey boundary, so individual returns roll up into the tower bands exactly (specification v0.3, section 5).
Why the learned-age bands are finer at the young end. The committee wanted detail at the ages when people start: “when we asked people at what age they learned we felt useful to go more granular into younger ages” (Mark, reporting the questionnaire review meeting, summer 2026). The August specification removed an overlap at 15 (the first draft had 12–15 and 15–17), so that 1988’s “under 15” can be rebuilt by adding the first two bands. The build of 5 November now matches the specification.
What it costs. 1988 split at 15 and 20, not 18 and 22. So:
- current age compares with 1988 only as under 40, 40–59 and 60 and over;
- age first learned compares as under 15 and from 30 upwards;
- A-B2 compares at 40–59 and 60 and over, and has no split by sex.
Considered and not adopted: splitting 18–21 into 18–19 and 20–21. This would restore 1988’s boundary at 20 for one more answer option. Mark decided against it in August 2026: “I prefer just keeping 18-21. Perfectly aligning to the split isn’t critical IMO.” The tower rep specification recorded the same at version 0.3 (“age bands revert to 18–21 (no 18–19/20–21 split)”). The 1988 bridge at age 20 is an accepted, logged limitation.
Reasoning: recorded.
Sources.
- House of Commons, written answer on higher education, Hansard, 10 March 1986, vol. 93, cols 332–3W (age participation index from 1972): https://api.parliament.uk/historic-hansard/written-answers/1986/mar/10/higher-education
- House of Commons, Higher Education (Age Participation Rate), Hansard, 2 May 1989, vol. 152, cols 3–5: https://api.parliament.uk/historic-hansard/commons/1989/may/02/higher-education-age-participation-rate
- FE News, Higher education participation statistics (reporting the Department for Education’s cohort measure, 2010/11 cohort, 2020/21): https://www.fenews.co.uk/resources/higher-education-participation-statistics/
- 1988 questionnaires: appendix C of the 1988 survey.
- Tower Representative Survey specification, versions 0.3 (end of August
- and 0.4 (20 September 2026), section 5.
Where people ring: every tie between a ringer and a tower is reported from both ends
Questions: Tower rep A4 (who usually rings here), A-B1 (band size), A-B2 (ages), A-B4 (other towers), A-C1 to A-C5 (census-week, Sunday and twelve-month counts), and A-T5 in the Tail (reliance on visitors). Individual Q46 (number of bands), Q49 and Q50 (bands that rotate), Q37 (main tower) and its follow-up, Q27 and Q53 (other towers), and Q39 to Q43 (where you rang in census week and on Census Sunday).
The choice. The census does not assume one tower, one band. Each tower rep says what kind of band rings there and names the towers it shares ringers with. Each ringer names every tower whose captain would count them as one of the band, says whether a band of theirs rotates round several towers, and names where they actually rang in census week. The analysis joins the two ends. Overlaps are removed there, not in anyone’s head while filling in a form. This is the design decided on 31 October (D-060 to D-062), as built in QuestionPro by 5 November. The boxes at the end list where the build falls short of it.
Five things the band questions must satisfy at once
| Constraint | What it demands of the questions |
|---|---|
| 1988 comparison | Keep 1988’s words where a 1988 figure rests on them: the roster instruction (ringers per bell), the incumbents’ categories for who rang (share of towers with a shared band), and the ringer’s “member of more than one band” and “band provides the ringers for more than one church” |
| The sampling frame | The frame is a list of rings (every ring of 4 or more bells in Dove’s Guide, keyed by RingID; D-007, D-017). Ringers are reached through rings: chased towers, Golden Ticket bands. So every ringer’s link to frame rings must be recorded, by name, to work out how likely they were to be reached |
| The tower rep survey | A five-minute Core that one person can answer from the tower’s side. The rep knows the band and where it rings; the rep cannot know every other tie each ringer has |
| The individual survey | An open link with no names. The ringer knows their own ties to towers; the ringer cannot see the band as a whole |
| How ringing actually works | Benefice bands rotating round several churches, ringers in two bands, towers rung only by neighbours, regular helpers, practice elsewhere, society bands, and “band” read as a list rather than as who rings. None of these fits one tower, one band |
The constraints pull against each other. The frame needs exact links; the Core needs brevity; 1988 needs its old words; and the old words now carry meanings they did not have in 1988.
Two cases that shaped the design. A Council representative described (September 2026) a benefice of six towers rung by one band of 13 on a bi-monthly rota: in census week the tower survey might find five ringers at one practice and no service ringing, which would badly understate a healthy band. And Tina pointed out (summer 2026) that one band serves all of Dundee with no preferred tower; giving everyone a single “principal tower” would make the other towers look unrung. Both are why the census asks about rotation directly and never assumes a home tower.
Each kind of tie, and how it is seen from each end
| Relationship | What the tower rep reports | What the ringer reports | 1988’s measure |
|---|---|---|---|
| One tower, one band (the simple case) | A4 “Tower band, which rings only here”; A-B1 counts the band | Q37 names the tower; follow-up: the captain would count them | Incumbents 4: “a regular band associated with your church only” |
| A band that rotates round several towers, often a benefice band | A4 “Band that regularly rings here and at other towers too”; A-B1 counts the whole band; A-B4 names the other towers | Q49 Yes; Q50 gives how many towers the rotation covers; Q37 and Q53 may name some or all of them | Incumbents 4: “a regular band that comes to your church and rings at others too” (Table 11.1); Ringers 1: “does your band provide the ringers for more than one church?” (12%) |
| A ringer in two bands | Each band’s rep counts them in A-B1 | Q46 = 2 or more; Q37 plus Q27 and Q53 name both towers | Ringers 1: “a member of more than one band of ringers” (8%) |
| A tower rung by other towers’ bands, with no band of its own | A4 “Ringers from other towers’ bands; no tower band”; A-B1 and A-B2 skipped; A-B4 names the towers the ringers come from | The ringers name it in Q41 or Q43 (where they rang), not as one of their own towers | Incumbents 4: “visitors who don’t normally ring at your church” |
| A regular helper, ringing most Sundays at a nearby tower without being one of its band | Not in A4 (the tower has its own band). Counted at that tower in A-C1, A-C3 and A-C5; A-T5 (Tail) shows the tower depends on visitors | Q41 and Q43 show it for census week only. Nothing records it as a habit | Ringers 1: “Are there any other towers where you ring regularly on Sundays?” (19%; 29% of tower captains) |
| Practice at another tower | Counted at that tower in A-C1 that week | Q41 names it, for census week only | Ringers 3: regular practice at other towers (26%) |
| One tower, more than one band (a university or society band ringing regularly at a tower that has its own band; added 17 November, D-081) | A4 “Tower band, and another band also rings here regularly”; A-B1 counts everyone in all the bands; if the bands send separate returns, both are kept | Students name the host tower in Q37 or Q53 if its captain, or their own society’s, would count them | None: 1988’s incumbents would have ticked “a regular band associated with your church only” |
| A society band at a tower it does not belong to | A4 “Society or Guild band”; the society’s steward answers | Members name the tower if its captain would count them | None |
| On a list but not ringing, or ringing but not on a list | A-B1: count by who rings with the band, “not by whether they are affiliated to this tower in your association’s records or on any church list” (D-060) | Q37 follow-up and Q27 ask whether the captain would count you, not whether you are a “member” (D-062) | Did not arise: 1988’s agents counted the captain’s own list of who rang |
The 1988 percentages are of ringers, or of tower captains where stated. They come from chapter 11 of the 1988 report; the questions are in its appendix C.
How the two ends are joined
One test for “belongs to a tower”, stated from both sides. A tower’s band is everyone who rings there on Sundays and/or at practice and whom the captain counts as one of the band. A-B1 asks the captain to count them. Q37’s follow-up and Q27 ask the ringer whether the captain would count them. The two must stay the same test. If they drift, the next step goes wrong.
Weights for ringers who belong to several towers. A ringer who belongs to three frame towers has about three chances of being in a chased band. The standard fix, multiplicity estimation (Sirken, 1970), divides each ringer’s weight by the number of frame rings through which they could have been reached. For most ringers that number is the towers they name as their own (Q37 if the captain would count them, plus Q53). For a member of a rotating band it is the towers in the rotation (Q50). Only frame rings count, because only frame rings can be drawn; the picker also offers rings of three, mini-rings and towers abroad (D-018), and those stay out of the divisor.
Band counts are band lists, not a headcount. Because A-B1 counts a shared band in full at each of its towers, adding A-B1 over towers counts shared ringers more than once, exactly as 1988’s lists did (41,100 before 1988’s own correction, about 36,000 after). That is right for chasing: an enumerator chasing a tower with a shared band chases the whole band. The headcount comes from linking returns. Towers joined by any rep’s A-B4 or any ringer’s Q53 form clusters, and the rule for counting each ringer once within a cluster is still open (Q-024).
Census-week counts are not de-duplicated at the tower. A ringer who rang at three towers that week really did ring at each, so A-C1 to A-C5 stay as they are. The ringer’s own Q41 and Q43 say where.
Rotation is asked about, not inferred. The census-week questions are a one-week snapshot. A band that rings at each of its churches once a month or less may not appear at most of them in census week. Q49, Q50, A4 and A-B4 record the rotation directly, and Q40’s option “My tower doesn’t ring every week, and it wasn’t our turn” explains a silent week.
What still compares with 1988
| 1988 figure | 2026 source | Comparable? |
|---|---|---|
| Ringers per bell, from band lists | A-B1 added over towers | Yes, as worded. 1988’s instruction is kept word for word; D-060’s two sentences follow it and only rule out readings 1988 never faced |
| Share of towers with a shared band (Table 11.1, nearly one tower in ten) | A4, options 2 and 3 | Roughly. Same categories, but 1988 asked who rang on the survey Sunday and 2026 asks who usually rings |
| Ringers in more than one band (8%); whose band serves more than one church (12%) | Q46, Q49 | Roughly. See the note below on who answered |
| Ringers who ring regularly elsewhere on Sundays (19%) | Not asked in the current build | No. Q43 gives one Sunday, not a habit |
| Ringers who practise regularly at other towers (26%) | Q41, census week only | No: one week is not “regularly” |
Who answered differs. 1988 drew its ringers from band lists, every third name, and its form opened “You are a member of the band at <name of church>”. Every 1988 respondent was on a band list by construction. The 2026 individual survey is an open link, and some who reply will be no tower’s band member: lapsed ringers, regular helpers, learners. Comparisons with 1988’s ringer figures should be restricted to respondents whom some captain would count (Yes to the Q37 follow-up, or a tower named in Q53), and weighted.
What was considered and not adopted
| Option | Why not (Mark, 31 October) |
|---|---|
| Replace A-B1 with an attendance test: “How many people would you expect to see ringing here in a typical month?” | Gives up 1988’s roster series, and leaves out ringers on a rota that reaches this tower every second month |
| Ask reps how many ringers are affiliated to the tower in the association’s records | Registers are better collected from associations after the census, as already planned; reps may not know |
| Name safeguarding in the band count | Makes the census read as a compliance check, which could lower response where records are weakest |
| Ask ringers to name a first tower for Sunday service | Q37 deliberately does not imply a home tower, and practice-only ringers have no Sunday tower |
| Replace Q46, Q49 and Q50 with one “why do you ring at more than one tower?” question (suggested 30 October) | Not taken up: Q46 and Q49 already match 1988’s two follow-ups, and Q50 gives the rotation size the weights need |
| One six-option question on band membership (“single band at a single tower / more than one band / a band at several towers / both / handbell band…”), the August design | Testers found it hard to answer: it mixed three things (how many bands, whether one rotates, and handbells). Split in early October into Q48, Q46 and Q49, which recover the same categories |
Where the current build may not be optimal
Checked against the designs fetched from QuestionPro on 5 November 2026 (questionnaire/design/*_test.md). Question ids are given because some codes are used twice.
Open The Q37 follow-up’s “No” answer still says “member” (id 1616664179).
The stem now reads “Would the tower captain there count you as one of their band?”, as D-062 intended. But the second answer reads “No, I ring there but I’m not a member of the band”. That is the answer a regular ringer who rejects the word “member” will pick, which is the misreading D-062 set out to stop. If they pick it, their main tower drops out of their multiplicity count and their weight comes out too high. Suggested wording: “No, I ring there, but they wouldn’t count me as one of the band”. The stem says “band” where D-062 said “ringers”; “band” is closer to A-B1’s own words, so it is probably the better choice, but the decision entry should record it.
Open: minor Q27’s definition differs from Q37’s (id 1616375258).
Q27’s help text opens “You’re a member of a band if…”, reintroducing the word, and leaves out Q37’s sentence on learners (“If you’re learning, count yourself in only if you can ring unaided”). A learner in two bands could then name a second tower under a looser test than their first. The fix is to copy Q37’s definition into Q27 word for word.
Open: minor The no-lists sentence is in the tower rep survey but not the individual survey.
A-B1 carries both of D-060’s sentences. The ringer-side definitions (Q37 follow-up and Q27) leave out “It doesn’t matter whether you’re affiliated to the tower in your association’s records or on any church list”, which D-062 included. Mark left it out on 3 November because it made the help text too long.
The case against leaving it out. The two ends should apply the same test. A ringer who is on an association register for a tower they rarely ring at may say the captain would count them, while a rep following A-B1 would not.
Why the omission is probably safe. The ringer is asked what the captain would do, and A-B1 now tells the captain to count by ringing, so the ringer’s answer inherits the rule. Q37 itself asks for a tower “where you regularly ring”. And the risk D-062 addressed on the ringer’s side was the word “member”, not lists. So the omission is defensible, but D-062 should gain a note, or a new entry, saying the sentence was dropped and why. If a shorter guard is wanted, “Being on an association or church list doesn’t matter.” is eleven words.
Open 1988’s Sunday question has not been added: “Are there any other towers where you ring regularly on Sundays?”
D-062 added it, asked of everyone, after Q53. The 5 November build does not have it.
The case for adding it.
- It is the only question that sees the regular helper as a habit. A helper is not one of the other tower’s band, so is not named in Q53, and A4 does not show them because that tower has its own band. Q43 sees them only if they helped on 22 November. In 1988, 19% of ringers and 29% of tower captains rang regularly elsewhere on Sundays; in Bedfordshire in 2021, 23% of Sunday service ringers helped at other towers. It is the commonest tie after the band itself.
- It is 1988’s own question, the source of one of its headline findings, so it gives a direct comparison (subject to the different respondents, above).
- It costs one yes/no. No picker is needed.
The case against.
- It does not change the weights or the count. Helpers are not on the helped tower’s band list, so they add nothing to the multiplicity divisor. The question serves description and comparison only.
- Q43 gives a proxy. The share of Census Sunday ringers who rang at a tower they do not belong to is a one-Sunday estimate of the same thing.
- “Regularly” is vague, and every item lengthens a survey aimed at people with low technical confidence.
Recommendation. Add it, worded as 1988 worded it. Helping out is how much of ringing is kept going, especially in cities, where it is informal and invisible in both Church and Dove records. A single yes/no is a small price for the one figure on it that compares with 1988.
Open: minor Q46 says “member”, and comes before Q37 (id 1616825994).
Q46, “How many ringing bands are you a member of?”, keeps “member” on purpose: it matches 1988’s “a member of more than one band of ringers” (D-062). But it sits in the block before Q37, so the first framing a ringer meets is the one the Q37 follow-up was reworded to avoid. A regular ringer who rejects “member” may answer Q46 “None” and then say Yes at Q37.
The 1988 comparison is worth the risk, and the inconsistency can be measured rather than designed out: Q46 should roughly equal the number of towers named as one’s own, after allowing for rotating bands (Q49, Q50). Where it does not, that is evidence on how often “member” is rejected. No change proposed; this records the trade-off.
Open: minor A4 as built differs from D-061 (id 1616430004).
The options are shortened (“Tower band, which rings only here” for “Our own band, which rings only here”), “Visiting bands by arrangement only” now comes before “Society or Guild band”, and option 2 has lost “(for example, a benefice band)”. The shortening and the order are harmless. The lost example matters a little: it was the cue that tells a rep a benefice rotation counts as “regularly rings here and at other towers”. Individual Q49 keeps that example, so the two ends now describe rotation differently. Restoring the four words is suggested. D-061 should record the order as built.
Open: minor Routing and text to check in the live build.
QuestionPro’s API does not return skip logic, so these cannot be confirmed from the fetched design:
- A-B4 should show after A4 options 2 and 3 only.
- A-B1 and A-B2 should be skipped after “Ringers from other towers’ bands; no tower band”, “Visiting bands by arrangement only” and “No band”. Without the skip, a borrowing tower’s rep may count the lending band, which would then be counted at both towers with nothing to link them.
- A-B1’s text begins with a stray quotation mark.
Open: minor A-C2’s visitor test misreads rotating bands.
A-C2 defines a visitor as “someone you would not expect to see here in a typical month”. A member of a band that reaches this tower every second month meets that test at their own band’s tower. The census-week total (A-C1) is unaffected; only the visitor share at rotation towers is. No wording change proposed: at analysis, towers with A4 option 2 should not be read as relying on visitors from A-C2 alone.
Open: minor Rules the analysis must write before it uses these answers (Q-024, D-010).
- Only frame rings in the multiplicity divisor. A ringer may name rings outside the frame. Those are recorded but not counted.
- Q50 against named towers. A rotating band’s member may name only their main tower in Q37 and Q53, and give the rotation’s size in Q50. The rule must say which wins, and how Q50 is used when the rotation’s other towers are not named anywhere.
- Towers with two rings (13 in the picker, D-012). A band ringing both rings at one tower could be counted under each ring. The ring-or-tower question is deferred (D-010).
Reasoning: recorded (D-058, D-060 to D-062; analysis team meeting of 30 October; Individual survey discussion of 6 October).
Sources.
- Central Council of Church Bell Ringers (1988), Survey of Ringing 1988: Questionnaire for Individual Ringers, questions 1 and 3; Questionnaire to Tower Captains, questions 7(e), 9 and 11; Questionnaire for Incumbents, questions 4 and 15; the agents’ instruction for the band list. Transcribed in appendix C of the 1988 survey; the 1988 figures quoted are from its chapters 4 and 11.
- Decisions D-007, D-010, D-012, D-017, D-018, D-049, D-058, D-060, D-061, D-062 and open question Q-024, in
provenance/decisions.mdandprovenance/decision-register.md. - The analysis chapter
analysis/shared-ringers.qmd(“Shared ringers: one ringer, several towers”): the 1988 evidence, benefices, the Bedfordshire 2021 and Devon 2023 surveys, and the three meanings of “band member”. - QuestionPro designs fetched 5 November 2026:
questionnaire/design/tower_test.md(survey 1603084948) andquestionnaire/design/individual_test.md(survey 1603035043). - Sirken, M. G. (1970). Household surveys with multiplicity. Journal of the American Statistical Association, 65(329), 257. https://doi.org/10.2307/2283590
Full-circle ringing only: what every count leaves out
Questions: Tower rep A-C i1 (the census-week introduction), A-C1, A-C3, A-C4, A-C5 and A3; Individual Q39 and Q48.
The choice. Every count in both surveys means full-circle ringing, and handbells in the individual survey. Clock strikes, automatic chimes, tune machines, carillons and hand chiming, including Ellacombe chimes, are not counted (D-066, agreed by Tina on 3 November). Tolling a single bell that is down is not full-circle ringing either, so it is out too; Mark is to confirm that reading (D-066).
Why. A Council representative asked whether towers where Ellacombe chimes are rung regularly should count, in the spirit of “every tower, every ringer”. Tina’s answer: the Council is in the business of promoting change ringing, not any ringing; the census can say chiming was considered. No earlier survey (1988, the 1999 working group, 2007–8, the 2015 drafts) mentions chiming, so this is a first definition, not a break in a series.
How it is built. The build of 5 November puts one sentence at the top of the census-week section (A-C i1): “Count full-circle ringing only. Do not count chiming (e.g. on an Ellacombe apparatus), tolling, a carillon or a clock.” The individual survey’s Q39 says “Chiming (for example on an Ellacombe apparatus), tolling and carillons don’t count”, and Q48 says “Tower bells (full circle)”. The build note of 3 November had proposed the sentence at each count instead; one introduction is shorter and draws the line in one place.
Alternatives considered (put to the officers by Mark, 3 November): recording hand chiming in one optional Tail question, which would give a floor rather than an estimate because the Tail is answered least at small towers; or counting chiming as its own activity in the Core, which would lengthen the five-minute Core with the build nearly frozen.
Open: minor A-C4 does not repeat “full circle” in its own words.
A-C4 (“Were the bells rung at all on SUNDAY 22nd November 2026?”) is asked of exactly the towers most likely to have their bells chimed or tolled instead: no band, unringable bells, or an enumerator’s observation. The build note of 3 November proposed “rung full circle” in the stem for that reason. The section introduction covers it if the respondent sees it, so this is a check rather than a fault: confirm that the routing to A-C4 does not skip the introduction.
Reasoning: recorded (D-066; questionnaire/build-changes-2026-11-03.md).
Teaching payoff: added after the freeze to test whether moving away changed who towers teach
Questions: Ringer Q55 (id 1617285418), “Do you still ring regularly at the tower where you learned?”; Tower A-R5 (id 1617285419), the movement grid; Tower A-R3 and A-R4 (ids 1617285608 and 1617285609), the best and least promising age groups for recruits. Added after the freeze of 13 November, and built by Vicki on 16 November (D-079). Tower Q31 and Q32 were removed at the same time (D-080).
Drafted by Claude (Opus 5.5) from Mark’s discussion of 14 to 16 November 2026, for Mark to check. Mark made each choice recorded here.
The hypothesis (Elva, 14 November). Up to the 1970s, teaching a young learner paid off for the tower that taught them. Young ringers had fewer competing hobbies, and most stayed in the area. Then a growing share of young people, often the brightest and most ambitious, went to university and settled somewhere else. A tower that teaches a twelve-year-old now often loses them at eighteen. A tower captain who needs enough ringers for the bells they have does better to teach older adults: slower to learn, less likely ever to ring advanced methods, but likely to stay. The slow decline of young and advanced ringers (the Red zone) follows from that change in the captain’s sums.
Mark’s corollary. If that is right, individual towers cannot be expected to become centres for teaching young people, because the tower pays and someone else benefits. Only a larger body, such as an association or the Central Council, gets both the effort and the reward. The recent success of the young ringers event is relevant here.
The strongest case against, which is why the test is worth having
- The timing is awkward. In Mark’s trends workbook (
Calcs for trends visuals - 2025.xlsx, sheet Population Mix), under-20s were 41% of ringers in 1972, 20% in 1988 and 10% in 2007. Higher education entry barely moved between the first two dates: the age participation index was 14.3% in 1972 and 15.1% in 1988–89 (see Age bands for the sources). So the first halving of the young share cannot have come from moving away to university. The 1972 and 2007 surveys had weak sampling, and a growing 60+ group shrinks the other shares on its own, so the figures are rough. - A rival explanation in the same workbook. Its Church of England series for young people’s usual Sunday attendance (sheet Baptisms) falls from about 385 in 1970 to 226 in 1990 and 119 in 2010. The church route that brought children to the rope was drying up. That predicts the same result, fewer young learners, from the supply side.
- Supply, not choice. Retired people have daytime hours and come asking. Captains may be teaching whoever turns up, not choosing older learners.
- Cost, not payoff. Safeguarding has made teaching children more work since the 2000s. That also predicts fewer young learners, but points to a different remedy.
Mark’s working view: moving away is probably a real and growing cause after about 1990, not the only one. The questions below are chosen so the census can tell these explanations apart.
What each question tests
| Question | What it measures | How it tests the hypothesis | 1988 |
|---|---|---|---|
| Ringer Q55 | Whether the tower that taught you still has you: yes / near enough that I could / too far away now / no, for another reason | With age when learned (Q3), years ringing (Q4), “University” as the route in (Q35) and the place type of the main tower: the share of young starters now ringing far from where they learned, by decade learned and kind of place | Captains were asked “did you learn to ring at the tower you now ring at?” (yes/no) about themselves |
| Tower A-R5 | Ringers in the last 12 months who joined by learning to ring unaided here; joined after moving into the area; left because they moved away; stopped ringing altogether | Which kinds of place gain and lose ringers by moving. Nationally, moved-in against moved-away gives a rough rate of movers lost to ringing. Stopping, set against the band’s ages (A-B2), shows the cost of a band that is mostly older | “Moved away from the area” and “given up ringing” are two of the three rows of captains’ question 31 (Table 9.5) |
| Tower A-R3, A-R4 | Which age group the respondent thinks makes the best, and the least promising, recruits | What captains prefer. Set against who they actually have (A-B2, and the ages of recent learners from Q3 and Q4), it separates the captain’s choice from who turns up | Captains’ question 26. In 1988, 13% of those naming a best group chose “40 or over” and 39% named it least promising; under-15s split 31% best, 32% least (Chapter 7) |
The question that decides the remedy. Do young ringers who move away keep ringing somewhere else? If they do, teaching them still pays off for ringing as a whole, just not for the tower that did the teaching, and Mark’s case for associations and the Council carrying the cost holds. If they mostly stop when they move, the better remedy is catching people at the move: university societies and a welcome in the place they move to. Q55 sees only the movers who still ring; A-R5 counts all who leave and all who arrive. Together they give a first answer.
Choices made in the wording
- Q55 asks about the tower where you learned, not your main tower. The first draft compared the learning tower with the main tower chosen in Q37. Mark pointed out that this fails for people in more than one band: someone who learned at their second band’s tower would have had to say it was “another tower”. The question now asks only whether the learning tower still has them.
- No fixed distance. The draft had “within about 15 miles”. A fixed distance means something different in rural Devon and in London, and the hypothesis is about whether the tower still has the ringer, not about miles. “Near enough that I could” is the respondent’s own judgement.
- The fourth Q55 option (“No, for another reason, e.g. there’s no ringing there now”) keeps towers that have fallen silent out of the two moving answers.
- Q55 sits after Q35, so “when, how and where you learned” come together, before the tower pickers, which would otherwise pull people towards the towers they had just chosen.
- A-R5 asks for four counts by tapping, not typing, with 0 to “5 or more” and “Not known”. Counts are small: in 1988 the net gain was about 0.7 ringers a tower a year (Chapter 9). Mark cut a six-number version as too much to ask of tower reps.
- “Stopped ringing altogether”, not 1988’s “given up ringing”: the same meaning, without suggesting failure to an 85-year-old who has retired from the rope. No reason is asked, because reasons drift into health, which the census does not collect (D-039). Stopping because of age can be inferred across towers from the band’s ages.
- “Learning to ring unaided” is the line used by the band count (A-B1) and the learner count (A-C6). So the first row of A-R5 counts exactly the people who moved from A-C6 into A-B1 in the year. It is not 1988’s question 19, which counted learners who started.
- A-R3 and A-R4 are single choice, as in 1988 (“one B and one L please”). Forcing a choice shows which way the balance tips; with multi-select many would tick most groups, and the 1988 comparison would break.
- Age groups: under 15, 15–17, 18–21, 22–39, 40–59, 60 or over. These keep 1988’s “under 15” exactly and rebuild its “40 or over” by adding two groups, so both ends of the 1988 comparison survive. They also match the band ages in A-B2 (under 18 = under 15 plus 15–17), and give 18–21, the university years, their own group. 1988’s 15–19 and 20–39 cannot be rebuilt.
Limits to state when the results are reported
- Q55 sees survivors only. Movers who stopped ringing are counted only by A-R5.
- Age, period and cohort are tangled. In a single census, the decade someone learned in cannot be fully separated from how long they have had to move. Most moving happens between about 18 and 35, so comparing ringers aged 40 and over who started young, across the decades they learned in, is the fairest test.
- A-R5 counts are only usable added up, nationally or by place type, never tower by tower.
- Returns from an enumerator’s own knowledge (Q47, “my own observation”) will give weak counts. Analyse with and without them.
- A band that rings at several towers may report the same leaver from each. Handle it as for the band counts, through A-B4.
- The books do not balance. Moves between towers in the same area, and returners after a break, are deliberately not asked.
- Splitting age groups can shift answers. Offering 40–59 and 60+ separately may draw slightly more choices than 1988’s single “40 or over”. This is documented for probability judgements as partition dependence (Fox and Clemen, 2005, Management Science); how far it carries over to this question is not known. Treat it as a caveat on the 1988 comparison.
- A-R3 compares with 1988 only for tower captains. Filter on the role question (Q46, id 1616891731).
Considered and not adopted
| Considered | What happened | Reasoning |
|---|---|---|
| A fixed distance in Q55 (“within about 15 miles”) | Replaced by “near enough that I could” (Mark, 14 November) | A fixed distance favours one kind of place over another |
| A tower picker for the tower where you learned | Not proposed for the build | The richest answer (exact flows between places), but a new picker question days before launch |
| An age split in the learned-here row of A-R5 (under 18 / 18 or over) | Dropped (Mark, 16 November) | Finicky and unnatural for tower reps. The ages of recent learners come instead from the ringer questionnaire (Q3 with Q4), though under-18s will answer less often |
| 1988’s full rows: moved to other towers in the area (out and in), and returners | Dropped (Mark, 16 November) | Local moves cancel out nationally; the “moved into the area” wording already leaves out returners |
| A grid of learners over ten years, by age when they started and whether they stayed | Not proposed for the build | The most direct measure of the payoff, but relies on long memories |
| Multi-select for A-R3 and A-R4 | Not adopted (Mark, 16 November) | Breaks the 1988 comparison and hides the tilt the question is for |
| 1988’s four age groups exactly | Not adopted (Mark, 16 November) | Keeps the middle of the 1988 comparison, but loses the university years and the match with A-B2 |
Reasoning: recorded (Mark’s discussion with Claude, 14 to 16 November 2026; D-079, D-080). The 1988 figures are from Chapters 7, 9 and 17 and appendix C of the 1988 survey.
Part 2. The Tower Representative Survey, question by question
What the survey is for. One return per tower, from whoever knows its week-to-week picture. It bridges Dove’s list of towers to the ringers. The specification sets five jobs: a band count on 1988’s definition; a census-week count, the modern measure; tower facts for the model of who replies; checks against the individual survey; and a record of how each return was obtained. Its stated priorities, in order: (a) answers the analysis can use, (b) the same reading by about 6,000 different respondents, (c) completion time (specification v0.4, section 1).
Core and Tail. The Core takes about five minutes (less at a tower with little ringing) and is what enumerators chase to near 100%. The Tail takes about five minutes more, is optional, and is never chased. The reason is 1988’s response rates: the tower captain’s long postal form came back from 60% of sampled towers, but the short band count, asked separately through local agents, from 94% (specification v0.2, section 2). The Core saves as a complete return on its own.
Opening
Q16 · Welcome text
Asks: a welcome naming the Central Council, the purpose (“how best to support recruitment and retention”), and the time: “less than 5 minutes to complete the core part of this survey, and under 10 minutes in total”.
Why: the welcome sets expectations, and the time claim is a promise that shapes who starts. Origin: rewritten after the pilot (build list of 29–30 October): the earlier text said respondents were “one of a small number of people who have been selected” and “20–30 minutes”, both wrong for a survey open to every tower. 1988: the covering page also said the purpose was to see “what needs to be done to ensure that as many bells as possible are rung every Sunday”.
Reasoning: recorded (build list, 30 October). The Core time is now stated as “less than 5 minutes”, not the “about 2 minutes” in the specification: QM-01. From 8 November the design, the emails and these notes all call it about 5 minutes; 2 minutes holds only where there is little ringing to report.
Q45 · Before you start: the privacy notice
Asks: a short notice: the Central Council is responsible for the data; the use is to estimate how many people ring and plan support; the basis is the Council’s legitimate interest; “We never ask you to name any ringer”; storage in the UK or EU; results only as totals; the right to object.
Why: under the legitimate interests basis the privacy notice must be given, not agreed to. A required tick box would look like consent and would cost returns from anyone who hesitates (implementation design, “Why: the 1988 evidence, and why no required tick”). Consent was rejected as a basis because people withdrawing consent unevenly would bias the counts. Origin: data protection review accepted by the committee on 16 October; the earlier specification had a required tick.
Reasoning: recorded. The contact address and the link to the full notice are still placeholders (“[address]”): QM-02.
Core: about your tower
A1 · Which tower is this return for?
Asks: the tower, chosen from a search box over the locked list built from Dove’s Guide. Never typed.
Why: tower names are vague and duplicated, so free text would make returns impossible to match. The answer stores Dove’s ring identifier, RingID, which is the key every return is matched on (D-007). The list holds every ringable ring in every country, including rings of three and mini-rings, so anyone can find their tower (D-018, D-040), with a “not listed” escape after three searches (D-056). Origin: agreed at the first committee meeting (6 January 2026: “tower names selected from a pre-defined list to ensure accuracy”). No county or district step, because some people don’t know their association structure (Tina, summer 2026). 1988: the form named the tower for the captain.
Reasoning: recorded.
Q46 (id 1617089050) · Who is filling in this form?
Asks: “I’m answering as the representative for this tower” or “I’m a census enumerator entering answers for this tower”.
Why: how a return was obtained is itself an input to the model, because an enumerator’s zero and a captain’s zero are different observations. The answer routes enumerators past the privacy notice and contact details, and flags the return (implementation design, §3). Origin: Mark, 29–30 October. The specification had planned four different invitation links, one per way of completing the return. Mark did not trust volunteers not to confuse the links, and some returns will be typed in while the rep is on the phone, which a link cannot capture. So one link, and a required first question.
Reasoning: recorded (discussion of 29–30 October; D-049 uses the answer in the status table).
Q47 (id 1617089051) · Where are these answers coming from?
Asks (enumerators only): from the rep or a band member, with me or on the phone now; told me earlier; or my own observation, local knowledge or records.
Why: these give the four return categories of the design: rep unprompted, rep after chasing, enumerator with the rep, enumerator from observation where no band is left (D-049). “Told me earlier” is second-hand and is flagged at analysis. The question is asked first, so the enumerator states the source before giving any numbers. Origin: the same discussion as Q46. It replaced the specification’s A-F1 (“How did you establish the position at this tower?”), which came at the end and only for observed returns.
Reasoning: recorded. Whether “unprompted” and “after chasing” are told apart from the date and the chasing log, as proposed, and not asked: QM-03.
Q46 (id 1616891731) · What is your role with respect to the tower?
Asks: tick any of tower captain / ringing master, another band officer, a band member, representative of the church, other.
Why: who answers shapes how good the counts are likely to be. The 1988 return came from the captain; 2026 accepts “whoever knows the week-to-week picture”. Origin: not recorded; it predates the specification. The proposal of 30 October was to show it only to reps, make it single choice, add “Tower captain” and decide whether “Representative of the church” stays.
Partly recorded. The build keeps it as tick-all, with the church representative option: QM-04.
A2 · What is the state of ringing here?
Asks: one of: more ringers than we can cope with; the band is thriving, at full capacity; enough ringers but would welcome more; not quite enough ringers; hardly any ringers; no band at all; bells unringable; not known.
Why: a quick, near-universal measure of tower health, the census’s first secondary objective. It also sorts towers for routing: “No band at all” and “Bells unringable” skip the band questions and go to the Sunday question. Origin: Elva’s scale from her first draft, with “More ringers than we can cope with” added at the questionnaire review meeting (summer 2026). That option was kept deliberately: the oversubscribed tower is rare but useful, and a growing / stable / shrinking scale would hide it (specification v0.4). “Unknown” became “Not known” on 30 October, to match the other questions. 1988: none directly; 1972 asked whether Sunday ringing had worsened or improved in recent years.
Reasoning: recorded. The specification also planned the same question for individual ringers, so a ringer’s view of their tower could be set against the captain’s as a check on optimism; it is not in the build: QM-05.
A3 · How many bells can be rung in changes?
Asks: a number; 0 if none; blank if not known.
Why: a check on Dove’s bell count, which can be out of date. The specification asked for a soft warning, never a block, if it differs from Dove. Origin: specification v0.1. 1988: the captain’s form opened “…which has …… bells”.
Reasoning: recorded. D-066 suggested “rung full circle in changes” as an optional clarification; not built, and judged harmless.
A4 · Who usually rings here?
Asks: tower band, which rings only here; band that regularly rings here and at other towers too; ringers from other towers’ bands, no tower band; visiting bands by arrangement only; society or guild band; no band; not known; other.
Why: it records the kind of band, including the two kinds of sharing, which need opposite treatment: a lending tower’s band is counted at both towers, a borrowing tower’s ringers belong elsewhere. Society-band towers, such as the College Youths’ London towers, are answered by the society’s steward. Origin: D-061 (31 October) replaced the earlier three-option question and the separate yes/no on shared bands (A-B3). 1988: the categories are 1988’s incumbents’ categories, so Table 11.1 can be compared directly, from the tower’s side.
Reasoning: recorded. The full account, and the small differences between D-061 and the build, are in Where people ring.
Open: minor A new answer for towers with more than one band (D-081, 17 November; build item QM-A21).
Simon Linford asked on 17 November what a university band ringing at someone else’s tower should do: “One tower one band is fine. One band more than one tower is catered for. But not one tower multiple bands?” A4 had no honest answer for the host captain. D-081 adds “Tower band, and another band also rings here regularly (e.g. a university or society band)”, routed like the first answer and recoded to it for the 1988 comparison, which is what a 1988 incumbent would have ticked. It also flags the tower, so all its returns are kept (D-049 is amended for flagged towers). Mark’s principle: honour people’s own account of how their tower works, and give them several places to say so.
Core: about ringers at your tower
A-B1 · How many people do you count as members of this tower’s band?
Asks: a number, with 1988’s definition: everyone who rings here on Sundays and/or at practice whom you consider a member of the band, including irregular and infrequent attenders; people away at school or university if they ring here regularly for part of the year; learners only if they can ring unaided. Two sentences added in 2026: count by who rings with the band, not association or church lists; if the band also rings at other towers, count the whole band.
Why: this is the series that compares directly with 1988’s headline, about 41,000 ringers from band lists, about 36,000 after removing double counting. The 1988 estimate rested on the band lists, not on the count of who rang on the survey Sunday, which 1988 used for silent towers. So the definition must not drift. Origin: specification v0.2, from 1988’s instruction to its agents and the two rulings it issued (students and learners); the added sentences are D-060. 1988: word for word, plus the two 2026 sentences.
Towers with more than one band (D-081, 17 November). One sentence is to be added: “If more than one band rings here regularly, count everyone in all of them” (build item QM-A22). Without it, “whom you consider one of the band” lets a host captain leave out a university society that rings there, while its students name the tower as theirs. This keeps the count what 1988 meant by it, the people who ring here, and keeps the two ends of each tie matched.
Reasoning: recorded. A whole number, not a band: the earlier drafts offered “Less than 6 / 7–10 / 11–15”, and the August specification rejected banding, with Mark’s agreement, because a count that is added up and modelled loses information when banded, for no saving in effort (August 2026). Full account in Where people ring.
A-B2 · Of those, how many are within each age range?
Asks: numbers under 18, 18–21, 22–39, 40–59 and 60+.
Why: the age split of every band is the one total the census will know for all responding towers, so ringer returns can be weighted to it. It also gives the under-18 count for Ringing 2030, as a number only, never names. Origin: specification v0.3; the bands are explained in Age bands. 1988: the captain gave the band’s ages by sex in five bands; 2026 drops the split by sex.
Reasoning: partly recorded. Why the split by sex was dropped is not on the record: QM-06. The specification’s soft check that the ages add up to A-B1 is still a “[TO DO]” note in the build (Q38), and the specification asked for it never to block: QM-07.
A-B4 · Which other towers share ringers with this one?
Asks: the other towers, chosen with the multi-tower picker. Shown only after A4’s two sharing options.
Why: naming the towers lets the analysis tell a tower rung by another tower’s band from a silent one, remove double counting across shared bands, and test the place clusters against how bands really work. Origin: D-058 (29 October) replaced a count of other churches with the names; D-061 moved it behind A4. 1988: asked only the number of other churches.
Reasoning: recorded. See Where people ring.
Core: ringing during census week
The census week is Monday 16 to Sunday 22 November 2026, a dull mid-term week like 1988’s, with students in their term-time towers and no festivals. The surveys open on Monday 23 November and ask about that week. Nobody chased is contacted before the week ends, so that chasing cannot change what happens in the week itself.
A-C i1 · Introduction: full-circle ringing only
Asks: nothing; it says “Count full-circle ringing only” and lists what not to count. Why and origin: D-066; see Full-circle ringing.
A-C1 · How many people rang at the tower in census week?
Asks: a number. Count anyone who took hold of a rope that week, for any purpose (service, practice, wedding, quarter or peal, handling lesson), each person once, including visitors and learners.
Why: the modern, objective measure, one of the census’s four time anchors. It does not depend on anyone’s idea of who is “in the band”. 1988 found turnout averaged about 55% of band size, which gives a check on each return. Origin: specification v0.1, with “took hold of a rope intending to ring” as the definition of having rung. 1988: none; 1988 counted ringers present on the Sunday only.
Reasoning: recorded. “Include visitors and learners for now” reads as an instruction that may change: QM-08.
A-C2 · Of those, how many were visitors?
Asks: a number. A visitor is “someone you would not expect to see here in a typical month”.
Why: the tower’s half of the measure of ringers who ring at several towers; the ringer’s list of towers rung that week is the other half. Disagreement between them is information, not error (specification §4.3). 1988: the Sunday table asked for “visiting ringers (ie, people who don’t normally ring at the tower on a Sunday)”.
Reasoning: recorded. A rotating band’s own members can meet this test at their own towers; see Where people ring.
A-C3 · How many people rang at your tower on Sunday 22 November?
Asks: a number; 0 if none.
Why: all of 1988’s landmark findings are about Sundays: 62.8% of towers rang regularly on Sundays, 30.2% were ringable but silent, 7.0% unringable; 36% of ringers did not ring on the survey Sunday. A week-long count cannot be compared with any of them, so Census Sunday has its own question, and the specification calls it “not optional”. Origin: specification v0.2. 1988: Tower captains question 7.
Reasoning: recorded.
A-C4 · Were the bells rung at all on Sunday 22 November?
Asks: yes / no / not known.
Why: dead and silent towers are where this is the only observation of ringing there is, and they are what the 1988 comparison is largely about. It is asked only where A-C3 was not: towers with no band or unringable bells, and returns an enumerator makes from observation. For everyone else the answer follows from A-C3, and asking both would be the same question twice in a five-minute Core (specification v0.3). 1988: Incumbents question 1, “Were your bells rung last Sunday?”.
Reasoning: recorded. See Full-circle ringing on whether the “full circle” cue reaches these towers.
A-C4b · Was there a service in this church on Sunday 22 November?
Asks: yes / no / not known / not applicable (no services held here).
Why: a silent Sunday cannot be read without it. A “No” at A-C4 or a 0 at A-C3 could mean no ringers, or no service that Sunday. Church of England figures for October Sundays show that a large minority of its churches, and a majority of rural ones, do not have a service every Sunday, so many silent Sundays will be rota Sundays. Origin: D-067 (Mark, 5 November). Asked of every rep, not only after a “No”, so it can also give the share of towers rung when there was a service. Placed after A-C4 so the 1988 questions are answered in the same context as before. 1988: split its shared-band figures by whether the tower rang on the survey Sunday (Table 11.1).
Reasoning: recorded. Routing to every rep, given A-C4 itself is conditional, still to confirm (D-067).
A-C5 · How many people rang at your tower at least once in the last 12 months?
Asks: a number, “a best estimate is fine”, leaving out one-off visitors on an outing or a visiting peal band.
Why: some towers ring on a cycle, every four or six weeks, and some bands rotate round several towers, so a silent census week can be a true answer from a healthy tower. The twelve-month count is immune to that. It also catches occasional ringers beyond the band. Origin: specification v0.2, one of the four time anchors. 1988: none.
Reasoning: recorded. The planned soft check against the week’s count is a “[NB …]” note in the build (A-C5 advisory): QM-07.
A-C6 · How many people are learning to handle a bell but can’t yet ring unaided?
Asks: a number.
Why: learners who can’t yet ring unaided are left out of the band count by 1988’s definition, so they are counted here, as the recruitment pipeline Ringing 2030 cares about. Origin: specification v0.1. 1988: none in this form; 1972 asked for the number of learners.
Reasoning: recorded.
Open: minor A-C6 has a typo: “can’t yet ringing unaided” should be “can’t yet ring unaided”.
Core: completion
Proposed · What prompted you to fill this in today?
Open Proposed by Mark on 7 November 2026, not yet agreed or built. It must be agreed, and logged as a decision, before the build freezes for 23 November.
Asks (self-completers only, i.e. not on the enumerator path; tick all that apply):
- A notice or message sent to lots of ringers (Ringing World, an association or branch email or newsletter, social media, a notice in the tower)
- Someone from the census contacted me about this tower’s return in particular: by phone, in person, or an email just to me
- Someone in my band or at my church asked me to
- Nobody: I knew about it and got round to it
- Not sure
Why: the calibration rests on telling towers that replied on their own from towers that replied because they were chased. The working rule is a date: a sampled tower that returns by the cut-off (provisionally Sunday 29 November) is voluntary, and one that returns later counts as chased (see _sampling-protocol.md). That rule has two weak spots, and this item measures both:
- Early contact. Enumerators hold provisional lists from 23 November and are asked not to contact their towers before 30 November. “Finding contact details” can slide into asking someone at the tower. A personal tick on a return before the cut-off is the clearest sign of that.
- Late voluntary returns. Some sampled towers will return after the cut-off without ever being contacted. The rule counts them as chased, which is conservative. A “nobody” or “general notice” tick lets analysis treat them as late-voluntary instead.
It also checks the completion mode recorded by Q46 and Q47. And it shows which general channels actually moved people, including association reminders, which is useful for the comms plan in its own right.
Wording. The personal option is defined by who and what (“someone from the census”, “about this tower’s return in particular”), not by channel. A mail-merged association email that greets them by name is still general, and so is a volunteer announcing the census to a whole branch. Neither is about this tower. “An email just to me” covers the one case where the channel matters. The general option names its examples, so respondents can recognise an association reminder as general without having to judge it. “Someone in my band or at my church” keeps a captain prompted by their own ringer or incumbent out of the personal option. “Not sure” stops people who can’t tell from guessing.
Limits. Self-report, so evidence, not proof. A captain whose enumerator is a friend nearby may tick “band or church”. Recall weakens for returns made weeks after a contact. Read it together with the return date and the sample flag. “Tick all” adds a little to the five-minute Core, but it is one screen; a single “main reason” would be quicker but would lose overlaps, such as an association reminder followed by a phone call.
Origin: Mark, 7 November 2026, while settling when enumerators get their tower lists. Wording drafted with Claude the same day, replacing a first yes/no draft (“had a census volunteer been in touch?”) that could not separate personal chasing from general reminders. Bears on QM-03. 1988: none. In 1988 agents carried the forms, so every return was agent-mediated.
Reasoning: recorded (discussion of 7 November 2026; not yet logged as a decision).
Q46 (id 1616891732) · May we contact you about this return?
Asks: optional contact details, used only to resolve queries about the return and deleted once results are compiled.
Why: some returns will need checking. Contact details are optional, held separately from analysis, and never asked of individual ringers. Hidden on the enumerator path. Origin: the specification’s required email became optional in the data protection review.
Reasoning: recorded.
A-E4 · Please complete the individual survey yourself as well
Asks: nothing; an instruction.
Why: in 1988 tower captains were surveyed separately from ordinary ringers and were systematically different (87% male against 58%, older). If captains treat the tower return as “their” survey and skip the individual one, the individual returns silently lose that group. So the census says it plainly (August 2026 specification, from the 1988 report). It also asks the rep to encourage the band, since the individual survey link is general, not sent through reps.
Reasoning: recorded. The specification also had a question “Will you circulate the individual survey to your band?” (yes / maybe / no); it is not in the build: QM-09.
Q47 (id 1617089052) · That’s your tower counted. Some optional questions?
Asks: yes / no.
Why: this is the Core / Tail break. Nothing in the Tail may block the Core, and the Core must save as a complete return here. Origin: specification v0.2; made a required question on 30 October so the pipeline can tell a finished Core from an abandoned one.
Reasoning: recorded.
Tail: how your tower runs (optional, never chased)
A-T3 · How often is there a practice?
Asks: every week / 2–3 times a month / once a month / less often / never / not known.
Why: practice is where ringers are made, and a tower’s practice pattern is a basic fact about its health. Origin: specification v0.2 (“weekly / less often / none”); the finer options are in the build. 1988: captains’ question 10, with six options including “once a week in part of the year only”.
Partly recorded. Who widened the options, and whether matching 1988’s categories was considered: QM-10.
A-T2 · How many bells are typically rung at practice?
Asks: a number.
Why: a measure of what ringing actually happens, which Dove’s bell count does not give. Origin: added at the questionnaire review meeting (summer 2026). 1988: captains’ question 10 asked “are all the bells usually rung?”.
Reasoning: recorded.
A-T4 · How often are the bells rung for Sunday service?
Asks: every week / 2–3 times a month / once a month / less often / never / not known / not applicable (no services).
Why: the usual pattern, against which census Sunday is read. A tower silent on 22 November but rung most Sundays is a different case from one rarely rung. Origin: specification v0.2. 1988: captains’ question 8 (every Sunday / most / at least monthly / less often) and incumbents’ question 3.
Partly recorded. As with A-T3, why the options differ from 1988’s: QM-10. A-T4 asks how often the bells are rung for service, not how often there is a service, which is why A-C4b was added to the Core (D-067).
A-T5 · How often does Sunday ringing depend on visiting ringers?
Asks: never / occasionally / most weeks / could not ring without them / not known.
Why: it shows towers kept going by helpers from elsewhere. A captain may under-report this, so it is set against the ringers’ own accounts of where they rang. Origin: specification v0.2. 1988: none directly. 1972 asked “Are you regularly assisted by ringers from other towers on a Sunday?”.
Reasoning: recorded.
A-T6 · How would you describe the regular congregation?
Asks: growing / stable / declining / not known / not applicable (no regular congregation).
Why: congregation health is the census’s main threat. It may drive both whether a tower replies and how many ringers it has, and nothing outside the census measures it for every tower. If the question works, it goes into the model. If most reps answer “not known”, that is the result: congregation health becomes an unmeasured cause to be handled by sensitivity analysis instead (specification, “pilot item”). Origin: the pilot objectives in the methodology outline (April 2026). The clergy interview that 1988 used for this (93% response) was dropped because there is no route to clergy without a central contact list (Mark, 19 October). Church of England attendance data may help after fieldwork (Q-036).
Reasoning: recorded.
Tail: tower resources
A-T1 · How would you describe your bells?
Asks: agree or disagree with: well maintained; easy to ring; effective sound control; easy to hear.
Why: these are facts about the tower, so they belong on the one-per-tower return rather than being asked of every ringer. Origin: Elva’s grid from her first draft, moved to the tower rep in August. A fifth row, “welcoming belfry”, was cut: it is a person’s experience, not a tower fact, and it overlapped with “I don’t feel welcome” in the individual survey (August specification).
Reasoning: recorded. A “Not known” column was proposed on 30 October and is not in the build: QM-11.
A-T11 · What teaching and training devices does the tower have?
Asks: tick all of simulator, dumbbells, Hawkear or Strikeometer, CCTV, handbells, belleplates, metronome, boomwhackers, whiteboard, not known, none, other.
Why: not recorded. Origin: added in the build; the specification marks it “NOT IN SPEC”. A yes/no gate (A-T10) before it has gone.
Reasoning: not yet recorded: QM-12.
A-T12 · Are these available to others outside the band?
Asks: yes / no / not known.
Reasoning: not yet recorded: QM-12.
Tail: recruitment
The three questions here were added after the freeze, on 16 November, to test whether moving away has changed who towers teach (D-079). Their shared reasoning is in Teaching payoff; each entry below gives the short version. The two questions that stood here before, Q31 and Q32, were removed (D-080).
Open: minor In the build of 16 November, A-R5 sits above the “Recruitment” heading (Q30) instead of below it: QM-A16.
A-R5 · In the last 12 months, how many ringers have…
Asks: a grid of four rows: became members of the band by learning to ring unaided here; joined the band after moving into the area (already able to ring); left the band because they moved away from the area; stopped ringing altogether. Each row: 0, 1, 2, 3, 4, 5+, not known.
Why: the year’s flows in and out of the band, so the analysis can see which kinds of place gain and lose ringers by moving, and how many movers are lost to ringing altogether. Set against the band’s ages (A-B2), the “stopped” row shows how many ringers a band loses each year to stopping, compared with how old it is. Origin: Mark, 14 to 16 November, from 1988 questions 20 and 31. 1988: captains’ question 31 asked how many ringers had left in the last 12 months “to move to other towers in the area”, “because they have moved away from the area” and “because they have given up ringing” (Table 9.5). Two of the three rows compare. The joining rows are new: 1988’s question 20 counted all proficient joiners as one number, and question 19 counted learners who started, not who reached ringing unaided.
Reasoning: recorded. The first row reads “Became members” after the stem “how many ringers have…”: QM-A17.
A-R3 · Which age group makes the best recruits?
Asks: under 15, 15–17, 18–21, 22–39, 40–59, 60 or over, no view. Single choice.
Why: what captains prefer, as distinct from who they get. A shift towards older groups since 1988 would be the hypothesis in captains’ own answers. Origin: Mark, 14 to 16 November, from 1988. 1988: captains’ question 26: of under 15, 15–19, 20–39 and 40 or over, which make the best recruits, one choice. 13% of those naming a best group chose 40 or over; 31% chose under 15 (Chapter 7). Under 15 compares exactly; 40 or over is 40–59 plus 60 or over. Compare tower captains only (role question Q46).
Reasoning: recorded. The agreed last option was “No real difference / no view”; the build says “No view”: QM-A19.
A-R4 · And which age group makes the least promising recruits?
Asks: the same groups as A-R3. Single choice.
Why, origin and 1988: as A-R3. In 1988, 39% named 40 or over as least promising and 32% named under 15.
Reasoning: recorded. The stem says “which make”, A-R3 “which makes”: QM-A18.
Tail: anything else
Q34 · Anything else about this tower’s situation?
Asks: free text, optional.
Why: the one free-text box on the tower return, for what the questions miss. It is left out of published outputs, and anything identifying, or about health, is removed before analysis (specification §8; data protection review).
Reasoning: recorded.
Q45 (id 1617003378) · Survey testers only
A feedback box for the pilot. To be removed before the census opens (build list, 30 October).
Considered for the tower return and not asked
| Considered | What happened | Reasoning |
|---|---|---|
| A four-question interview with the clergy of chased towers | Dropped (Mark, 19 October) | Without a central contact list there is no route to clergy. 1988’s best response rate (93%) came from this, and clergy are the only authoritative source on congregation health, so it is a real loss, handled by A-T6 and sensitivity analysis |
| Separate yes/no: “Does your band also ring at other churches?” (A-B3) | Removed by D-061 | A4 option 2 asks the same thing, and A4 can also record a tower that borrows ringers |
| Counting how many other churches (the first A-B4) | Replaced by naming them (D-058) | The count follows from the names, and names let returns be joined |
| Band size in bands (Less than 6 / 7–10 / 11–15) | Rejected (August) | A count that is added up and modelled loses information when banded |
| “How did you establish the position?” for enumerator returns (A-F1) | Folded into Q47 at the start (30 October) | The source should be stated before the numbers |
| Hand chiming as an optional Tail question or a Core count | Not adopted (D-066) | See Full-circle ringing |
| Number of band members who are association members | Not asked | Association registers are better collected from associations, after the census (D-060) |
| 1988’s long captain’s form (tower officers, charges for ringing, books used, The Ringing World, striking contests, the captain’s own ringing) | Not asked | The short Core is the lesson of 1988’s response rates; the individual survey covers the captain’s own ringing |
| Q31: “How do you find getting learner recruits compared with a year ago?” (1988 captains’ question 24) | Removed after the freeze (D-080, Mark, 16 November) | Comparing now with an impression of the past is a poor way to track a trend, and one tower’s year is too much a matter of chance for its captain’s impression to mean much. The trend comes from comparing the 1988 and 2026 counts. In 1988 the answer also followed the captain’s own age (28% “increasingly difficult” among captains aged 20–39, 64% among the oldest; Table 17.1), and 2026 does not ask the respondent’s age |
| Q32: “And do recruits these days…” give up / stay / don’t differ (1988 captains’ question 25) | Removed after the freeze (D-080) | As Q31. The “stopped ringing” row of A-R5 now measures leaving as a count |
Part 3. The Individual Ringer Survey, question by question
What the survey is for. Every ringer completes it, tower reps included, and so can people who no longer ring. It gives the make-up of the ringing population (age, gender, ability, how and when people learned, motivation, what brings people back), tied to towers through the tower list. There is one general link and no names or email addresses. Duplicates are found from the answers at analysis and controlled by QuestionPro’s cookie setting (implementation design §3).
Where the questions came from. Elva designed the first full survey, building on surveys she had run before. Vicki built it in QuestionPro. Tina commented on it. At the questionnaire review meeting in the summer (probably the line-by-line workshop of 14 July 2026; QM-41), the committee reconciled the versions. This page calls it “the review meeting”. Mark then specified the census-week and tower questions and the links to the tower return (Individual Ringer Survey specification v0.2, summer 2026). A pilot round, reported on 21 October, had 98 starts and 43 completions. The committee agreed four changes on 16 October (D-039). Mark and Elva held a long design session on the individual survey in late October, which settled several of the current wordings; its transcript was not available for this page, and entries that depend on it say so.
Opening
Q30 (id 1616566439) · Welcome
Asks: a welcome; why the census matters; that ringers who ring regularly, occasionally or not at all are all wanted; confidentiality; “around 10 minutes”; and that it is hard to go back once you move on.
Why: the census needs lapsed and occasional ringers as well as keen ones, so the welcome invites them by name. Origin: Elva’s draft. “Around 10 minutes” came from the pilot, where one tester took 10–15 minutes and QuestionPro’s average was 8. The warning about going back reflects a QuestionPro limit on pages with branching.
Partly recorded. The pilot raised two points not yet reflected: the welcome does not say that the census wants what people did in census week rather than a typical week, and “No individual will be identified” sits oddly with the fact that two named towers can identify someone to the census team: QM-14.
Q50 (id 1616947431) · Before you start: the privacy notice
Asks: the same short notice as the tower return, with “results are published only as totals or anonymised quotes”.
Why and origin: as the tower return’s notice: under legitimate interests the notice is given, not agreed to (data protection review, accepted 16 October). The “anonymised quotes” clause covers the free-text answers. Reasoning: recorded. The address and link are placeholders: QM-02.
About you
Q1 · How old are you?
Asks: under 18, 18–21, 22–29, then by decade to 80+.
Why: age is the main demographic of the census, and the tower return gives every band’s age split, so ringer returns can be weighted to it. Age band only, never date of birth. Origin and 1988: see Age bands. Reasoning: recorded.
Q2 · What is your gender?
Asks: male / female / non-binary / prefer not to say / other.
Why: gender is one of the primary objectives’ named characteristics, and 1988 asked sex on every form, so the balance can be compared. It is also how the analysis can check whether some experiences, such as feeling intimidated in some towers, fall unevenly, without naming gender in those questions (Tina’s concern, October). Origin: the review meeting decided “What is your gender?” with Female / Male / Prefer not to say / Other. 1988: “Are you male or female?”.
Partly recorded. “Non-binary” is in the build but not in the August decision; Richard Heys asked in September how the census would handle gender: QM-15.
Q3 · How old were you when you first learned to ring?
Asks: under 12, 12–14, 15–17, 18–21, 22–29, then by decade to 80+.
Why: with current age and how long people have rung, it separates a real effect of starting young from survivorship (young starters who struggle drop out; late starters who remain are self-selected). That was a requirement of the methodology outline from the start. Origin and 1988: see Age bands. Reasoning: recorded.
Q35 · How did you first become involved in ringing?
Asks: family, friend, church, school, university, public or open day, social media or online search, other.
Why: routes into ringing matter to Ringing 2030’s recruitment. “University” was added by the committee on 16 October (D-039): a tester asked for it to show how important university societies are in bringing people in. Origin: Elva’s draft. 1988: “What influenced you to take up ringing?” was open text, and a separate question asked which family members rang.
Partly recorded. Why the rest of the list has these options: QM-16.
Q55 · Do you still ring regularly at the tower where you learned?
Asks: yes; no, but it’s near enough that I could; no, it’s too far away now; no, for another reason (e.g. there’s no ringing there now).
Why: whether the tower that taught someone still has them. With age when learned (Q3), years ringing (Q4), “University” as the route in (Q35) and the place type of the main tower, it shows how many young starters now ring far from where they learned, by the decade they learned in. Added after the freeze, on 16 November, to test whether moving away has changed who towers teach (D-079); the shared reasoning is in Teaching payoff.
Origin: Mark, 14 to 16 November. It asks about the tower where you learned rather than comparing it with your main tower, so that people in several bands can answer simply. It gives no distance: “near enough that I could” is the respondent’s own judgement, which works the same in a village and a city. It sits here, after “how did you first become involved”, so that when, how and where people learned come together, before the tower pickers.
1988: tower captains were asked about themselves, “And did you learn to ring at the tower you now ring at?” (yes/no). Not asked of ordinary ringers.
Reasoning: recorded. Lapsed ringers will answer “No” for reasons that have nothing to do with moving; the analysis separates them with Q34 (“Not at all these days”). The build shortens the first option to “Yes”, which the stem makes clear.
Q4 · How long have you been ringing?
Asks: a drop-down from less than 1 year to 61+ years.
Why: with current age and age learned, it places a ringer in time (how long since they started) and lets the pipeline flag impossible combinations. A tester asked whether it was needed, since it can be worked out from the two age questions. The answer recorded at the time: the working-out gives a wide range (aged 40–49 and learned at 12–14 gives 26 to 37 years), so the question stays. Origin: Elva’s version, extended to 61+ at the review meeting.
Partly recorded. The August specification proposed asking the year first learned instead, exact for adults, which would give 1988’s “decade learned” series and an exact check on the ages; it is not in the build. And the options as built overlap at 2 years (“1–2”, “2–3”) and leave a gap at 3: QM-17.
Q6 · Have you ever stopped ringing for a year or more (excluding Covid)?
Asks: yes / no / NA.
Why: lapsing and returning is one of the census’s objectives: “reasons for lapsed participation (for follow-up studies)”. Covid is excluded because an enforced, universal break would swamp the voluntary breaks the question is after. Origin: decided at the review meeting. 1988: “Have you had a period (say of a year or more) when you did no ringing?” — close enough to compare.
Reasoning: recorded. Who “NA” is for: QM-18.
Q45 (id 1616825993) · How long was your longest break, in years?
Asks: a number of years, about the longest break if several, excluding Covid.
Why: the length of a break, with how long someone has rung, gives years actually ringing. Origin: the review meeting asked for the bands to align with “how long ringing”; it is built as a number. A tester found it hard to recall exactly and asked for ranges.
Partly recorded. Whether to use ranges, as the pilot suggested, is not on the record as decided: QM-19.
Q8 · What encouraged you to return after that break?
Asks: tick all of: more time; to support the local ringers; missed ringing friends; was asked back; missed the ringing; a special occasion; moved house; children got older; exercise; a change in circumstances (e.g. health, work or family); to support the church; “I never returned — I don’t ring these days”; other.
Why: what brings people back is one of the primary objectives’ named items, and directly useful to recruitment. Origin: Elva’s list, with “moved somewhere”, “children older” and others added at the review meeting (Elva to finesse from earlier surveys’ “other” answers). The circumstances option names health as one example among several, so that ticking it reveals nothing about anyone’s health (D-039). Reasoning: recorded for the health wording; partly recorded for the rest. The “never returned” option makes this the place lapsed ringers are identified: QM-20.
Q34 · In a typical week, on how many days do you ring?
Asks: from “6+ days per week” to “once per week”, then several days a month, once a month, special occasions only, “not at all these days”.
Why: how often people ring is a secondary objective, and it is one of the three measures of keenness now used to compare early and late replies, since the five-year question was dropped (D-039). The last option also marks lapsed ringers. 1988: captains were asked “How many times a week (on average) do you go ringing?”.
Partly recorded. The stem says “in a typical week” but half the options are monthly or rarer: QM-21.
Bands overview
Q48 · What do you ring?
Asks: tower bells (full circle) only / handbells only / both.
Why: handbells are a different instrument, not a kind of band membership. Handbell-only ringers have no tower, so the frame, the Lucky Dip and the Golden Ticket cannot reach them; they are counted from voluntary returns as a lower bound, not estimated. The answer also routes them past the tower questions. “Full circle” added by D-066. Origin: proposed in early October when the single band-membership question was split (see Where people ring); handbell-ringer numbers are a secondary objective. Reasoning: recorded.
Q46, Q49, Q50 · Bands, and bands that rotate
Asks: how many ringing bands you are a member of; whether any rings regularly at more than one tower (with the benefice example); if so, how many towers it covers.
Why: 1988 found that about one ringer in five could be reached through more than one tower, because of shared bands (about 12%) and membership of several bands (about 8%). That is what took 1988’s estimate from 41,000 to 36,000. These questions measure the same thing and give the weights. Origin and 1988: see Where people ring. Reasoning: recorded.
About where you ring
Q37, its follow-up, Q27 and Q53 · Your towers
Asks: one tower where you regularly ring, picked from the list (“this allows us to link your response to Dove’s Guide and does not imply that this is your only or ‘home’ tower”); whether that tower’s captain would count you as one of the band; whether other towers would; which.
Why: these link each ringer to the towers whose band counts include them, which is what the weights and the removal of double counting need. The “not your home tower” wording follows Tina’s Dundee point. Origin and 1988: see Where people ring. Reasoning: recorded.
Census week activity
Q39 · Did you ring anywhere during census week?
Asks: yes / no / can’t remember, with the definition “taking hold of a rope intending to ring”, handbells counting, chiming and carillons not.
Why: the ringer’s side of the census-week count. The “No” answers are the temporarily inactive (away, unwell), which is exactly what a one-week count misses. Origin: the census-week module specified in August, matching the tower return’s definition so both ends measure the same event. D-066 added the chiming sentence. Reasoning: recorded.
Q41 · Which towers did you ring at that week?
Asks: every tower, including your own; skip if handbells only.
Why: a bare count of towers would correct the total in aggregate, but keen ringers who ring at many towers reply more, so the count needs modelling by kind of tower, which needs to know which towers. It is also the ringer’s half of each tower’s visitor count. Origin: August specification. Reasoning: recorded. Whether a tower answer becomes required after “Yes” is open (Q-032), as is the route for handbell-only ringers.
Q40 · Why not? What stopped you ringing that week?
Asks: tick all of: my tower doesn’t ring every week and it wasn’t our turn; no ringing was arranged; I was away; personal circumstances (e.g. health, work, family or caring); I’ve stopped ringing; something else.
Why: separates a rota week or a holiday from lapsing, so that a “No” at Q39 can be read. The first option exists for rotating bands, whose silent weeks are expected. Health is one example among several (D-039). Partly recorded: the option list’s origin: QM-22.
Q42 and Q43 · Did you ring on Sunday 22 November, and where?
Asks: yes / no / can’t remember; then the towers.
Why: 1988’s Sunday findings need a Sunday answer: 36% of 1988’s ringers did not ring on the survey Sunday. 1988: “On how many separate occasions did you ring on Sunday 20 November?” (once / twice / more than twice / you didn’t). 2026 asks yes or no and where, not how many times.
Partly recorded. Why the number of occasions was not kept: QM-23. Pilot testers disliked entering the same towers three times; the picker now offers towers already named as quick picks (D-054).
Q50 (id 1616875943) · Have you rung a peal, quarter or BellBoard performance in the last 12 months?
Asks: tick any of a peal, a quarter peal, any other performance published on BellBoard.
Why: two jobs. It is one of the three keenness measures for comparing early and late replies (D-039). And it links the census to outside counts: Pealbase and BellBoard know how many distinct people rang a recorded performance in a year, so the share of them who reply shows how much keen ringers are over-represented (implementation design §6). Reasoning: recorded. There is no “none of these” option, so a blank cannot be told from a skip: QM-24.
Ringing ability and development
Q5 · What is your current ringing capability?
Asks: one of: learning to handle a bell (Learning the Ropes level 1); simple rounds and call changes (level 2); plain methods, e.g. Grandsire, Plain Bob (levels 3–5); competitive standard call changes; advanced methods, e.g. Stedman, Surprise.
Why: ability is a primary objective. Origin: Elva’s draft, with Tina’s changes (summer 2026): drop the word “method”, because early learners may not know it; add the Learning the Ropes levels, which newer ringers know; and move competitive call changes and advanced methods to the end. The review meeting adopted these, “to finesse from pilot”. Claude noted at the time that competitive call change ringing is a tradition of its own, not a step on the method ladder; Mark agreed to test the order in the pilot. 1988: asked which methods people could ring inside; not directly comparable.
Reasoning: recorded. What the pilot showed about the order, and whether call changes should be a separate tick: QM-25.
Q52 (id 1617082003) · Which of these could you ring confidently now?
Asks (only after “Advanced methods” at Q5): tick any of Cambridge Surprise Minor or similar; Surprise Major (Yorkshire, Cambridge, Lincolnshire or Double Norwich); London or Bristol Surprise on any number of bells; none of these yet. “On an inside bell, through a quarter peal without help.”
Why: it places method ringers in Simon Linford’s ringing zones without asking anyone to rank themselves. BellBoard already describes the top end well, so the census need only find the Red zone ringers who rarely appear in recorded performances, and the gap between what people can ring and what they get to ring. It also gives a check against BellBoard’s count of about 3,300 Red zone ringers a year. Origin: Mark and Elva, after a forum question and an exchange with Simon Linford (19–28 October). Design choices: ask what people could ring now, not what they have rung, because what people have rung mostly reflects opportunity; no Stedman, which adds only about 5% to BellBoard’s Red zone count; no columns for numbers of bells and no Black zone row, which would read as a league table; Delight and Double Norwich named for local pockets where they replace Surprise; “none of these yet” as a graceful way out.
Reasoning: recorded (implementation design §3; analysis/ringing-zones.qmd). Vicki argued on 21 October for keeping the census a baseline and leaving zones to a later survey; the follow-ups are marked “not yet agreed by the committee” in the design but are in the build: QM-26.
Q52 (id 1617082235) · When did you last ring any of the above methods?
Asks: in the last month / 1–12 months ago / 1–5 years ago / 5+ years ago / never.
Why: with the question above, it measures the gap between ability and opportunity, Simon Linford’s “unfulfilled potential”. Reasoning: recorded as above. The design lists four options; the build adds “Never”: QM-26.
Q31 (id 1616566725) · How confident are you with the following?
Asks: for teaching basic bell handling, conducting touches, quarter peals and peals, running a tower, and basic steeplekeeping: not applicable / no experience, not confident, somewhat confident, confident.
Why: the census needs to know who can teach, lead and look after bells, because in many areas the shortage of teachers limits recruitment (a point Richard Heys also made in September). Origin: Elva’s grid. The scale was changed after the pilot: three testers and the analysis team found that the old first option, “I don’t do this”, mixed up not doing something with not being confident at it. The new first option, “Not applicable / No experience”, separates them. 1988: asked whether ringers had ever conducted a quarter or a peal, and which tower offices they had held.
Reasoning: recorded for the scale; partly recorded for the items. “Bell handing” in the first row should be “handling”.
Q10 · How do you feel about developing your ringing?
Asks: I’m actively working on it / I’d like to, if the opportunity came up / I’m happy with the ringing I do now / NA.
Why: a gate: the first two answers go on to the three development questions. Origin: pilot testers disliked the earlier yes / no (“Would you like to progress further?”), and some who said “No” were wrongly routed on. Mark asked for a non-binary version on 26 October. “I’d like to, if the opportunity came up” catches people with ambition but no opportunity, the gap the census most wants to see. “Happy with the ringing I do now” lets people decline with pride.
Partly recorded. The August specification, with Mark’s agreement, had used 1988’s own gate, “Do you feel that you would like to advance further in your ringing, but are unable to?”, because exactly half of 1988’s ringers said yes and it gives a 38-year comparison. The current wording loses that comparison. Who “NA” is for is also unclear: QM-27.
Q11 · Which aspects of your ringing would you like to develop?
Asks: tick any of 12 aspects, from bell handling to bell maintenance and tower management.
Why: shows what training associations should offer. Origin: Elva’s list. Partly recorded: the list’s origin: QM-28.
Q12 · Which of the following limit you developing your ringing?
Asks: tick any of: limited training or development opportunities; not enough experienced ringers or conductors; limited opportunities at my usual tower or band; not always feeling included, welcomed or given opportunities; lack of confidence or finding progression difficult; lack of time; difficulty travelling; personal circumstances (e.g. health, caring or family); no limits; other.
Why: the barriers to progress are where associations and Ringing 2030 can act. The list includes being included and welcomed because Tina raised, in the summer and again in October, that some people, women especially, are intimidated in some towers and that “not being asked” is a barrier in itself. Origin: Elva’s list, with additions from the review meeting (confidence; not invited; not feeling welcome). Health is one example among several (D-039). 1988: three reasons (not up to it; the band doesn’t provide the opportunities; the standard in the area is low).
Partly recorded. In August a version grouped by where the barrier sits (me, my tower, my area, being included) was specified to stop options overlapping; in October, rewordings were proposed to name intimidation directly. The build has neither: QM-29.
Q13 · What would help you develop?
Asks: tick any of 11 options, from more chances to practise to “nothing in particular”.
Why: turns barriers into actions. Origin: Elva’s list. Partly recorded: QM-28.
Motivations
Q45 (id 1616825866) · How important are the following to you?
Asks: 15 aspects of ringing (community, tradition, intergenerational activity, learning, being in a band, service to the church, teaching others, service to the community, ringing elsewhere, physical activity, quarters and peals, health and wellbeing, friendship, music, mental challenge), each from “not at all” to “very” important.
Why: motivation is a primary objective, and the census needs to know what keeps people ringing. “Health & wellbeing” stays because valuing wellbeing says nothing about anyone’s health (D-039). Origin: Elva’s grid, “to finesse from previous survey’s ‘other’ responses”. Tina added teaching others and tradition (summer). 1988: asked people to pick two of four attractions (part of the work of the church; mentally fascinating; working in teams; bringing people together); these map roughly onto four rows here.
Partly recorded. Tina also proposed “looking after the bells and the tower”, which is not in the list; the helper text tells people to choose NA, but the scale has no NA; and a tester suggested “Contributing to my health and wellbeing”: QM-30.
Q30 (id 1616566833) · What else is important to you?
Asks: free text. Why: Elva’s method: the “other” answers from one survey improve the option list of the next. Reasoning: recorded.
About your District, Association or Guild
Q50 (id 1616982222) · Which association would you like to give feedback on?
Asks: one society from a list of 86: the Central Council’s 70 member societies, 14 university societies, and two catch-alls.
Why: the section used to assume the respondent’s “main” association. For someone in a territorial association at home, a university society and a national society, that is too ambiguous to analyse, so they choose one (D-039). The question does not ask about membership: Mark’s view is that membership is defined so differently between societies that it is a distraction from who rings. Origin: D-039, D-059 (the Council’s own members; societies outside the Council are mostly names on BellBoard, not organising structures), D-064 (university societies, to encourage them to join and because many younger ringers learned there). The Guild of Clerical Ringers is listed like any other; the census makes no inference about religion from it. Reasoning: recorded (analysis/associations.qmd).
Q32 · How often do you attend District, Association or Guild events?
Asks: multiple times a month to never.
Why: reach of the associations. 1988 found attendance averaged twice a year, and 47% of ordinary ringers attended none, so a large “never” is a finding, not a data problem. Origin: Elva’s draft; the review meeting asked for wording about the individual, not the band. 1988: “How many meetings did you attend in the last 12 months?” (none / one / two or three / more than three). Reasoning: recorded. The options differ from 1988’s, which limits comparison: QM-31.
Q53 (id 1617132603) · How valuable are these local activities to you?
Asks: nine activities, from special practices to social events, each on a five-point scale from “not at all” to “extremely”.
Why: tells associations which activities their members value. Origin: Elva’s list; the review meeting listed district practices, youth events, special practices, striking contests, social events, AGMs, quizzes, webinars and training courses. The build has dropped district practices and youth events and added outings and newsletters (QM-32). After the pilot it was moved from a slider to a grid: with “not applicable” at 0 on the slider, a slider left at its start could not be told from a real “not applicable”, and both mixed with low scores (analysis team, October). Reasoning: recorded.
Open: minor The old slider (Q28, id 1616375941) is still in the build beside the new grid, with “[TO DO: Change to matrix]” showing; the grid has no “not applicable” option and two unlabelled points. See QM-32.
Q33 · How well informed do you feel about your association’s activities?
Asks: from “I don’t know anything about what goes on” to “I get too much information”.
Why: associations reach ringers mainly through communications, and Tina asked in the summer for a measure of awareness of the wider ringing world. Partly recorded: whether this is the narrow form of Tina’s idea: QM-33.
Finally
Q29 · How likely are you to recommend ringing to a friend?
Asks: 0 to 10.
Why: a measure of enthusiasm, and one of the three keenness measures used to compare early and late replies since the five-year question was dropped (D-039). Origin: decided at the review meeting. The specification had five labelled points; the build uses QuestionPro’s standard 0–10 Net Promoter question. Partly recorded: QM-34.
Q19 · Anything else about ringing or its future?
Asks: free text, with “Please don’t include health details, or anything that identifies other people”.
Why: the open question at the end of most surveys. The warning keeps special category data and other people’s details out (data protection review). Free text from under-18s is deleted before analysis. Reasoning: recorded.
Survey meta (pilot block)
Q20–Q24, Q51, Q53 (id 1617132601) · Pilot questions, device, how you heard
The block is headed “for pilot responses only”. Q21–Q24 are pilot feedback and go before launch. Two questions in the block may be worth keeping: Q51, the device used, which shows whether phones struggle with the tower picker; and Q53, “How did you hear about this census?”, which measures how well each channel works, and could help the model, since how people hear may shape who replies. Tina asked in the summer for a question on how people get ringing information. Which, if any, stay: QM-35.
Considered for the individual survey and not asked
| Considered | What happened | Reasoning |
|---|---|---|
| “How likely are you to still be ringing in five years?” | Dropped by the committee, 16 October (D-039) | For older ringers the answer mixes enthusiasm with expected health and ageing, so it can’t be read cleanly. The recommend-to-a-friend question carries enthusiasm |
| Questions that collect health data | Reworded, 16 October (D-039) | Health is special category data. Not collecting it is simpler and safer than relying on the research condition. Health now appears only as one example among several circumstances |
| Name and email address | Not asked (February; confirmed 16 October) | Keeps the survey nameless. Duplicates are found from the answers |
| Ethnicity | Noted as “agreed yes” at the 2 June meeting; not asked | Later superseded by the rule of no special category data (implementation design §8). The record of when and why: QM-36 |
| Religious belief or church attendance (a tester’s suggestion) | Not adopted | Special category data; the census makes no inference about religion |
| Occupation and life stage | Dropped, February and August | Less important to the objectives; life-stage questions about children kept for follow-up surveys |
| “Are you currently ringing?” as an opening screener | Decided in August (“upfront as a validation/screener”); not in the build | Lapsed ringers are now identifiable from Q34, Q8 and Q40: QM-37 |
| “Tremendously keen / moderately keen / interested but no more” (1988) | Kept in the August specification; not in the build | 1988’s strongest predictor of practice attendance, and the variable the early-versus-late comparison stands in for. D-039 names frequency, performances and the recommend question as the keenness measures instead: QM-38 |
| Year first learned, exact | Proposed in August, pending data protection sign-off; not in the build | QM-17 |
| Which methods you can ring inside (1988’s checklist) | Kept for the pilot only in August; replaced by the Q52 follow-ups | The capability question covers the purpose; the follow-ups cover the Red zone |
| A “welcoming belfry” rating | Cut in August | Overlaps with “not feeling welcome” in Q12 |
| A ringer’s own view of the state of ringing at their tower | In the August specification, to check captains’ optimism; not in the build | QM-05 |
| A familiarity grid (The Ringing World, ART, CCCBR, the association, national competitions) | Raised by Tina; held back in August pending a length check | QM-33 |
| Ringing World readership, BellBoard use, software, Central Council services (a tester’s list) | Not adopted | Outside the census’s purpose; for a later survey |
| Whether the ringer attended university, polytechnic or college (asked in 1988: 44% of ordinary ringers, 28% of tower captains, Table 10.1) | Not asked | Not recorded: QM-40 |
| A prize draw (as in 1988) | Not adopted | Not recorded: QM-39 |
| 1988’s attitude statements (“Peals do little to advance the quality of Sunday ringing” and others) | Not asked | Not recorded: QM-39 |
Sources for Parts 2 and 3
- QuestionPro designs fetched 5 November 2026:
questionnaire/design/tower_test.md(survey 1603084948) andquestionnaire/design/individual_test.md(survey 1603035043). - Tower Representative Survey — Specification, versions 0.1 to 0.4 (July to 20 September 2026), Google Drive, “Design (questionnaire, analysis)” folder.
- Individual Ringer Survey specification v0.2 and the QuestionPro build sheet (August 2026), drafted with Claude from the review meeting notes; not held in the repository.
- 2026 Tower Survey Plan — Methodology outline (24 March to 4 August 2026).
- Meeting notes: 6 January, 17 February, 21 April, 2 June and 30 June 2026; the questionnaire review meeting (notes in Mark’s discussion with Claude, summer 2026); committee meetings of 20 September, 16 October and 30 October (as recorded in the decision log).
- Pilot feedback, collated 21 October and reorganised by question on 26 October 2026 (not in the repository).
- Decisions D-039, D-049, D-054, D-058 to D-067, in
provenance/decisions.md; build notesquestionnaire/build-changes-2026-10-31.mdand-11-03.md. - Richard Heys, Reflections on the Bellringer’s Census Design, 17 September 2026 (personal capacity).
- The 1972 and 1988 questionnaires, appendices C and D of the 1988 survey, and its chapters 3, 4, 10, 11, 17, 19 and 20.