The problem arises because many important plant families, from the mint family that includes rosemary to oaks, pines and countless daisies, package their seeds inside indehiscent fruits that do not open to reveal their contents. Researchers routinely treat these fruits as seeds, both for convenience and because it seems reasonable to assume that a plant investing resources in fruit tissues would fill them with a viable embryo. Yet seedless fruits are widespread across the plant kingdom, arising from self-incompatibility, male sterility, degraded pollen cells, chromosomal irregularities during meiosis and resource shortages. In most cases they are morphologically indistinguishable from seeded fruits, so there is no easy way to sort them out before an experiment begins.

Garcia-Fayos was drawn to the issue by a puzzling feature of the rosemary literature. Reviewing studies on the species’ germination, he found that in roughly 40 percent of published experiments, germination under controlled conditions never exceeded 10 percent, and reports of germination above 50 percent were rare. More striking still, the literature was riddled with contradictions. One study concluded that cold stratification boosts rosemary germination while another found no effect. Some forestry manuals recommend stratification while others dismiss it. Darkness reduced germination in some experiments but not others, gibberellic acid produced increases, decreases or nothing at all depending on the study, and even the effects of heat, smoke and soil chemistry were reported inconsistently. Madeiras and colleagues had already suggested in 2009 that these contradictions might stem from differences in the proportion of seeded fruits used across experiments, since rosemary produces many empty fruits after self-pollination, and that proportion varies between populations and years.

To test this idea rigorously, the researcher turned to experiments first conducted in 2004 with fruits collected from four large rosemary populations, each with more than 5,000 individuals, spread across the full gradient of annual temperature and rainfall that the species occupies in eastern Iberia. In each population, fruits were gathered from 15 adult plants separated by at least two metres, air-dried, shaken free from their calyxes and stored. The design crossed three alternating temperature regimes, 5 to 15, 10 to 20 and 15 to 25 degrees Celsius, with three soil water potential levels, 0.00, minus 0.01 and minus 0.05 megapascals, simulated using polyethylene glycol solutions calibrated to each thermal regime. A separate experiment crossed temperature with a light treatment comparing the natural alternating day-night cycle with continuous darkness, achieved by wrapping Petri dishes in aluminium foil.

The scale of the seedlessness problem became apparent only at the end of the experiments, when every ungerminated fruit was opened under a binocular microscope. Of 17,184 fruits collected across the four populations, just 17.2 percent contained a seed, defined as one that completely filled the fruit’s internal space with white or creamy-white, turgid tissue. Of the 8,955 fruits actually used in the germination trials, only 1,732, or 19.3 percent, held seeds. The consequences for the headline numbers were dramatic. Calculated the conventional way, over all fruits, germination was a meagre 11.2 percent. Counted only from seeded fruits, it leapt to 57.9 percent, a four- to seven-fold increase depending on the population. The ranking of the populations themselves changed, with La Murta leading at 78.7 percent germination of seeded fruits, followed by Pobla de Vallbona, Barranc dels Horts and Puebla de San Miguel.

The statistical consequences were even more consequential than the raw percentages. Using generalised linear mixed models fitted separately for each population, first with all fruits and then with seeded fruits only, Garcia-Fayos found that the apparent effects of environmental factors shifted in every population in the temperature and water potential experiment. When all fruits were included, water stress reduced germination in three of the four populations but temperature appeared to have no effect at all. When only seeded fruits were analysed, water potential emerged as a negative factor in all four populations, and warmer temperature regimes produced a clear positive response in three of them, a signal that had been completely masked by the empty fruits. In the temperature and light experiment, darkness appeared to suppress germination in three populations when all fruits were counted, but among seeded fruits this negative effect survived in only one, Puebla de San Miguel, while temperature and light otherwise lost their apparent influence entirely.

What makes the findings particularly sobering is that the remaining differences between populations could not be confidently attributed to local adaptation. If populations from hotter, drier sites had evolved seeds that germinate better at higher temperatures or tolerate drier soils, the seeded-fruit models should have shown effect sizes tracking the environmental gradient. They did not. The positive effect of temperature was indistinguishable between the populations from the warmest and coolest sites, and the effect of water potential was similar regardless of a population’s average rainfall. This fits with earlier population genetics work on the same species, which found high genetic diversity and weak population structure across the same gradient, driven by extensive gene flow through the shrub’s prolonged flowering and its generalist insect pollinators, conditions that hinder fine-scale local adaptation.

Instead, the residual inconsistencies most likely reflect a statistical artefact: the severe loss of replicates that occurs once empty fruits are removed. In the temperature and water potential experiment, only two populations retained 25 or more seeded fruits in every treatment combination, while Pobla de Vallbona fell below that threshold in eight of nine treatment levels. Given that 50 replicates per treatment level is often considered the minimum for statistically consistent results, nearly every treatment combination in both experiments was underpowered. The problem compounds if researchers, as is customary, treat Petri dishes rather than individual fruits as replicates. In this study, 30 of 180 dishes in the water potential experiment and 5 of 120 in the light experiment showed no germination at all, and 14 and 4 of those respectively contained no seeded fruits whatsoever, making the detection of any germination physically impossible.

The remedy Garcia-Fayos proposes is twofold. First, researchers should estimate the proportion of seeded fruits in their material before the experiment and increase the number of fruits sown per treatment level accordingly, so that the effective number of seeded replicates reaches a statistically meaningful target. Second, and indispensably, every ungerminated fruit should be opened and inspected at the end of the experiment so that calculations can be corrected to include only fruits that actually contained seeds. Neither step alone is sufficient; scaling up fruit numbers without verification leaves the denominator inflated, while verification without adequate starting numbers leaves the analysis starved of power. The approach echoes recommendations long voiced in the seed science literature, which has repeatedly warned that ignoring seed viability distorts germination ecology, yet the practice remains far from universal.

The implications stretch well beyond one aromatic shrub. Indehiscent, single-seeded fruits characterise a vast swath of the world’s flora, and the same blind spot affects not only germination experiments but also studies of seed rain, seed banks, predation and dispersal, where fruits rather than seeds are what researchers actually count. Ironically, seedlessness itself can be adaptive, since retaining aborted or inbred seeds until maturity may dilute the impact of pre-dispersal seed predators that lay eggs indiscriminately, a form of botanical decoy documented in junipers, pistachios and other groups. But whatever its evolutionary benefits, the new study makes clear that for the scientists trying to read the environmental rules of germination, empty fruits are noise masquerading as data. As research on endangered and rare species, where fruits and replicates are inherently scarce, becomes ever more urgent in a changing climate, the message is blunt: open the fruits, count the seeds, or risk publishing conclusions that a single overlooked emptiness could overturn.

Subject of Research: The effect of seedless fruits on germination experiments in plants with indehiscent single-seeded fruits

Article Title: Seedless fruits mask germination response to key environmental factors

Article References: Garcia-Fayos, P. (2026). Seedless fruits mask germination response to key environmental factors. Web Ecology, 26(1), 61-70. https://doi.org/10.5194/we-26-61-2026

Image Credits: AI Generated

DOI: 10.5194/we-26-61-2026

Keywords: seed germination, seedless fruits, Salvia rosmarinus, rosemary, plant ecology, experimental design, statistical power, water potential, temperature, light, Mediterranean shrubs, seed viability