1. Tracking devices have become small enough to be widely applied to arthropods to study their movement. However, possible side effects of these devices on arthropod performance and behaviour are rarely considered.
2. We performed a systematic review of 173 papers about research in which tracking devices –Radio Frequency Identification (RFID), harmonic radar, and radio telemetry tags– were attached to terrestrial arthropods. The impact of such tags was quantified in only 12% of the papers, while in 40% the potential impact was completely disregarded. Often-cited rules of thumb for determining appropriate tag weight had either no empirical basis or were misconstrued.
3. Several properties of a tracking device (e.g. weight, balance, size, drag) can affect different aspects of an arthropod’s life history (e.g. energy, movement, foraging, mating). The impact can differ among species and environments. Taken together, these tag effects can influence the reliability of obtained movement data and conclusions drawn from them. We argue that the impact of tracking devices on arthropods should be quantified for each (1) study species, (2) tag type, and (3) environmental context. As an example, we include a low-effort impact study of the effect of an RFID tag on a digger wasp.
4. Technological advancements enable studying the movement of arthropods in unprecedented detail. However, we should adopt a more critical attitude towards the use of tracking devices on terrestrial arthropods. The benefits of tracking devices should be balanced against their potential side effects on arthropods and on the reliability of the resulting data.

We systematically reviewed the published literature on the use of tracking devices on terrestrial arthropods. We searched ISI Web of Science (conducted on 26 August 2019) and scanned reference lists to include additional papers if accessible through the internet or via the authors. Only primary studies about terrestrial arthropods were selected. Non-arthropod terrestrial invertebrates, such as slugs and snails (Gastropoda), were not included as their ecology is highly different from that of terrestrial arthropods. Other kinds of tags or markings (e.g. numbered labels, coloured markings, or biotelemetry; the latter being tags to measure physiological and energetic variables remotely; Cooke et al., 2004), were also not considered. We selected 173 relevant papers, extracted information about the species, tracking devices, and the assessment of possible effects of these devices. Relevant variables collected for this study are species name, order and family, IUCN status, pest status, tag type, mode of movement, duration of tracking, mass of species and/or tag, the tag-to-body mass ratio, and whether the impact was discussed or quantified.

We tested the effect of RFID tags on the general behaviour of the digger wasp Bembix rostrata. In the field, we used a calliper to measure inter-tegular distance (ITD), maximum width of forewing (FW), and maximum length of forewing (FL). ITD is highly correlated with dry mass in bees and wasps (Cane, 1987; Ohl & Thiele, 2007) so we used it as a proxy for mass. The area of an ellipse with FW and FL as axes was used to approximate wing area. Wing loading was then calculated as ITD divided by wing area. RFID tags measured 8 mm x 2 mm and were encapsulated in glass vials to withstand the digging in the sand. The entire tag weighed 55 mg (Eccel Technology LTB, UK). The tags were glued on the dorsal thorax (scutum) with cyanoacrylate glue. Using a fully crossed (BACI) design (Smith, 2002), we tested the impact of the tags on the behaviour of the individuals in cubic insect cages with a side length of 30 cm (BugDorm, Taiwan). For 12 minutes the behaviour of the individual was noted every 30 seconds as flying, crawling, resting, washing, digging, or biting. After twelve minutes, half of the wasps were tagged and the other half was similarly manipulated but without attaching the tag to mimic the tagging procedure. After the treatment, the behaviour of the wasp was again recorded for 12 minutes. The test was performed on 94 individuals. Percentages of a certain behaviour were analysed using a generalized linear mixed model with a binomial distribution and logit link function. Wasp individual was included as a random effect to account for repeated measurements. Fixed effects were BA (before or after; 2 levels), CI (control or impact; 2 levels), and WL (wing loading; continuous) or ITD (continuous). Our main interest is the tag effect (BA:CI-interaction) and how this tag effect depends on WL or ITD (3-way interaction). We used the function ‘mixed’ from the R-package ‘afex’ (Singmann, Bolker, & Westfall, 2019), which uses likelihood ratio tests.

For the dataset on the effect of RFID tags, following abbreviations were used: f flying, c crawling, r resting, w washing, d digging, b biting. Measurements (ITD, WL, WW) were all taken three times and averaged.