You do not need a relationship-science background to read a Guardian "Moment I Knew" column. The Carmen Pavlovic piece — Pete, who never wanted children, telling her out of the blue that he would have a baby with her — is moving on its own. But the longitudinal literature has names for what happened in that conversation. Nine terms is enough. They will not explain Pete's specific reversal. They will tell you what researchers measure when adults flip a stated preference on parenthood inside a serious partnership.
Parenthood Reversal Rate
The proportion of adults who change their stated preference on having children at least once during a longitudinal panel.
It matters because the folk model treats "wants kids" and "doesn't want kids" as stable traits. The panel data does not support that. Heaton, Jacobson and Holland (1999, *Journal of Marriage and the Family*) tracked fertility intentions across multiple survey waves and found that a non-trivial slice of respondents who described themselves as certain about not having children revised the position by the next interview. Hayford's 2009 *Demography* paper — using the National Survey of Families and Households trajectory data — reported broadly similar instability, particularly between ages 24 and 36.
Here is the working, plain. Take 100 self-identified childfree adults at baseline. Across an eight-year follow-up, roughly 30 to 40 will revise their stated intention at least once — the exact figure varies by panel and by how the question is phrased. Of those revisers, a smaller subset — call it ten to fifteen — will land on a final "yes" by the panel's last wave. The remainder oscillate. The Pete-style reversal sits inside that ten-to-fifteen subgroup, not in the supposedly fixed sixty-to-seventy.
Affective Forecasting Error
The systematic gap between predicted future feelings and actual future feelings.
Wilson and Gilbert (2005, *Current Directions in Psychological Science*) is the canonical synthesis. The core finding: people are reliably wrong about how an event — a promotion, a loss, a child — will make them feel, in both direction and magnitude. They over-predict the intensity of negative feelings (the "impact bias") and under-predict adaptation.
Why it matters here: "I never wanted children" is, mechanically, an affective forecast. It is a claim about how a future state would feel. The Pete reversal is not necessarily a change of values — it can be a corrected forecast. He has new data: this specific partner, this specific bond, this specific imagined child. The forecast updates. Gilbert's lab work (the 2002 affective adaptation papers) suggests the prior forecast was probably too confident anyway. The reader of the Guardian piece intuits this when they read "out of the blue" — but the literature would call it the lifting of impact bias under new information.
Self-Expansion Theory
Arthur and Elaine Aron's 1986 framework: humans are motivated to expand the self by incorporating the resources, perspectives, and identities of close others into their own self-concept.
The 1992 Aron, Aron, Tudor and Nelson paper in the *Journal of Personality and Social Psychology* operationalised this with the Inclusion of Other in the Self (IOS) scale — seven overlapping-circles diagrams, the participant picks the one that captures how much their identity has fused with the partner's. Higher IOS scores correlate, across studies, with greater willingness to absorb the partner's goals as one's own.
The Pete reversal fits the self-expansion mechanism cleanly. A goal Carmen held — motherhood — gets incorporated into Pete's self-concept rather than negotiated as an external demand. The shift is structural, not concessive. He does not give in to a baby. He becomes a person whose imagined self contains a baby with this particular partner. Whether the IOS-mediated route produces stable commitment is one of the open questions in the literature; Aron's lab generally finds yes, with caveats about the rate of expansion.
Mate Value Recalibration
The updating of a partner's perceived suitability for long-term investment as new information accumulates.
Evolutionary-psychology language can drift into pickup-artist territory; the academic version is bloodless. Buss and Shackelford's 1997 work on mate retention, and subsequent recalibration models (Conroy-Beam and colleagues at UC Santa Barbara, 2016 onwards), treat partner-value perception as a Bayesian estimate updated with each interaction. The estimate is not just sexual or genetic — it includes parenting-quality forecasts.
What this means in the Carmen-Pete shape: Pete's earlier "no children" position was made under an estimate of generic future partners. The estimate for Carmen specifically — five years of cohabitation, observed conflict resolution, observed kindness under pressure — is different. The recalibrated mate-value estimate makes a baby with Carmen a different decision than a baby with the median imagined partner. The literature would predict exactly this kind of partner-specific reversal in roughly the time frame the column describes.
Sliding Versus Deciding
Stanley, Rhoades and Markman's 2006 paper in *Family Relations* introduced the distinction. "Sliding" is moving into commitment milestones — cohabitation, joint mortgage, parenthood — by inertia and circumstance. "Deciding" is an explicit, verbalised, mutual choice.
Their argument, replicated across the Markman lab's later work, is that the deciding route predicts greater stability in the post-milestone phase. Slid couples are over-represented in subsequent divorce and dissolution data; deciding couples carry the milestone's weight more durably.
The Pavlovic vignette, as reported, is on the deciding side. "Out of the blue, he said he'd never wanted children but would have a baby with me" is a verbalised, explicit revision spoken in real time. It is not a fait accompli announced after circumstances narrowed the options. Stanley and Markman's framework would code this as a deciding moment, and would predict — population-level, not for any specific couple — better outcomes than the modal sliding-into-parenthood transition that their 2010 data showed dominates among US cohabiters.
Investment Model
Caryl Rusbult's 1980 model, formalised across the next two decades: commitment to a relationship is a function of three things — satisfaction with the relationship, quality of perceived alternatives, and accumulated investments that would be lost if the relationship ended.
The investment term is the underrated one. It includes time, shared friends, shared property, shared identity — and crucially, foregone alternatives. A partner who has spent eight years with someone has not only invested those years; they have not invested them somewhere else.
In the parenthood-reversal context, the investment-model reading is uncomfortable but worth naming. Pete's "I would have a baby with you" could be a commitment signal pushed by the investment term — eight years in, the cost of not having the child with this partner now looks higher than the cost of having one. The Rusbult framework does not moralise this. It is the same mechanism that explains why long-tenure couples often consolidate rather than recalibrate.
Cognitive Reappraisal
James Gross's 1998 paper in *Review of General Psychology* defined emotion regulation strategies. Cognitive reappraisal is the antecedent-focused strategy: changing how you construe a situation so that the emotional response itself changes.
This is not suppression. Suppression — Gross's other major strategy — masks an emotion that is already present. Reappraisal rebuilds the construal upstream so the emotion never arrives in the same form. Across 20+ years of replications, reappraisal correlates with better wellbeing, lower depression, and more stable relationships; suppression correlates with the opposite.
A partner who once flinched at "having a baby" and now leans toward it has, in Gross's terms, reappraised. The construal — what the baby means, what fatherhood costs, what the partnership becomes — has been rebuilt. The reader of the Guardian column may interpret this as a values change; the regulation literature would interpret it as construal change. The two are distinguishable: a values change predicts breadth (he would now have a baby with various hypothetical partners), a construal change predicts narrowness (only with this one).
Counterfactual Identity Construction
Markman and McMullen's 2003 *Personality and Social Psychology Review* paper on counterfactual reflection: humans build a sense of identity partly by imagining who they would be in alternative life paths, and using that imagined self to calibrate the current self.
Counterfactual upward comparisons ("I could have been a parent") generate regret. Counterfactual downward comparisons ("I could have ended up alone") generate satisfaction. The lab work shows both are used, often within the same person, to construct a stable identity narrative.
The Pete reversal is, partly, counterfactual identity work. Saying "I would have a baby with you" is a statement about the imagined alternative — the version of himself that did not. Naming the willingness shifts which counterfactual is operative. The Markman-McMullen framework would not predict that every man who runs this counterfactual flips on children. It would predict that the flip, when it happens, is preceded by a change in which alternative life feels worse than the current one.
Bayesian Belief Updating
The formal model of how a rational agent should revise a prior probability given new evidence. The posterior is proportional to the prior times the likelihood of the evidence under that prior.
Relationships research has imported this in soft form. The Conroy-Beam mate-preference papers, and the broader evolutionary-modelling literature, treat partner perception as a Bayesian estimate. So does the affective-forecasting work, in different language.
For the parenthood question: a prior of "I will not want children" gets multiplied by the likelihood of repeated daily evidence — Carmen's specific way of being, the texture of the relationship, the imagined child — and produces a posterior that can flip the modal answer. The mechanism is not mystical. It is not even uncommon. What is uncommon is the verbalised, audible, in-real-time articulation of the posterior, which is what makes the Guardian column moving and what makes it data. We would revise the framing above if a methodologically strong longitudinal study found that verbalised parenthood reversals like Pete's had the same dissolution rate as the unverbalised modal case. Until that paper exists, the deciding-versus-sliding distinction holds, and the reversal in the column reads, in the literature's terms, as the stronger of the two.
FAQ
What does the research say about how often a partner reverses their position on having children?
Across longitudinal panels of US and UK adults, somewhere between 30% and 40% of self-described childfree respondents revise that position at least once across an eight-to-ten-year follow-up. The figure varies with question phrasing and age band. Hayford's 2009 *Demography* trajectory analysis showed the most volatility in the late twenties and early thirties — the same window most "Moment I Knew" columns describe. Stable lifetime childfree adults exist; they are just a smaller fraction than self-reports at any single time point suggest.
Is Pete's kind of reversal — sudden, verbalised, partner-specific — predictive of follow-through?
The Stanley-Markman "sliding versus deciding" framework would say yes, relative to the comparison case. Verbalised, explicit decisions sit on the deciding side and correlate with greater post-milestone stability than slid-into transitions. The framework is population-level — it does not predict any single couple. It does predict that, between two couples who become parents, the one that named the decision aloud has, on average, the better odds.
Is "self-expansion" a softer way of saying he gave in?
No, and the distinction is the whole point of Aron's theory. Concession means accepting a partner's goal as an external demand you tolerate. Self-expansion means absorbing it into your own self-concept so that pursuing it feels like pursuing your own goal. The IOS scale was designed to discriminate the two. High-IOS partners report the absorbed-goal experience; low-IOS partners report the tolerated-demand experience. Outcomes diverge accordingly.
Could affective forecasting error work in the other direction — predicting parenthood will feel better than it does?
Yes, and the literature is clear on this. The impact bias runs in both directions. People over-predict the joy of positive future events as much as they over-predict the pain of negative ones. Gilbert's framing is that we are bad at predicting our own future emotional states generally, not biased toward pessimism specifically. A reversal driven by corrected forecasting is not automatically a correct forecast.
Where does mate value recalibration end and rationalisation begin?
The honest answer is that the published instruments cannot fully discriminate the two. Conroy-Beam's mate-preference recalibration work uses Bayesian-style updating models that fit the data well, but a rationalising mechanism would fit similar data. The cleaner test is prospective: does the recalibrated estimate predict behaviour the rationalisation account would not? Some of the 2018-onward partner-specific investment papers attempt this. The results are suggestive, not conclusive.
What would change the reading offered here?
A well-powered prospective study tracking verbalised parenthood reversals from the moment of articulation through ten-year outcomes — relationship stability, parenting quality measures, child wellbeing — and finding null or negative differences relative to non-reversers. That study does not yet exist at sufficient sample size. The Stanley-Markman work approximates it for the deciding/sliding distinction generally, not for parenthood reversals specifically. Until that gap is filled, the glossary above describes mechanisms, not outcomes.