African elephants (Loxodonta africana) use many sensory modes to gather information about their environment, including the detection of seismic, or ground-based, vibrations. Seismic information is known to include elephant-generated signals, but also potentially encompasses biotic cues that are commonly referred to as ‘noise’. To investigate seismic information transfer in elephants beyond communication, here we tested the hypothesis that wild elephants detect and discriminate between seismic vibrations that differ in their noise types, whether elephant- or human-generated. We played three types of seismic vibrations to elephants: seismic recordings of elephants (elephant-generated), white noise (human-generated), and a combined track (elephant & human generated). We found evidence of both detection of seismic noise and discrimination between the two treatments containing human-generated noise. In particular, we found evidence of retreat behaviour, where seismic tracks with human-generated noise caused elephants to move further away from the trial location. We conclude that seismic noise are cues that contain biologically-relevant information for elephants that they can associate with risk. This expands our understanding of how elephants use seismic information, with implications for elephant sensory ecology and conservation management.

Video data was collected during seismic playback experiments from a car separate from the car playing the seismic treatment (Figure 2a in main manuscript). Seismic playback was 6 minutes long, with 2 minutes of silence (a1 period), 2 minutes of seismic treatment (WN, Ele or Ele&WN, b period), then 2 minutes of silence (a2 period; Figure 1 in main manuscript). Videos were cropped into these 2 minute segments (a1, b or a2) and assigned a random number for blind analysis of elephant behaviour. These two minute videos have been uploaded (in either .ts or .mp4 format) along with a key for their treatment type, family/sub-family group and focal individual identity (see 'notes_on_videos.xlsx'). One video (33) uploaded was not used as elephant behavioural data in the main manuscript, but gives an example recording of car noise (Supplementary Figure 3).

Spreadsheet 'notes_on_videos' has three tabs. Treatment_key tab gives the number/title of the video and its corresponding family/sub-family group and the seismic treatment given. It also gives a key for the terms 'WN' 'Ele' 'Ele&WN' (seismic playback treatments) 'a1' 'b' and 'a2' (experimental periods). Focal_individuals tab gives details of which elephant in the video was the focal elephant, how visible they are in the video, notes on their identity, family/subfamily group and seismic treatment. Supp_Fig_3 tab gives more information on the videos that match the example recordings given in Supplementary Figure 3 - the elephant families, period, start and end times, distance from source and terrain related to these example recordings.