The study, published on 14 October 2026 as an open-access article, recruited a German-speaking cohort that included patients with psychotic disorders as well as healthy control individuals who were further characterized by low or high levels of schizotypy, a personality construct reflecting subclinical traits associated with schizophrenia-spectrum conditions. Each participant completed structured interviews, open discourse tasks, and picture description exercises in which they described at least six pictures. This multi-task design was deliberate. By recording the same speakers across different speaking conditions, the researchers could ask not only which features distinguish patients from controls, but also how much speech is actually needed to obtain a reliable classification.

From the audio recordings and their transcriptions, the team extracted a broad range of features that fall into two fundamentally different categories. Acoustic features capture the physical properties of the sound signal itself: pitch, intensity, timing, pauses, and spectral characteristics that reflect prosody and voice quality. Linguistic features, by contrast, are derived from the content of speech once it has been transcribed into text, encompassing vocabulary richness, sentence structure, word frequency, and measures of semantic coherence that quantify how tightly related the ideas expressed in a passage are to one another. These features then served as inputs to Random Forest classifiers, an ensemble machine learning method that builds many decision trees on random subsets of the data and aggregates their votes.

The results were strikingly asymmetric. Linguistic features achieved area under the receiver operating characteristic curve, or AUC, values of around 0.8 to 0.9 consistently across all interview formats. An AUC of 0.5 corresponds to chance-level guessing, while 1.0 represents perfect discrimination, so values in this range indicate clinically meaningful separation between groups. Acoustic features performed significantly worse, and their weakness was most pronounced in the picture description task, where the constrained, stimulus-driven nature of the speaking condition appears to strip away much of the prosodic and temporal variability that acoustic models rely on. In other words, when everyone is describing the same picture in a similar way, the sound of the voice alone offers comparatively little diagnostic leverage.

One of the most practically important findings concerns efficiency. The researchers showed that, using linguistic features, increasing the number of pictures described by each participant significantly improved model accuracy, but the gains did not continue indefinitely. Performance plateaued after the combination of three pictures. This matters enormously for clinical translation. A diagnostic or screening instrument that requires a participant to describe six or more pictures imposes a burden on both the patient and the clinician, lengthening assessment sessions and complicating data collection. Knowing that three pictures suffice to reach near-peak accuracy means that a future speech-based screening tool could be built around a task lasting only a few minutes, without sacrificing the classification performance that longer protocols deliver.

What exactly were the models learning? The team’s feature importance analysis provided a linguistically coherent picture of the differences between patients and controls. Patients with psychotic disorders produced simpler language, shorter sentences, and less speech overall compared to healthy controls, and they exhibited lower semantic coherence. These findings align with long-standing clinical observations of thought disorder in psychosis, in which the organized flow of ideas becomes fragmented and the connective tissue of discourse, the referential links that make speech easy to follow, weakens. What is new is that these phenomena can be quantified automatically from transcripts, without any human rater, and that the resulting numbers carry enough signal to drive a classifier to AUC values approaching 0.9.

The inclusion of schizotypy in the control group adds a layer of sophistication to the study design. Healthy individuals with high levels of schizotypy share some subtle cognitive and perceptual characteristics with patients, making them a more demanding comparison group than controls selected at random. By testing whether speech analysis could separate patients not just from ordinary controls but from controls with elevated psychosis-like traits, the researchers probed whether the linguistic signature they detected reflects something specific to the illness itself rather than to broad personality variation. The consistent performance of linguistic features across formats suggests that the content-based signal is robust to these gradations within the healthy range.

The choice of a German-speaking cohort is also significant. Much of the published work on automated speech analysis in psychiatry has been conducted in English, and it is not obvious that findings transfer cleanly across languages. Semantic coherence measures, in particular, depend on language models and embedding techniques trained on specific languages, and sentence-length statistics interact with the grammatical conventions of the language being spoken. Demonstrating that linguistic features reach AUC values of 0.8 to 0.9 in German provides evidence that the approach generalizes beyond English, which is a prerequisite for any effort to deploy speech-based screening across the diverse linguistic landscapes of Europe and beyond.

The research was supported by the European Union under grant agreement 101080251 as part of the TRUSTING project, and by the Swiss National Science Foundation, with additional support for individual investigators from the Brain & Behavior Research Foundation and the OPO Foundation. The TRUSTING project, which focuses on trustworthy AI for mental health monitoring, signals the direction in which this line of work is heading: not toward replacing clinicians, but toward building systems whose predictions are reliable, transparent, and validated across languages and clinical settings. The authors declared no competing interests, and the article is published under a Creative Commons Attribution 4.0 license, making the findings freely accessible to researchers worldwide.

Cautious optimism is the appropriate response. The study demonstrates that automated linguistic analysis of relatively brief speech samples can objectively detect the discourse changes associated with psychosis, and that a three-picture description task may be enough to capture the signal. At the same time, AUC values in the 0.8 to 0.9 range, while strong, are not diagnostic certainty, and any real-world deployment will require validation in independent cohorts, attention to false positives in screening contexts, and careful handling of privacy and consent when voice data are involved. What this study delivers is a clear methodological roadmap: prioritize what is said over how it is said, keep the task short, and measure semantic coherence. If those principles hold up in larger and more diverse samples, the quiet voice of a patient describing a picture could one day become one of the most informative data points in psychiatric assessment.

Subject of Research: Automated linguistic and acoustic speech analysis for detecting psychotic disorders in a German-speaking cohort

Article Title: Automated speech analysis for psychosis detection: a systematic comparison of task efficiency and feature groups

Article References: Zimmermann, N., Heitz, J., Steiner, A., Dannecker, N., Hüppi, R. M., Cecere, G., Omlor, W., Surbeck, W., Homan, P., & Langer, N. (2026). Automated speech analysis for psychosis detection: a systematic comparison of task efficiency and feature groups. Schizophrenia. https://doi.org/10.1038/s41537-026-00816-0

Image Credits: AI Generated

DOI: 10.1038/s41537-026-00816-0

Keywords: psychosis, schizophrenia, automated speech analysis, machine learning, linguistic features, acoustic features, semantic coherence, schizotypy, Random Forest, digital biomarkers, University of Zurich, picture description task

The research team, led by Marco Giesselmann, analyzed data from the Swiss Household Panel, a longitudinal social science survey that follows the same households and individuals across many years. This design matters enormously for questions about partnership formation, because cross-sectional snapshots can easily confuse cohort effects with genuine behavioral change. By tracking partnership status, educational pairing, and income relationships across a full generation of Swiss adults, the researchers could observe how mating market dynamics actually evolved as female educational attainment tripled. Their central question was deceptively simple: when women become more highly educated on average while still seeking partners with the same or a higher level of education, does it become harder for those women to find anyone at all?

The empirical answer is emphatic. The study found no evidence of a marriage squeeze for highly educated women in Switzerland. Across the entire 25-year observation window, the likelihood that a highly educated woman aged 30 to 40 lived in a partnership remained consistently above 80 percent, and in fact showed a slight upward trend. This stability defied the researchers’ own expectations. This result came as a surprise to us, Giesselmann acknowledged, noting that the team had anticipated a different trajectory given the dramatic shift in the educational composition of the population. The persistence of high partnership rates among graduate women suggests that the classic hypergamy model, in which women marry up and men marry across or down, has proven far more flexible than many theorists assumed.

The key to this resilience, according to the study, lies not in the behavior of women but in the adaptive responses of men. Women with tertiary education continue to prefer relationships with men who hold the same level of education, a preference that has not eroded over time. What has changed is the supply side of the equation. Men have been the ones primarily adapting to the changing educational composition, the researchers conclude. In 1999, only about one-quarter of highly educated men were partnered with a fellow higher education graduate. By 2024, educationally equal relationships had become the dominant partnership pattern for this group. The normative boundary that once made it unusual for a graduate man to pair with a graduate woman has effectively dissolved, expanding the pool of acceptable matches for educated women without requiring them to lower their standards.

The adaptation extends well beyond the highly educated male population. Men with medium levels of education also became increasingly likely to form relationships with highly educated partners, doing so nearly four times as often in 2024 as they had in 1999. This diffusion of cross-educational pairing across the middle of the distribution has kept the mating market of highly educated women in balance, absorbing the growing female surplus at the top of the educational hierarchy. The picture is considerably less rosy at the bottom. The likelihood that men with low levels of education live without a partner has nearly tripled since 1999, a stark indicator of how the costs of educational expansion are being distributed unevenly across the social structure. Yet even here, the researchers add an important demographic caveat: the share of men with low educational attainment in the total population has been steadily declining for years, which partially offsets the individual-level partnership penalty this group faces.

At first glance, these findings might suggest that the traditional European partnership model, in which the man holds more education than his female partner, is simply fading away. The data tell a more subtle story. Educational hypergamy, understood strictly in terms of degrees and credentials, has indeed declined as equal-education couples have become the norm among graduates. But when the researchers turned from education to annual income, they discovered that the traditional architecture of the couple remained remarkably intact. Inequality within Swiss partnerships has not disappeared; it has migrated from one dimension of socioeconomic status to another. Credentials have converged, but paychecks have not.

The income findings are among the most consequential in the study. Only 25 percent of men in the top income third have a partner who earns roughly the same amount as they do, even though a majority of these high-earning men are now in relationships with highly educated women. In other words, even when both partners hold advanced degrees, the man typically remains the substantially higher earner. Gender-based income disparities continue to reinforce traditional economically unequal partnership patterns, Giesselmann explains. Traditional differences in socioeconomic status have not simply vanished; they have, to some extent, merely shifted from the educational column of the household ledger to the income column. This pattern, which the researchers describe in terms of masked hypergamy, implies that the economic foundations of the male-breadwinner model persist beneath a surface of educational equality.

Why does this masking occur? The study itself does not run a full causal decomposition, but the pattern is consistent with well-documented mechanisms in labor market and family research. Gender gaps in labor force participation, working hours, and career interruptions around childbirth continue to depress women’s earnings relative to those of equally educated male partners. When a couple has a highly educated woman and a highly educated man, the woman is still more likely to reduce paid work or accept lower-paying positions to absorb domestic and caregiving responsibilities. The result is that educational homogamy, the pairing of like with like, coexists with persistent economic asymmetry within the household. For policy purposes, this distinction is critical: expanding women’s access to higher education, however valuable in its own right, has not by itself dissolved the economic hierarchies inside Swiss couples.

The findings also carry weight well beyond the marriage market, speaking directly to one of Switzerland’s most pressing public debates: the decline in birth rates. Many commentators have assumed that falling fertility reflects a matching crisis, with educated women unable to find suitable partners and therefore delaying or forgoing family formation. The Zurich study undercuts that assumption. Given the relative stability of partnership formation documented across 25 years, the researchers argue that the drop in birth rates has less to do with the marriage market than with institutional and cultural factors. Among the candidates they identify are changing expectations around self-fulfillment, the persistent challenge of balancing work and family life, the fact that starting a family is no longer treated as a matter of course, and rising housing costs that raise the economic threshold for parenthood. If partnership formation is not the bottleneck, then policies aimed at easing the transition to parenthood, such as childcare provision, housing support, and family-friendly workplace arrangements, may be more effective levers than any intervention targeting the dating market.

Methodologically, the study demonstrates the value of longitudinal panel data for resolving debates that cross-sectional statistics cannot settle. Because the Swiss Household Panel tracks individuals over time, the researchers could distinguish genuine changes in partnering behavior from artifacts of generational replacement. The work, authored by Giesselmann together with K. Rost, P. Sager, M. Döhne, J. Rössel, and C. Vantis and published in volume 52, issue 3 of the Swiss Journal of Sociology, contributes to a broader international literature on educational assortative mating, gender inequality, and family demography. Its central message resonates far beyond the Swiss case: as women outpace men in higher education across the industrialized world, the feared collapse of the partnership market for educated women has largely failed to materialize, because men and norms have adapted. What has not adapted, at anything like the same pace, is the economic order within couples. The gender revolution in the classroom, it turns out, is still waiting for its counterpart in the paycheck, and the persistence of that gap may shape family life in Switzerland and beyond for decades to come.

Subject of Research: Marriage market dynamics and partnership patterns among highly educated women in Switzerland

Article Title: Highly educated women continue to find partners

Article References: Highly educated women continue to find partners. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: Switzerland, marriage market, higher education, hypergamy, gender inequality, partnership formation, income disparity, Swiss Household Panel, demography, fertility decline, sociology, University of Zurich

Researchers at the Institute for Regenerative Medicine at the University of Zurich, working together with colleagues at the University of California, Santa Cruz, have now identified a way to protect these fragile cells during laboratory processing. Their study, published in EMBO Molecular Medicine, focuses on a signaling molecule with an unexpected dual identity. Activated protein C, a protease best known for its role in regulating blood coagulation and dampening inflammation, turns out to act on human blood stem cells in a way that preserves their stemness rather than depleting it. The finding points to a simple, time-efficient intervention that could meaningfully improve the outcomes of stem cell transplantation and gene therapy.

The challenge the team set out to address is rooted in the biology of the cells themselves. Hematopoietic stem cells exist mostly in a state of deep rest, a condition biologists call quiescence. In this state, the cells divide rarely and resist the temptation to differentiate into specialized blood cells. Quiescence is not idleness; it is a protective program that shields the cells from the wear and tear of continuous division and preserves their long-term capacity to rebuild the blood system. When stem cells are harvested from a patient or donor and then manipulated in the laboratory for gene therapy, they are exposed to culture conditions, viral vectors, and inflammatory cues that can push them out of quiescence. Once activated, they begin dividing and differentiating, and with each step away from the resting state they lose some of the regenerative potential that the transplant depends on.

To find a remedy, the Zurich team investigated how signaling through a specific cell surface receptor influences human blood stem cells. The receptor in question, protease-activated receptor 1, or PAR1, is a well-studied sensor of proteolytic activity in the blood. What makes PAR1 scientifically intriguing is that it can be switched on by different proteases with dramatically different consequences. Thrombin, the central enzyme of the coagulation cascade, activates PAR1 and is classically associated with pro-inflammatory and differentiation-promoting signals. Activated protein C also engages PAR1, but it is known from vascular biology to trigger a distinctly protective, cytoprotective program in endothelial cells. The researchers asked whether this same dichotomy would hold true in the context of hematopoietic stem and progenitor cells.

The answer, according to first author Simon Pöllmann of the Institute for Regenerative Medicine, was a clear demonstration that the identity of the activating protease matters enormously. Although both molecules activate the same receptor, PAR1, they produce very different outcomes, he explained. Treatment with activated protein C helped the blood stem cells remain in a resting, protected state. In practical terms, the aPC-treated cells divided less frequently, were less likely to differentiate prematurely into specialized blood cells, and retained the defining characteristics that stem cells need in order to function over the long term. Thrombin, by contrast, drove the cells toward greater differentiation, a shift that diminishes their stem cell potential and, by extension, the durability of any transplant built from them.

This receptor-level insight reframes how scientists think about manipulating stem cells ex vivo. PAR1 is not simply an on-off switch; it is a molecular platform whose output depends on which protease is bound to it. Activated protein C appears to bias the signaling machinery toward survival and quiescence, while thrombin biases it toward activation and lineage commitment. For laboratory protocols that expose stem cells to stress, the implication is that the protease environment surrounding the cells can be tuned to favor preservation over depletion. That kind of tuning could be integrated into existing gene therapy workflows, where cells must spend time outside the body while being corrected and expanded.

The most striking practical result of the study was how little exposure was needed to make a difference. Treating the human blood stem cells with activated protein C for just one hour was sufficient to improve their transplantability. The team tested this in an animal model by transplanting the treated cells into mice. The aPC-treated cells produced more human blood cells in the recipient animals and showed a better capacity to sustain blood formation even after repeated transplantation, a demanding test that probes the long-term self-renewal reserve of the graft. Sustained engraftment across serial transplants is one of the most rigorous indicators that a stem cell population has retained genuine, durable stem function rather than a short-lived burst of activity.

Beyond improving engraftment, the treatment also shielded the cells from a common hazard of laboratory handling: inflammatory signaling. Inflammation is a well-known antagonist of stem cell quiescence. Inflammatory signals that circulate during illness or that arise during cell processing can rouse stem cells from their resting state, driving them into cycle and pushing them toward differentiation. The study showed that activated protein C protected the stem cells from being activated by these inflammatory cues, effectively buffering the cells against the environmental pressures that would normally erode their potential. This protective effect is particularly relevant for gene therapy, where cells from patients with inflammatory blood disorders may already be primed toward activation before they ever reach the laboratory.

Last author Ute Modlich, professor in the Gene and Cell Therapy Division at the Institute for Regenerative Medicine, emphasized the translational significance of the engraftment data. The treated cells produced more human blood cells and displayed a better capacity to sustain blood formation even after repeated transplantation, she explained, underscoring that the benefit was not merely a laboratory artifact but a functional improvement in the property that matters most to patients: the ability of the graft to reconstitute a working blood system and keep it running. For conditions that depend on a small number of successfully engrafted, gene-corrected stem cells, even a modest improvement in preservation could translate into a substantially higher chance of therapeutic success.

The study, conducted as a randomized controlled research effort with animal models as the primary experimental subject, was a collaboration between the University of Zurich and the University of California, Santa Cruz, with contributions from Marcel Rommel and Camilla Forsberg on the American side. Taken together, the results establish signaling through activated protein C as a promising strategy for preserving the quality and regenerative capacity of human blood stem cells during the vulnerable window between harvest and transplantation. The researchers believe that, in the long term, this approach could improve the success of stem cell transplantations and gene therapies for a range of blood disorders, delivering a real benefit to patients whose treatments currently fall short of their full potential. As gene therapy continues to expand across hematological medicine, protecting the stem cells at the heart of these procedures may prove to be one of the most consequential advances of the coming years, and a one-hour treatment with a molecule the body already knows how to make is an appealingly simple place to start.

Subject of Research: The role of activated protein C signaling through PAR1 in preserving hematopoietic stem cell quiescence and improving transplantation

Article Title: Alternative signaling pathway improves transplantation of blood stem cells

Article References: Alternative signaling pathway improves transplantation of blood stem cells. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: blood stem cells, activated protein C, PAR1, thrombin, quiescence, gene therapy, bone marrow transplantation, hematopoiesis, engraftment, University of Zurich, EMBO Molecular Medicine, regenerative medicine

The study, published in The Quarterly Journal of Economics, used an elegant experimental design to probe moral intuitions about distributive justice. Participants were asked to decide how bonuses should be allocated to real workers on an online labor market platform, choosing between allocation based on measured performance and allocation through a random lottery. This approach allowed the researchers to observe whether people reward merit or tolerate chance when real money is at stake for actual workers, rather than merely recording abstract opinions. The same participants were then asked about their support for redistribution through policy measures, including higher taxes on the wealthy, giving the team a direct window into the connection between private fairness judgments and public economic preferences.

The headline result is that, on average, the global population is substantially more willing to accept an unequal distribution of wealth when it is based on work rather than on pure chance. Earnings that flow from effort, skill and achievement are widely regarded as deserved, while windfalls produced by the roll of a dice are viewed with far more suspicion and are far more likely to trigger support for redistribution. In other words, inequality is not a single moral category. Two societies with identical levels of income dispersion may judge that dispersion in profoundly different ways depending on whether the rich earned their position or simply got lucky.

Yet this merit-based view of fairness is not universal, and here the study delivers its most provocative findings. The belief that prosperity achieved without personal effort is undeserved is strongest in countries with democratic political systems that are richer, more equal and more individualistic. Switzerland emerges as a paradigmatic case: the Swiss population places exceptional value on achievement through personal effort and is comparatively tolerant of income gaps produced by hard work and performance. At the same time, the Swiss remain open to a degree of redistribution, particularly in cases where inequality is clearly grounded in luck rather than merit. The moral logic seems to be that society should cushion the unlucky but need not penalize the successful.

In sharp contrast, the researchers documented a high level of acceptance of luck in several non-Western countries, with China and India standing out as prominent examples. In these societies, majorities were not only more accepting of prosperity gained by chance; they were also opposed to redistributing wealth that had been acquired through luck. This finding upends the assumption, common in Western economics and political philosophy, that chance-based inequality is morally illegitimate everywhere and that luck should be neutralized through taxation and transfer programs. Almås suggests that cultural and other factors underpin these divergent attitudes, and she indicates that future work will examine to what extent differences in redistributive preferences are associated with different religious beliefs, which may shape whether fortune is seen as a private blessing or a collective concern.

The study also tested an alternative explanation for why people might tolerate inequality even when they find it unfair. Redistribution, after all, is not costless. High earners may be believed to work less when redistributive taxes are in place, shrinking the economic pie and reducing incentives. If such efficiency concerns drive tolerance of inequality, then attitudes should correlate with beliefs about the economic costs of taxation rather than with moral judgments about the source of income. The analysis, however, found that this factor plays a much smaller role in the global acceptance of inequality than the origin of that inequality itself. Luck versus merit, it turns out, is the decisive axis along which people around the world evaluate whether a distribution deserves their approval.

These results carry important methodological lessons for the social sciences. Much of what economists and psychologists claim about human fairness norms has been built on studies of so-called WEIRD populations, that is, Western, educated, industrialized, rich and democratic societies. The findings reveal a significant contrast between Western and non-Western societies, demonstrating that conclusions drawn from Western samples cannot simply be generalized to humanity at large. A person in Zurich and a person in Delhi may look at the same distribution of income and reach opposite moral verdicts, not because one is mistaken about the facts, but because they are drawing on fundamentally different frameworks for justifying economic outcomes.

The practical implications extend well beyond the academy. Debates over progressive taxation, welfare programs and executive compensation are often framed as battles between egalitarians and defenders of the free market, with the underlying assumption that people respond to inequality itself. This research suggests that framing is incomplete. Public support for redistribution is likely to depend on the narrative attached to inequality, whether citizens believe the wealthy earned their position through effort or stumbled into it through inheritance, connections or chance. Policymakers who ignore the moral source of economic disparities may misread public opinion, and campaigns for or against redistribution may succeed or fail based on which account of merit dominates the public conversation.

For Almås and her colleagues, the dataset represents a foundation for a new generation of research on the moral justification of inequality. By covering eighty percent of the world’s population with a consistent experimental protocol, the study offers an unprecedented map of how fairness norms vary across cultures, political systems and levels of economic development. Understanding how societies morally justify inequality, and how those justifications in turn shape attitudes toward redistribution, is essential for anyone seeking to explain why some democracies embrace high taxes and generous welfare states while others resist them. The study does not settle whether luck or merit should matter, but it makes unmistakably clear that, in the court of global public opinion, not all inequality is judged equally, and the verdict differs dramatically depending on where you ask the question.

Subject of Research: Cross-national study of fairness preferences and attitudes toward income inequality and redistribution

Article Title: Luck or merit? Not all inequality is equally acceptable.

Article References: Luck or merit? Not all inequality is equally acceptable.. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: inequality, fairness, redistribution, merit, luck, economic preferences, University of Zurich, cross-cultural study, taxation, behavioral economics, Quarterly Journal of Economics, Ingvild Almås