Three common marmoset family groups, each consisting of eight individuals, were observed at the Biomedical Primate Research Centre in Rijswijk, the Netherlands. Marmoset breeding pairs typically produce twin offspring every 5 months [Sutcliffe & Poole, 1984]. The groups consisted of one breeding pair and three generations of its twin offspring. Sixteen adults (>14 months) or sub-adults (>9–14 months) [cf., Cawthon-Lang, 2005a] from three different family groups were observed for this study (Table II). We excluded infants and juveniles, since their behavior may differ from that of adults and sub-adults [as found for vocalizations, see Martins Bezerra & Souto, 2008; Martins Bezerra et al., 2009]. Dominance within a group is thought to be age-dependent, with older individuals occupying a higher dominance position than younger (immature) ones [Sutcliffe & Poole, 1984].
Table II. Groups and Subjects Observed
|Name||Sex||Birth datea||Age category|
Animals were housed in an inside enclosure of 2 × 2 × 3 m connected to an outside enclosure of 2 × 2 × 3 m. Standard enrichment, such as branches, boxes, baskets, garden ropes, fire hoses, and nets, were always present in the cages. In addition, a different additional play object was presented every week. The cages were equipped with bio floors containing crickets and other insects that were available for the marmosets [Vernes & Louwerse, 2010]. The marmosets were fed monkey chow (Sniff®) in the morning and this was replenished if necessary at the end of the afternoon. As a complement Arabic gum, fruit, raisins, biscuits, or puffed wheat with honey were given once a day. Water was available ad libitum.
The study only concerned behavioral observations that did not affect the normal behavior of the animals. This research complied with protocols approved by the Animal Ethical Committee of the Biomedical Primate Research Centre and adhered with the legal requirements of the Netherlands and the American Society of Primatologists principles for the ethical treatment of primates.
Pre–post event observations concern the significant association in time between a visual display (the event at point zero) and a particular behavioral context by comparing behavioral rates at specific time points with a baseline. Therefore, interactions between individuals were followed in detail, including who gives and who receives the behavior, and these were collected with focal observations.
Focal observations [Box, 1975; Martin & Bateson, 1993] were collected during a 4-month period (December 2010–March 2011). Observations were carried out between 09.30–12.00 and 13.00–15.30 hr. The timing of the observations was balanced over the day, to exclude effects of time of the day. Subjects were observed on average 293 min (mean: 293.13 min, standard deviation: 4.43 min). The focal observations of December and January were 10 min long, those in February and March were 20 min (see below for an explanation). Two weeks before the observation period started, the marmosets were habituated to the observer (RAdB) sitting in front of their cage.
The ethogram used in this study was based on published descriptions of common marmoset behavior [Barros et al., 2007; Cilia & Piper, 1997; Digby, 1995; Epple, 1968; Gerber & Schnell, 2004; Kaplan & Rogers, 1999, 2006; Moynihan, 1967, 1970; Sparks, 1967; Stevenson & Poole, 1976; Sutcliffe & Poole, 1984; see also www.marmosetcare.com]. Two types of behavior were distinguished. First, we identified behavior that can be unequivocally related to one behavioral context, distinguishing four contexts: aggressive, fear, affiliation, and play behavior. These behaviors were either actions or vocalizations that in the literature were assigned to a specific behavioral context. We did not observe sexual behavior in the observation period and, therefore, could not distinguish a sexual context. Second, we identified “visual displays,” entailing a facial expression, body posture, and/or the (partial) pilo-rection of the fur (Table I; Fig. 1; Supplementary Video).
Figure 1. Photographs of several common marmoset visual displays: tongue in-out; arched bristle; slit stare; head tilt; pilo-erection tip tail; hind legs rampant; tufts up; and tufts down.
Download figure to PowerPoint
The behavioral context “aggression” consists of both intra-group and inter-group aggressive behavior (i.e., chase, cuff, fight, lift tail, push, slap/claw, snap bite, uninhibited bite, chatter, tsetsetse, twitter, receiving of: cringe, escape, squeal, withdrawal posture and scream). Intra-group aggression was seen in all three social groups, but occurred at low rates. However, all groups regularly responded aggressively to neighboring groups. The characteristic “twitter” vocalization, indicating intergroup aggression [Epple, 1968; Martins Bezerra & Souto, 2008], was heard often. Inter- and intra-group aggression were collapsed into one context, in order to obtain sufficient aggressive interactions for statistical analyses. The behavioral context “fear” consisted of behavior that indicated submission, fear or anxiety (i.e., cringe, escape, flee, mob, scratch, self-groom, startle response, wet dog shake, withdrawal posture, alarm call, egg and ogg, squeal, scream, tsik, receiving of: chase, cuf, snap bite, push, slap/claw, and uninhibited bite). “Affiliation” was comprised of behavior that was considered friendly or positive social behavior (i.e., allogroom, approach, being together, body rub, carry infant, follow, grooming invitation, huddle, hug, lick, nuzzle rub, share food, suckle and transfer infant). Lastly, “play” included behavior seen in play (i.e., bounce, extricate, grip, mock bite, object play, play chase, play escape, play invitation, play wrestle, sliding on side, somersault, and stalk).
Visual displays concern facial expressions that is movement of tufts, eyes, mouth, or tongue; body postures that is arching back, raising body, and pilo-rection on the whole or part of the tail or back (Table I). For all but one of these visual displays at least one behavioral context has been reported in the literature, but often more than one context has been suggested, indicating that no clear consensus exists on their meaning (Table I).
Observations were conducted using The Observer (version XT 9.0). Data were entered in the format: focal animal; the behavior; recipient of behavior. The behavior could either be a visual display or aggressive, fear, affiliation and play behavior defining a context. For both these context-defining behaviors and visual displays the direction was noted that is “giving” or “receiving”. However, of seven visual displays (i.e., “coiled tail,” “head tilt,” “hind legs,” “hind legs rampant,” “side to side,” “tongue in-out”), the head and visual attention of the sending individual were typically not aimed at another individual and, therefore, it was difficult to determine whether these behaviors were directed to one specific individual. Therefore, only “giving” and not “receiving” these visual displays was analyzed.
The aim of the pre–post-event analysis was to test whether certain visual displays were found in close association with the presence or absence of behavior defining a particular context, following the methods of Preuschoft et al. . When a visual display was recorded during a focal observation, a time window of 3 min before and 3 min after the visual display was examined to calculate the baseline rate of behavior determining each context (Fig. 2). The time window was subdivided into blocks of 10 sec. We used 0/1 sampling to note in each “10-sec block” if any behavior that was part of a behavioral context was absent or present. These behaviors included both state and point behaviors. Thus, although several behaviors of one behavioral context could be displayed within a 10-sec block, the maximum score for each context in each time block was 1. The counting was done separately for all focal animals and for all behavioral contexts and visual displays.
Figure 2. This figure provides the detailed presentation of the PPE analysis for the visual display “arched bristle.” Box plots indicate the median of the distribution in the rate of aggression. Each box plot represents 10 sec. The visual display is given at time 0. Before and after this visual display, the X-axis displays eighteen 10-sec blocks, totaling 3 min. For each 10-sec block the median of the distribution of behavior signifying this context, in this case aggressive behavior, is provided with a box plot. The baseline frequency of the behavioral context was calculated from these data and is indicated here by the black line. Figures 3 and 4 provide the base line and the box plots from 10-sec block −2 to 2. The horizontal line indicates the average rate, “*” indicates a significant difference from the average.
Download figure to PowerPoint
The timing of a visual display within a focal observation varied, and sometimes we did not observe the full 3 min before or after the visual display was seen. Therefore, we sometimes analyzed incomplete time frames with only the full 3 min before or after the physical appearance. Instead of discarding these data, we corrected for the number of times each 10-sec block was observed. To do so, all 10-sec block counts were divided by the number of times that block was observed. To increase the number of visual display observations with a full 6 min block, we conducted focal observations of 20 min in the second half of the study.
For each visual display, we calculated per individual for each 10-sec block the rate of each behavioral context. Next, we calculated for each 10-sec block the average rate for all individuals that showed a particular visual display. Subsequently, the “baseline” rate for each of the four behavioral contexts was calculated by averaging the absence or presence of behavior over all 10-sec blocks of all animals over the 6 min time span surrounding the visual display (the solid line in the figures). We regarded 6 min as long enough for a rapid-moving animal such as the common marmoset to represent a reliable baseline.
We analyzed whether certain behavioral contexts were more often or less often seen in close temporal association with the visual displays. We concentrated the analysis on the five 10-sec blocks that directly surrounded the visual display (−20 sec, −10 sec, 10 sec in which the element was shown, plus 10 sec and plus 20 sec), as Preuschoft et al.  has shown that these points were specifically important in indicating the relation between a visual display and behavioral context. The division into 10-sec blocks also gave us the opportunity to note quick responses to a visual display. Henceforth, these surrounding 10-sec blocks will be referred to as time −2, −1, 0, 1, and 2.
We tested for animals for each of the five surrounding blocks to see whether the mean frequency of a particular behavioral context was significantly higher than the baseline. The variation between animals was taken into account by comparing the frequency for each animal separately with the baseline of the behavioral context. Thus, the chance of a bias due to outliers in the data was minimized. All individual values were compared to the baseline using a Wilcoxon-signed rank test (PSAW Statistics 18). The significance level was set at 0.05, however, since for each visual display five tests were done for each social behavior context, a Bonferroni correction was applied resulting in an alpha value of 0.01. Only visual displays where results had a P-value smaller than 0.01 (significant) or 0.05 (trend) are mentioned. All statistical tests were two-tailed.