Tracking the impact of sickness on social networks in the field

Out latest paper, entitled “Tracking sickness effects on social encounters via continuous proximity sensing in wild vampire bats” is out in Behavioral Ecology (once again just in time for Halloween). This field experiment led by postdoc Simon Ripperger is the followup to a series of lab studies on sickness behavior led by grad student Basti Stockmaier.

To predict how a pathogen will spread across a social network, it can be very useful to understand changes in the social behavior of individuals before and after they get sick. Our past lab studies showed that sick lethargic bats are much less likely to call, approach, or groom others (see example video below). Our new field experiment confirmed these social effects of sickness behavior in a natural setting and precisely measured them over time.

It’s important to clarify what our study does and does not show. Eusocial societies of ants, and (some) human societies perform active and altruistic forms of “social distancing” that involve sick individuals self-isolating as a public good to protect the whole colony. Vampire bats do not do this, and this does not make evolutionarily sense for them given their kinship structure. Despite trying our best to make this clear, some journalists and bloggers continue to tell the story this way. It’s an easy mistake to make because vampire bats are highly cooperative. But vampire bats do not live in stable cooperative groups (like say cooperatively breeding meerkats). Instead, they have networks of differentiated social relationships that exist despite roosting groups being fairly unstable in membership. They don’t invest in the whole colony.

What we did observe is a more passive and automatic form of social distancing which occurs simply as a byproduct of sickness behavior. This phenomenon occurs in other bats, mice, and possibly most or all mammals. Sick bats move around less, and they might even move away from others to avoid the jostling chaos of social interactions or to avoid overheating if they have a fever. But none of these behaviors are altruistically intended to help the group as whole.

We did this field experiment on a trip to Belize (actually during my birthday) with my postdoc Simon Ripperger, me, and my wife Michelle Nowak. Graduate student Basti Stockmaier created the experimental treatments and helped us with the experimental design. For study subjects, we captured over 100 vampire bats from a roost at the Lamanai Archeological Reserve, a site of Mayan Ruins (see photos below).

We then selected samples of non-pregnant females for the treatment and control group. To create “sick” bats, we injected 16 female bats with the immune-challenging substance, LPS, while the control group of 15 females received saline injections. Next, we used proximity sensors to continuously track the bats’ social encounters over a time period in which the sickness effect of LPS should disappear. As the data were being downloaded from the tree, we relaxed in our hammocks enjoying a cold beverage. We liked this new form of bat fieldwork.

Just as we had seen in captivity, the “sick” bats associated with fewer bats, spent less time near others, and were less socially connected to more well-connected individuals. The cool new thing is that we could see these effects across multiple bats simultaneously, and how the effect changed hour-by-hour, even minute by minute. Below is a plot of the difference in social connectedness between the control and test group every hour over three days.

One of the great joys of science is picturing a plot in your mind, and then making it and seeing how well the real data match your expectations. Making this plot was one of those moments (well, I mean the moment after the very first plot, which had a bug in the code so that nothing made any sense).

The highlighted periods between the vertical lines is when we expected to see the sickness effect (i.e. when the bats are awake inside the tree). During this period of the night, the effects of the sickness treatment were dramatic. When the bats left the roost to forage around 3 am, the effect should be partially reversed because more healthy bats left the roost to forage while more sick bats appeared to stay behind. When the bats returned to sleep around sunrise, there was again no clear difference because neither the treated and untreated bats were moving about. Also, we could see the effect disappear as the injected LPS treatment wore off.

The proximity sensors we used update social encounters every 2 seconds. These kinds of high-resolution proximity data would allow researchers to flexibly define association rates based on how a particular pathogen is transmitted. For example, one could define a “contact” as either being at least 10 seconds or at least 1 hour. Or one could define a contact as being within a 5 cm or a within 5 meters. We therefore also inspected how different ways of defining “contact” would changed the observed effect.

For example, if we said a contact requires more than 15 minutes of close proximity, then the number of encounters became too few to reliably detect the statistical effect of sickness on social behavior (at least within this number of subjects). The same thing was true if made the threshold for proximity too high. Because the data have estimates of both contact duration and proximity, it would be possible to simulate how pathogen transmission would be impacted by social behavior changes for a pathogen that requires 1 minute at 1 cm for transmission versus a pathogen that requires 10 minutes at 100 cm.

In collaboration with grad student Elsa Cardenas Canales, we are now looking at social behavior changes in vampire bats before and after they contract a real virus, rabies. We are also using the same technology to study social foraging and emergence of new cooperative relationships.