Flowers come in a variety of colours, shapes, sizes, and odours. Flowers also differ in the quality and quantity of nutritional rewards they provide to entice potential pollinators to visit. Given this diversity, generalist flower-visiting insects face the considerable challenge of deciding which flowers to feed on and which to ignore.

Working with real flowers poses logistical challenges due to correlations between flower traits, maintenance costs, and uncontrolled variables. Here we overcome this challenge by designing multi-modal artificial flowers that varied in visual, olfactory, and reward attributes. We used artificial flowers to investigate the impact of seven floral attributes (three visual cues, two olfactory cues, and two rewarding attributes) on flower visitation and species richness. We investigated how flower attributes influenced two phases of the decision-making process: the decision to land on a flower, and the decision to feed on a flower.

Artificial flowers attracted 890 individual insects representing 15 morphospecies spanning seven arthropod orders. Honeybees were the most common visitors accounting for 46% of visitors. Higher visitation rates were driven by the presence of nectar, the presence of linalool, flower shape, and flower colour, and was negatively impacted by the presence of citral. Species richness was driven by the presence of nectar, the presence of linalool, and flower colour. For hymenopterans, the probability of landing the artificial flowers was influenced by the presence of nectar or pollen, shape, and the presence of citral and/or linalool. The probability of feeding increased when flowers contained nectar. For dipterans, the probability of landing on artificial flowers increased when the flower was yellow and contained linalool. The probability of feeding increased when flowers contained pollen, nectar, and linalool.

Our results demonstrate the multi-attribute nature of flower preferences and highlight the usefulness of artificial flowers as tools for studying flower visitation in wild insects.

We developed and tested artificial flowers aimed at attracting wild, free-flying insects. Based on preliminary literature searches and our experience, we tested the effect of seven flower attributes: the presence of nectar, the presence of pollen, the presence of citral, the presence of linalool, flower colour, flower shape, and flower size. Each attribute had two levels, with the exception of ‘flower shape’ which had 4 levels. For nectar, pollen, citral, and linalool, the levels were simply ‘presence’ or ‘absence’. For shape, the levels were daisy, daisy clustered, simple disc, and simple clustered. For colour, the levels were ‘yellow’ or ‘blue.’ For ‘display size’, the levels were ‘small’ (5 cm across) and ‘large’ (10 cm across). Investigating every combination of our seven attributes in a full factorial design was logistically infeasible as it would have required 256 unique flower treatments. To overcome this challenge, we used a fractional factorial experimental design to reduce the number of treatments to a feasible subset of 16 unique flower designs 

We deployed patches of artificial flowers around The University of Sydney Camperdown campus in New South Wales, Australia (33.8886°S, 151.1873°E) on the unceded land of the Gadigal people of the Eora nation, between 10:00 and 11:00 am during Australian autumn (March–May) and early winter (June – August) 2021. We tested artificial flowers on days forecast to be sunny with no rain. Maximum temperatures on study days ranged from 18 to 31˚C. Experiments ran between 10:00 and 16:00 hours.

Insects were identified to morphospecies on the wing.