The agenda · what the money is for
The science that does not exist.
Every other page here is a list of things that are known. This one is a list of things that are not, and it is the answer to the fair question of what a men's entity would actually do with a budget. Four questions. Each one carries the research that already exists, then the sentence in which its own authors say what their study could not answer. We are not claiming anybody got it wrong. We are quoting them on what they could not do, and asking who is supposed to do it.
How to read this page
Each gap opens with the question and why it matters, then two disclosures. The first is the prior art: what has been tried, and the authors' own limitation, quoted. The second is the practical part: who holds the data and what the study would actually be. A gap only appears here if we ran searches to establish the absence rather than assumed it, and if the claim survived being narrowed by what already exists.
Three of these four can be answered from data that has already been collected and is sitting in an archive right now. That is the part worth being angry about.
01
Whether family courts decide differently by parent sex
Across an adequately powered, representative sample of adjudicated custody cases, does the outcome differ by the sex of the parent, holding pre-proceedings caregiving constant?
This is the single most argued question in the whole field, in both directions, and it has never been answered on data capable of answering it. Every position currently rests on studies whose authors say, in the study, that they could not test it.
What already exists, and what its authors say it cannot do
The most cited English study on this question examined 197 court files across five county courts in a six-month window in 2011, of which 174 were parent against parent. Twenty-five cases went to a fully contested final hearing, 14 per cent of the sample; 57 per cent of final orders were marked as consent orders. Its finding on fathers is carried here in full, because it cuts our way and the limitation still stands.
As there were so few applications for contact made by mothers, it was not possible to meaningfully examine the success rates of such applications on the basis of gender. However, applications by fathers for contact were overwhelming successful.
Harding M, Newnham A (2015). How do county courts share the care of children between parents? Nuffield Foundation and University of Warwick.
The 2025 judgment analysis for the review of the presumption of parental involvement covered 245 cases across eight courts. Its Table 4.4 codes the sex of every applicant and respondent: 88 mother applicants against 135 father applicants. Its Table 5.1 codes the outcome. The two are never crossed, and the outcome is recorded as involvement with the non-resident parent without the report ever stating which parent that was.
This has resulted in a patchy dataset from which it was not possible to perform regression or correlation analysis to meaningfully understand the statistical relationship between variables.
Ministry of Justice and NatCen (2025). Review of the Presumption of Parental Involvement: judgment analysis, published 22 October 2025.
Norway's mapping of parental disputes drew 17,174 cases from the Lovisa court system covering 2008 to 2014. It records which party filed and the decision type. It does not record the substantive claim, the sex of the parties, or which of them prevailed.
Hovedkrav er ikke registrert i datasettet vi har mottatt [...] Vi kan derfor ikke undersøke om organiseringen og saksbehandlingen er ulik for foreldreansvar, daglig omsorg og samvær.
Oxford Research in cooperation with Proba samfunnsanalyse (2016). Domstolsbehandlingen av foreldretvister: kartlegging av foreldretvister etter barneloven, for Barne-, likestillings- og inkluderingsdepartementet.
Why that does not answer it
Sex is coded. Outcome is coded. Nobody has crossed them on a representative sample of cases that were actually adjudicated rather than settled, with a measure of who was doing the caregiving before proceedings began as the control. Without that control the comparison is worthless in either direction, which is why we will not run the analysis without it.
Who holds the data, and what the study would be
Where it is
- Data First, the Ministry of Justice and ADR UK linked administrative dataset, holds family court records from January 2011 to January 2021 linked to Cafcass records from April 2007 to January 2023 and to Census 2021, which supplies sex. It is held in the SAIL Databank and reachable under Digital Economy Act 2017 accredited researcher access.
- Domstolsverket in Sweden holds central registers of cases, parties and judgments. A 2017 parliamentary motion asked for annual custody outcome statistics to be compiled from them. It was not implemented.
What it needs
Accredited researcher access to the linked dataset, a pre-proceedings caregiving baseline drawn from the Cafcass records, and a pre-registered analysis plan published before the data are opened, so the result cannot be steered by whoever paid for it.
Status Specified. The data exists and the access route exists. Nobody has commissioned the study.
02
Whether attachment instruments measure fathers and mothers the same way
Do the Strange Situation Procedure and the observer-rated Attachment Q-Set show configural, metric and scalar invariance across parent, which is what licenses comparing mean security between mothers and fathers at all?
The field's largest meta-analysis of observed attachment compares exactly those means. Its author lists the untested assumption as his own limitation. Every mother-father comparison in the behavioural literature rests on it.
What already exists, and what its authors say it cannot do
The meta-analysis pooled 95 studies and 7,557 children on behavioural attachment measures and reports a mother-father correlation of r = 0.32 and a mean-level comparison. Its third stated limitation is that the studies assumed, without testing, that the same instrument measures security equivalently with mothers and fathers.
Measurement non-equivalence cannot be ruled out.
Pinquart M (2022). Attachment security to mothers and fathers: a meta-analysis on mean-level differences and correlations of behavioural measures. Infant and Child Development 31(6), e2364.
The largest validation of the Attachment Q-Set covered 139 samples and 13,835 children. Mothers described the relationship in 77 studies and fathers in 10. Trained observers rated attachment to the mother in 54 studies, and to the father in one.
The validity of the AQS for fathers still has to be documented. Incidentally, the same holds true for the SSP with fathers.
van IJzendoorn MH, Vereijken CMJL, Bakermans-Kranenburg MJ, Riksen-Walraven JM (2004). Assessing attachment security with the Attachment Q Sort. Child Development 75(4), 1188-1213.
Two studies have compared a self-report attachment scale across parent using a bootstrap overlap procedure on standardised loadings. Neither runs the configural, metric and scalar ladder, and one of them runs that ladder for child sex instead. A third paper reports no moderating effect of subject on Q-Set validity coefficients, which is a moderator test on validity rather than a test of the measurement model.
Our readingPrior art that narrows this claim rather than removing it. The question has been approached on questionnaires, and never on the observational instruments the mean comparisons actually use. That is our reading of these three papers, not a sentence any of them wrote.
Fernandes C et al. (2021), Attachment and Human Development 23(1), 56-74; Marci T et al. (2018), European Journal of Developmental Psychology 15(4), 481-492; Cadman T, Diamond PR, Fearon P (2018), Infant and Child Development 27, e2034.
Why that does not answer it
Searches run to establish the absence rather than assume it: PubMed for measurement invariance and attachment and father returns four records, none of them across parent on either observational instrument. OpenAlex returns nothing for invariance and Attachment Q-Sort and father, and nothing for measurement equivalence across mother and father on observational measures. Comparing group means without scalar invariance is not a small technical omission. It is the step that makes the comparison mean anything.
Who holds the data, and what the study would be
Where it is
- The Collaboration on Attachment Transmission Synthesis runs a secured remote-access data commons holding pooled individual data, syntax and analysis software from 59 samples and 4,498 dyads, coordinated at Vrije Universiteit Amsterdam. Its own overview paper does not mention fathers once.
- The Kochanska Family Study at the University of Iowa ran the Strange Situation with both parents on the same children, 101 with mother and 100 with father at 15 months, which is the within-child design the test requires.
- The Risky Situation arm at the Université de Montréal now has the procedure administered to mothers as well as fathers in at least three published samples, so cross-parent invariance is directly testable on data that already exists.
What it needs
A within-child design, the same child observed with both parents, using the Ainsworth interactive scale scores and the disorganisation scale as the items rather than the final classification, with correlated residuals for the same indicator across parent. The data has already been collected. What is missing is somebody funded to run the model on it.
Status Specified, and testable on archives that already exist. Nobody has commissioned it.
03
Which fathers, in which circumstances
Does the association between a father's sensitivity and his child's attachment security hold outside low-risk North American and European samples, and does it differ by risk, culture or family circumstance?
Not one of these questions is answered badly. They are unanswerable from the evidence that exists, and the authors of the largest synthesis say so in their own limitations section.
What already exists, and what its authors say it cannot do
The largest synthesis of parental sensitivity and attachment pooled 174 studies and 22,914 dyads. Fathers supplied 7 per cent of the data: 1,626 child-father dyads against 21,483 child-mother dyads.
Of the 23,109 dyads included in this meta-analysis, only 7% (n = 1,626) were child-father dyads whereas 93% were child-mother dyads (n = 21,483).
Madigan S, Deneault AA, Duschinsky R, Bakermans-Kranenburg MJ, Schuengel C, van IJzendoorn MH, Ly A, Fearon RMP, Eirich R, Verhage ML (2024). Psychological Bulletin 150(7), 839-872, Study Limitations.
The consequence is stated by the same authors: a pooled paternal effect size could be derived, but its moderators could not be tested, and the sub-analyses for avoidant, resistant and disorganised attachment in father dyads could not be run at all.
We were limited in our testing of moderators of this association and for sub-analyses for avoidant, resistant, and disorganized attachment.
Madigan S et al. (2024). Psychological Bulletin 150(7), 839-872, Study Limitations.
The samples that do exist are drawn overwhelmingly from populations with no clinical, medical or socio-demographic risk, and every single study of paternal sensitivity in the set was carried out in North America or Europe.
None of the studies were conducted outside of North America or Europe.
Madigan S et al. (2024). Psychological Bulletin 150(7), 839-872, Study Limitations.
Why that does not answer it
A body of evidence a twentieth the size of its counterpart, drawn from one part of the world and from families under no particular strain, cannot answer a question about fathers in general. It was never designed to. Nobody was ever asked to design it that way.
Who holds the data, and what the study would be
Where it is
- No single archive. This one needs new collection, in the places and populations the existing set excludes, using the instruments already validated for the purpose.
What it needs
Observational father-child samples outside North America and Europe, and samples with the clinical, medical and socio-economic risk profiles the current set screens out. This is the most expensive item on this page and the only one that cannot be answered from data already sitting in an archive.
Status Specified. Requires new fieldwork, which is precisely the kind of thing an institution with a budget exists to commission.
04
Sexual violence against boys, at the same coverage as girls
What is the prevalence of sexual violence in childhood against boys, measured on the same indicator, in the same countries, at the same coverage as it is measured for girls?
The indicator was written to cover both. The measuring was not done for both. This is the clearest example on the whole site of a gap that is nobody's decision to exclude anyone, and everybody's failure to count.
What already exists, and what its authors say it cannot do
Sustainable Development Goal indicator 16.2.3 covers young women and men alike. The comparable data actually collected covers about 68 low- and middle-income countries for women and about 12 for men, with high-income countries outside the count for both sexes.
The indicator excluded nobody. The measuring did.
UN Statistics Division, SDG indicator 16.2.3 metadata and country coverage.
Why that does not answer it
Every statement anyone makes about sexual violence against boys at global scale is drawn from a twelfth of the countries the equivalent statement about girls draws on. That is not a finding about boys. It is a finding about the survey instrument.
Who holds the data, and what the study would be
Where it is
- The Demographic and Health Surveys programme and the Violence Against Children and Youth Surveys already run the modules in-country. Extending the module to male respondents in the countries where it is already fielded for female respondents is an addition to existing fieldwork, not a new programme.
What it needs
Funding for the male module in surveys already going into the field, and a UN entity with a mandate to ask for it. There is currently no such entity, which is the whole of what this campaign is asking for.
Status Specified, and the cheapest item here, because the surveys are already happening.
·
What we will not say about any of this
The research is missing because somebody decided men should not be studied.
We have no evidence of that and we are not going to imply it. Every gap on this page has a duller and better documented explanation: instruments were validated on the parent who was available, indicators were written well and funded narrowly, and administrative data was collected for running a court rather than for answering a research question. Actions, not motives. What we are asking for is that somebody be assigned to close them, not that anyone be blamed for opening them.
Closing these gaps will prove the campaign right.
It might not, and we have to be able to say so before the money is spent rather than after. The point of a pre-registered analysis plan on the family court question is that the answer binds us too. If an adequately powered study with a caregiving baseline finds no difference by parent sex, that is the finding, and this page will carry it at the top with the same prominence as everything else. A research programme that can only produce one answer is not research.
Why this page exists
The route women's advocacy took was to get funding, understand more, and research more, and it worked. We are asking for the same route, and we are not asking for it at anyone's expense: nothing here takes a survey away from women or a study away from anyone. The male module on a survey already in the field is an addition to that survey.
The commitment that has to come with it is the uncomfortable one. A research programme that can only produce one answer is not research. If the family court study comes back showing no difference by parent sex once caregiving is controlled for, that is the finding, and it goes at the top of this page. That is why the analysis plan is pre-registered before anyone sees the data, including us.
every number we do publish, with its source → who we are and who pays for this →
