Processionary caterpillars of Thaumetopoea pityocampa (in Europe) and Ochrogaster lunifer (in Australia) (Lepidoptera: Notodontidae) form single files of larvae crawling head-to-tail when moving to feeding and pupation sites. We investigated if the processions are guided by polarisation vision. The heading orientation of processions could be manipulated with linear polarising filters held above the leading caterpillar. Exposure to changes in the angle of polarisation around the caterpillar resulted in orthogonal changes in heading angles. Anatomical analysis indicated specialisations for polarisation vision of stemma I in both species. Stemma I has a rhabdom with orthogonal and aligned microvilli, and an opaque and rugged surface; which are optimisations for skylight polarisation vision, similar to the dorsal rim of adult insects. Stemmata II-VI have a smooth and shiny surface and lobed rhabdoms with non-orthogonal and non-aligned microvilli; thus, optimised for general vision with minimal polarisation sensitivity. Behavioural and anatomical evidence reveal that polarised light cues are important for larval orientation and can be robustly detected with a simple visual system.

Materials and Methods

Behavioural analyses

In August 2018 and 2020, outdoor experiments with first instar (L1) T. pityocampa larvae were conducted on sunny days between 0930 – 1130 h (GMT +2) at Tregnago, Veneto, Italy (45°51’ N, 11°17’ E). A sheet of 50 cm² white paper was used as the experimental arena where 10 first instars at a time were released in the middle of the sheet and behavioural observations were made; N = 58 observations (31 single larvae (singletons) and 29 two or more larvae (processions)). After release, the larvae clustered for a few minutes, then formed processions or travelled as a singleton in various orientations. Four treatments were applied to the processions/singletons after the larva(e) established a course. A flexible 25 cm² linear polarising filter (PF) for visible light (XP42HE-40, ITOS, Mainz, Germany) was bent into a tunnel and held above the procession leader or singleton (Supplementary Fig. S1). The PF that filtered the light incident to both sides of the head was held either (1) ‘horizontally’ or (2) ‘vertically’ by rotating the filter 90°; so the horizontally or vertically polarised light was transmitted at low elevations, respectively, creating two orthogonal polarisation patterns around the larvae. After application, the larvae proceeded under (3) unobstructed sky without a filter. The PF created shade (transmission to unpolarised light 40%), and additionally decreased the incident light by max. 30%, depending on the angle with respect to the polarised skylight [21]. As the stemmata are non-image forming organs with large fields of view, the possible intensity artefact slightly affected the total, but not the differential signal in the orthogonal photoreceptor pairs. Thus, (4) control experiments were performed using a 45% neutral density filter (NDF) (Lee 298 and 209 combined; Lee Filters, Hampshire, UK), in place of PF. Each treatment lasted for 2 min, and the larval orientation was recorded 20 s after commencement (Fig. 1). The larvae were changed after every trial.

The same experiments were conducted on final 8th instar (L8) O. lunifer (N = 12 processions) and 5th instar (L5) T. pityocampa (N = 7 processions) larvae in their natural habitat without the arena. Ochrogaster lunifer were observed in late March 2019 at 0600 – 1200 h (GMT+10) at The University of Queensland, Gatton campus, Queensland, Australia (‑27°56’ S, 152°34’ E). Summer feeding T. pityocampa  [22] were observed in September 2019 at Leiria, Portugal (39°31’ N, 9°07’ W) at 0800 – 1300 h (GMT+1).

Morphological analyses

Preparation of stemmata for scanning electron (SEM), light (LM) and transmission electron (TEM) microscopy was performed as described previously [23]. Details can be found online [24].

Statistical analyses

All analyses were performed using R Studio version 1.1.419 and an alpha value of P < 0.05 was taken as statistically significant. In some trials, the PF was applied and removed repeatedly on the same procession/singleton. Before pooling the data for analyses, we tested if the orientation of processions was affected by previous exposure to the PF. The Wallraff Test of Angular Distances was performed using the RStudio software package ‘circular’ [25] on all processions/singletons. There were no significant differences (all P > 0.1), therefore, the data were used as independent for each treatment. To determine if the procession/singleton reacted to the treatments, the angular difference was calculated by subtracting the initial orientation of travel from the manipulated orientation after PF/NDF exposure/removal. Summarised angular differences of each group after the treatments were graphed as circular plots with a kernel density estimation (Fig. 1). The angular difference was given a score of 0 or 1 according to the difference being less (no change) or greater than 22.5° (change; 22.5° is the minimal cardinal value on the compass). Circular Logistics Regression Model (CLRM) for Binomial Data [26] was applied on the angular difference using the variables: Treatment, Azimuth angle (degrees), Procession/singleton ID, Number of larvae in the procession and Starting orientation. The models were reduced to the model of best fit by removing non-significant covariables and by lowering the Akaike’s information criterion value. Environmental temperatures of the study sites were collected, but not used as a variable because of collinearity (r > 0.70) with the Azimuth angle. Details of R codes used can be found online [24].

For further details on Morphological analyses and Statistical analyses, download the Dryad datasets.