AbstractIn this paper we report the results of our first efforts to evaluate the functional significance to signaler and perceiver of variation in tail flagging (Fig. 1 and Fig. 2A-C) by California ground squirrels (Spermophilus beecheyi). We first report a series of anecdotes in which we describe the circumstances of a variety of tail movements by California ground squirrels, including the different kinds of tail flagging. Secondly and primarily we identify the information afforded by snake-elicited tail flagging. Tail flagging is a signal used by California ground squirrels primarily when they are harassing a potential snake predator (Fig. 4). It attracts other squirrels who may also begin harassing the snake. The risk to squirrels in encounters with snakes continuously varies, and the squirrels adjust their behavior accordingly. Consequently in this situation we expected to find shifts in the information afforded by different tail-flagging variants. We view the information afforded by tail movement and other signals as a consequence, not of selection for making that information available, but of the correlations resulting from situational constraints on the signaler's behavior, e.g., correlations between tail movement variation and variation in significant events. We used two complementary approaches to help determine the information afforded by tail flagging. In one, we asked whether information important to percipients is afforded by tail flagging. In the second, we searched for situational correlates of tail-flagging variants. We applied the second approach to each individual separately and to the group comprising these individuals. This allowed us to look for idiosyncrasies in signaler behavior. Although a relatively simple signal, tail flagging varies along several structural and temporal parameters. From video recordings we quantified a structural parameter- number of movement cycles in a bout of tail flagging - and two temporal parameters- rate and temporal clustering of tail flagging. Our results show that squirrels adjust their tail-flagging behavior in the following ways. 1. When a rattlesnake rattles, harassing squirrels increase the number of cycles per bout of tail flagging. 2. Lone snake-directed squirrels temporally clump their flagging bouts more than snake-directed squirrels accompanied by other snake-directed individuals. 3. The structure of flagging varied with the squirrel's behavior vis-a-vis the snake. While dealing directly with a snake, squirrels emitted relatively few 1-cycle bouts of flagging. While in the vicinity and monitoring snake-related events, but engaged primarily in other activities such as feeding or grooming, squirrels emitted relatively greater numbers of 1-cycle bouts. As one would predict from this difference, individuals were farther from the snake on the average while emitting 1-cycle bouts than while emitting 2-cycle bouts. Adults used far more 2-cycle bouts than 1, and 3 or higher. In the field, bouts of 4-cycles or greater were very rare. We found that the information afforded by a bout of tail flagging was much greater when we considered structural variation, than when we did not. For example, flaggers were more likely on the average to "pause" before than after tail flagging. However, although the same difference held for 1-cycle bouts, just the reverse was true for 3-cycle bouts, and there was little difference in the probability of pausing for 2-cycle bouts. We found similar differences for other behavioral correlates of tail flagging. A percipient could much more precisely predict a tail flagger's behavior by considering signal variation. Our results indicate a percipient can infer from a high proportion of 3-cycle bouts that the flagger is beginning an episode of snake-directed activity, whereas 1-cycle bouts indicate a temporary cessation of snake-directed behavior. We expect to find even greater situational specificity of tail flagging when we simultaneously consider multiple structural dimensions, such as axis of movement and number of cycles. When we looked at the information afforded by a bout of tail flagging separately for individual squirrels, it was clear that the specific behavioral profiles associated with different variants of tail flagging were somewhat idiosyncratic. Thus, a percipient squirrel should be able to infer more from the tail flags of a familiar squirrel than an unfamiliar squirrel. We propose that signals are used to elicit a particular kind of performance from the target(s). The effectiveness of such action depends upon a knowledge of the current status of the individual's target(s). We conclude that variation in tail flagging is constrained in at least three ways: 1) by the number and quality of targets of tail flagging (e.g., snakes, squirrels); 2) by the signaler's certainty about the status of the target(s) (by eliciting behavior in targets, tail flagging may be used in part to extract information about the target's state); and 3) by the quality and availability of feedback (the success of tail flagging is continually assessed on the basis of feedback). The understanding of the functional significance of tail flag variability therefore becomes the problem of understanding how the flagger uses signal adjustments to deal with changes in its circumstances in terms of these three general constraints.
