Nowhere is the problem more insidious than in the eyes of ordinary observers. Because the three species are so similar externally, anglers, pond owners, and citizen scientists who submit photo-based records may simply be guessing wrong. A new study published in Nature Conservation put that concern to a direct test, asking hundreds of visitors at a major Czech angling fair to identify photographs of the three species. The results reveal a striking pattern: identification success depended far more on which individual fish was shown than on who was doing the identifying, and the errors were not random but flowed in directions that could actively distort conservation data.

The experiment took place at the For Fishing 2023 fair in Prague, an enormous gathering that drew more than 27,000 anglers to over 18,000 square metres of exhibition space. As part of outreach for the Save the Crucian Carp campaign, researchers set up a forced-choice identification quiz. Each participant received a printed sheet with nine photographs, three per species, all taken under controlled lighting, background, distance, and angle. Crucially, the fish selected were not idealised textbook specimens but realistic individuals spanning the genuine phenotypic variation encountered in the field, including deliberately confusion-prone types such as small-bodied gibel carp and a dark, naturalised goldfish. Participants had to label every photograph with one of three species names; no do-not-know option was offered, and outside help was prohibited.

After quality control, responses from 325 participants yielded 2,925 identification trials. Overall accuracy across answered trials was a moderate 64.3 percent, a figure that might seem encouraging until it is unpacked. Descriptively, goldfish fared best at 77.6 percent, crucian carp came in at 63.1 percent, and gibel carp languished at 52.4 percent, barely better than a coin flip. But the statistical modelling told a subtler story. When the researchers fitted binomial generalised linear mixed models with random intercepts for both respondents and photographs, the species-level differences dissolved into statistical insignificance, while the photograph itself emerged as the dominant source of variation.

That image-level effect was enormous. Adding photograph identity as a random effect dramatically improved model fit, and within each species the three images differed significantly in how easily they were recognised. The gibel carp photographs were the most extreme case: two small-bodied individuals were correctly identified only about a third of the time, while a large, robust adult was identified 89.5 percent of the time. Model-based predicted probabilities of a correct answer ranged from 0.33 to 0.92 depending purely on which photograph was shown. In other words, the salience of diagnostic traits in a specific individual fish, not the species it belonged to, largely determined whether an observer got it right.

The direction of the mistakes is what should concern conservationists most. For the two small gibel carp images, the most frequent wrong answer was crucian carp, selected in over 51 percent of answered trials for each. Meanwhile, one crucian carp photograph was labelled as gibel carp in 39.3 percent of responses, and the naturalised, bronze-coloured goldfish drew substantial confusion, with 31.6 percent of respondents calling it a crucian carp. Only the bright orange ornamental goldfish images were essentially unambiguous, scoring above 88 percent. The pattern fits classic categorisation research on typicality effects: people recognise prototypical examples easily but stumble badly when a species presents a phenotype that overlaps visually with another.

Why does this matter for the fate of a critically endangered fish? Because photo-based citizen science records feed directly into monitoring, early detection of invasives, and management decisions. Directional confusions translate into systematic false positives and false negatives rather than noise that averages out. When small gibel carp are routinely reported as crucian carp, the apparent range of the endangered native species is inflated while the spread of the invader is masked. Worse, well-intentioned anglers or pond managers might move, stock, or protect fish they believe are crucian carp, inadvertently aiding gibel carp dispersal. The reverse error, labelling a genuine crucian carp as an invasive, can discourage protection or even trigger lethal management of the wrong fish. Recent participatory monitoring of crucian carp in Central Europe underscored the stakes: field verification confirmed the species in only 35 percent of submitted tips.

Experience helped, but only modestly. Model selection based on the corrected Akaike Information Criterion supported a parsimonious model in which angling-related expertise and education contributed to explaining accuracy, with anglers showing a small but statistically reliable improvement over other visitors, an odds ratio of 1.56. Age and sex received little support. The key insight is that even among fair attendees, a self-selected crowd with strong interest in fish, performance remained constrained by image ambiguity. Expertise and training can reduce misidentification, but they cannot compensate for diagnostic cues that are simply invisible or weakly expressed in a particular photograph, a limitation also documented in professional fisheries monitoring contexts.

The underlying biology compounds the problem. Crucian carp body depth varies with ecological conditions, gibel carp shows substantial phenotypic plasticity, and goldfish spans an enormous spectrum from ornamental orange to dark wild-type forms after decades of domestication and feralisation in European waters. Moreover, several of the formally diagnostic characters used by taxonomists, including gill-raker counts, lateral-line scale counts, peritoneum colour, and the serration of fin rays, simply cannot be assessed from a standard lateral photograph. Some individuals may therefore remain genuinely unassignable from images alone, even for experienced observers.

The study’s authors argue that the solution lies in redesigning how photo-based records are collected and validated. Reporting tools should record uncertainty explicitly, for example through confidence ratings, rather than forcing a categorical choice that converts doubt into a false definitive claim. Training and outreach materials should deliberately include small-bodied and ambiguous individuals, not just prototypical specimens, so the public learns the true range of variation. Records matching known confusion-prone phenotypes should trigger verification workflows, with expert review and, where feasible, molecular confirmation for atypical individuals or suspected hybrids. Even the rare skipped answer, just 0.62 percent of trials here, could serve as a behavioural signal of ambiguity worth capturing in reporting interfaces. As citizen science takes on a growing role in freshwater fish conservation across Europe, the humble goldfish and its relatives have delivered a clear lesson: in the battle to save the crucian carp, knowing exactly what you are looking at is not a luxury but a conservation necessity.

Subject of Research: Public photo-based identification accuracy of crucian carp, gibel carp, and goldfish and its implications for citizen science monitoring

Article Title: Goldfish, gibel carp, or crucian carp? Testing public recognition of Carassius species

Article References: Gottwald, M., Stepanyshyna, Y., & Šmejkal, M. (2026). Goldfish, gibel carp, or crucian carp? Testing public recognition of Carassius species. Nature Conservation, 65, 135-152. https://doi.org/10.3897/natureconservation.65.185650

Image Credits: AI Generated

DOI: 10.3897/natureconservation.65.185650

Keywords: crucian carp, gibel carp, goldfish, Carassius, citizen science, misidentification, invasive species, freshwater conservation, phenotypic plasticity, species identification, Czech Republic, angling