In this study population, we observed 33 different abnormal behaviors and seven combinations of two behavior patterns (see Table 1). Across all 18 chimpanzees, the percent of scans spent on abnormal behavior did not differ significantly between sampling periods, i.e., abnormal behavior was equally frequent in 2003, 2004 and 2005 in the social setting (see Table 2). The same was true when early (ED) and late deprived (LD) subjects, male and female chimpanzees and the AM group and the MS2 group (Table 2) were tested separately. In the MS1 group, by contrast, we found a significant increase in the performance of abnormal behaviors from 2003 to 2004 (Table 2).

When tested within each period (Figure 2), the percent of scans spent on overall abnormal behavior did not differ significantly between ED and LD chimpanzees in the period subsequent to re-socialization, i.e., in 2003 (mean ± SEM: see Table 2; Mann-Whitney U Test: U = 24.500, p = 0.183). However, in the first year following re-socialization, i.e., in 2004, the ED chimpanzees performed abnormal behaviors significantly more frequently than did LD ones (U = 15.000, p = 0.027), and in the second year following re-socialization, i.e., in 2005, this trend was still recognizable (U = 18.000, p = 0.055), but not significant due to a higher variability among EDs. ED males did not differ in their mean percent of scans spent on overall abnormal behavior between the period of single caging (ED males: N = 6, mean ± SEM = 20.7 ± 8.2%) and subsequent to re-socialization, i.e., in 2003 (mean ± SEM = 13.7 ± 4.6%; U = 14.000, p = 0.589; see Figure 2). The overall level of abnormal behavior of male and female chimpanzees did not differ, neither in 2003 (mean ± SEM: see Table 2; U = 36.000, p = 0.762) and 2004 (U = 40.000, p = 0.965), nor in 2005 (U = 27.500, p = 0.286). In 2003, as well as in 2005, we found no difference in the percentage of scans spent on overall abnormal behavior between the three social groups (mean ± SEM: see Table 2; 2003: Kruskal-Wallis Test: H = 2.088, p = 0.352; 2005: H = 5.889, p = 0.053). However, in the first year following re-socialization, i.e., in 2004, the AM group significantly exceeded the MS2 group in their performance of abnormal behaviors (H = 7.940, p = 0.019, Dunn’s Post Test: AM vs. MS2: p < 0.05).

The impact of the different predictors could only be tested for the two most prevalent abnormal behavioral categories “Eating Disorders” and “Self-directed Behaviors”. “Eating Disorders” were significantly more prevalent in LDs than in EDs (100% vs. 70%, Chi-Square Test: χ² = 5.294, df = 1, p = 0.021) and in the two mixed-sex groups compared to the AM group (AM vs. MS1 vs. MS2: 57% vs. 100% vs. 100%, χ² = 14.389, df = 2, p = 0.001). It did not differ significantly between male and female individuals (70% vs. 88%, χ² = 2.051, df = 1, p = 0.152). There was a trend for “Self-directed Behaviors” to be more prevalent in EDs compared to LDs (100% vs. 75%, χ² = 3.571, df = 1, p = 0.059). “Self-directed Behaviors” were significantly more widespread in the AM and MS1 group compared to the MS2 group (AM vs. MS1 vs. MS2: 100% vs. 100% vs. 67%, χ² = 8.157, df = 2, p = 0.017). It did not differ significantly between male and female chimpanzees (90% vs. 88%, χ² = 0.022, df = 1, p = 0.881).

The comparison of the three most prevalent abnormal behaviors “Coprophagy & fecal manipulation”, “Hair pulling” and “Pat self” between ED and LD chimpanzees revealed that “Coprophagy & fecal manipulation” was significantly more prevalent among LD than ED chimpanzees (88% vs. 60%, Chi-Square Test: χ² = 5.297, df = 1, p = 0.026), in the two mixed-sex groups compared to the AM group (AM vs. MS1 vs. MS2: 29% vs. 100% vs. 100%, χ² = 44.026, df = 2, p < 0.001) and among females than males (100% vs. 50%, χ² = 16.667, df = 1, p < 0.001). “Pat self”, by contrast, was significantly more prevalent in ED compared to LD chimpanzees (90% vs. 25%, χ² = 36.739, df = 1, p < 0.001). We found “Pat self” to be more widespread in the AM and the MS1 group compared to the MS2 group (AM vs. MS1 vs. MS2: 86% vs. 80% vs. 17%, χ² = 47.902, df = 2, p < 0.001). It did not differ between males and females (60% vs. 63%, χ² = 0.073, df = 1, p = 0.857). “Hair pulling” did not differ between the ED and LD subjects (70% vs. 63%, χ² = 0.368, df = 1, p = 0.603) and between males and females (70% vs. 63%, χ² = 0.368, df = 1, p = 0.603. However, “Hair pulling” was significantly more widespread in the AM and the MS1 group compared to the MS2 group (AM vs. MS1 vs. MS2: 71% vs. 80% vs. 50%, χ² = 7.075, df = 2, p = 0.030).

“Eating Disorders” varied substantially within subjects across the three sample periods (Kendall’s Rank Correlation: 2003 vs. 2004: τ = 0.161, p = 0.506; 2003 vs. 2005: τ = 0.204, p = 0.401; 2004 vs. 2005: τ = 0.316, p = 0.192). We found no correlation in the occurrence of “Self-directed Behaviors” of individual chimpanzees between 2003 and 2004 (τ = 0.396, p = 0.103), as well as between 2003 and 2005 (τ = 0.255, p = 0.293). Between 2004 and 2005, however, the occurrence of “Self-directed Behaviors” within subjects stabilized (τ = 0.837, p = 0.001).

We found the occurrence of “Coprophagy” to be highly variable within subjects across the three sample periods (2003 vs. 2004: τ = 0.125, p = 0.606; 2003 vs. 2005: τ = 0.204, p = 0.401; 2004 vs. 2005: τ = 0.255, p = 0.293). We found no correlation in the occurrence of “Pat self” of individual chimpanzees between 2003 and 2004 (τ = 0.396, p = 0.103), as well as between 2003 and 2005 (τ = 0.255, p = 0.293) due to an increase in the number of chimpanzees performing the respective behavior (see Table S1). Between 2004 and 2005, however, spreading of “Pat self” within subjects did not change (τ = 0.671, p = 0.006). The occurrence of “Hair pulling” within subjects remained more or less constant between 2003 and 2004 (τ = 0.553, p = 0.023), but varied between 2003 and 2005 (τ = 0.236, p = 0.331), as well as between 2004 and 2005 (τ = 0.224, p = 0.357).

If we differentiate between the three sample periods, the 18 subjects of our study population did not differ significantly with respect to their diversity of abnormal behaviors (see Table 5). The same was true for ED and LD chimpanzees and for male and female subjects (see Table 5). The AM and the MS2 group did not differ in their diversity of abnormal behaviors over the years, whereas the MS1 group exhibited a significantly higher number of abnormal behaviors in the first year compared to the second year following re-socialization (Table 5). For the AM group, we ran a second analysis where we also included the diversity of abnormal behaviors during single caging (cage: mean ± SEM = 3.9 ± 0.9). However, no significant difference between the four activity periods became apparent (Friedman Test: χ² = 1.286, df = 2, p = 0.733).

ED vs. LD individuals did not differ with respect to the diversity of abnormal behaviors in the social setting (mean ± SEM: see Table 5; Mann-Whitney U Test: 2003: U = 29.500, p = 0.372; 2004: U = 27.000, p = 0.266; 2005: U = 33.500, p = 0.593). Likewise, we found no difference between males and females (mean ± SEM: see Table 5; 2003: U = 26.000, p = 0.229; 2004: U = 24.000, p = 0.168; 2005: U = 30.000, p = 0.397). The three social groups did not differ in their diversity of abnormal behaviors in 2003 and in 2005 (mean ± SEM: see Table 5; Kruskal-Wallis Test and Dunn’s Post Test: 2003: H = 4.492, p = 0.106; 2005: H = 1.209, p = 0.567). However, in 2004, individuals of the MS1 group significantly exceeded those of the MS2 group concerning the number of abnormal behaviors (mean ± SEM: see Table 5; H = 10.013, p = 0.007, Dunn’s Post Test: MS1 vs. MS2: p < 0.01).

Since the levels of abnormal behavior did not change significantly over the two-year period, neither within the whole study population, nor within early- or late-deprived subjects, males or females and within two out of three social groups, it might be suggested that an emancipation from original causes [27] took place in most of the abnormal behaviors performed by our study population. However, more detailed analyses revealed that the individual performance of the most prevalent abnormal behaviors varied substantially across the three sample periods. Thus, it seems that most of these abnormal behaviors are malleable. In other words, although the time spent with abnormal behavior was stable, the composition of abnormal behaviors did vary. As we cannot rule out that some short-lasting abnormal behaviors may have been missed by conducting scan sampling, we will clarify this question in detail according to Mason & Latham’s [28] classification in an ongoing study based on focal sampling data recorded in 2004.

Unexpectedly, in one of the two mixed-sex groups, the level of abnormal behavior even increased in the first year following re-socialization. Social group formation may cause an increase in the level of abnormal behaviors, at least in the short-run [20], since individuals have to adapt to the unfamiliar social stimulation. However, mean values of abnormal behaviors returned to the initial level in the second year of group-living. At least for the males of our study population, it could be demonstrated that elevated stress levels did not contribute to these increasing levels of abnormal behavior, since the fecal cortisol metabolite levels decreased during the early re-socialization period compared to the levels found during the habituation period [31]. Thus, it can be assumed that our study subjects are able to cope with social stressors. Nevertheless, a contrary effect became obvious within the ED individuals whose mean levels of abnormal behavior decreased subsequent to re-socialization, but one year after successful social rehabilitation, they increased again to values comparable to that found in the ED males during single caging and stayed at this high level throughout the second year of group-living. LD individuals differed significantly from EDs within these two years after re-socialization, which is in line with Warniment and Brent [16], who report of significant effects of age at onset of social deprivation on the level of abnormal behavior. Similar to the return of our re-socialized ED chimpanzees to the original levels exhibited during single caging, Baker [32] found no difference in the level of abnormal behavior between single and pair- or trio-housed chimpanzees after living in these settings for years, though the level in these biomedical chimpanzees lay in between that of our ED and LD individuals. Moreover, while ED chimpanzees spent about five percent of scans on affiliative social behavior subsequent to re-socialization, this level halved in the first and second year of group-living (Kalcher-Sommersguter, unpublished data). Accordingly, we assume that the abnormal behavior may have been replaced to some extent by social activity in ED chimpanzees in the short-run, but afterwards returned to their pattern of “non-social” abnormal behavior. LD chimpanzees, by comparison, invariably spent about eleven percent of their time on affiliative behavior (Kalcher-Sommersguter, unpublished data). Fritz and Howell [33] observed a decrease in affiliative interactions in rehabilitated chimpanzees as well and, as a solution, proposed to vary the composition of social groups after one to three years to stimulate social activity again. Sex did not affect the level of abnormal behavior per sample period, just as reported by Birkett and Newton-Fisher [17], whereas effects of group composition became apparent, as the ED-dominated all-male group significantly exceeded the LD-dominated mixed-sex group 2 in the first year following re-socialization.

We found 29 different abnormal behavior patterns that occurred in the social setting, which is virtually equal to the 26 different behaviors found by Walsh et al. [13], even though only half of the observed abnormal behaviors were identical between these two studies. Just as in zoo-living chimpanzees [17], coprophagy and hair pulling were most prevalent in this population of ex-laboratory chimpanzees. With respect to coprophagy, even median percent of time and scans spent on this behavior, respectively, are virtually identical, whereas our chimpanzees spent much more time on regurgitation compared to the zoo-living individuals [17]. We did not distinguish between eating or manipulating the feces, as these patterns mostly occurred in combination in our study population.

Eating disorders, including coprophagy and regurgitation, are presumed to be associated with the feeding regime in captivity [34], such as insufficient dietary roughage and lack of foraging opportunities [35], as well as with social deprivation and lack of adequate stimulation [13]. In our study population, eating disorders, mainly in terms of coprophagy, were significantly more prevalent in the LD compared to ED chimpanzees. Nash et al. [15] found coprophagy to be more widespread among mother-reared chimpanzees compared to hand-reared ones and suggested coprophagy to spread by social learning. We did not observe social transmission of coprophagy across the three sample periods, however, the longer time period of early social housing of LDs might at least partially explain this outcome. Unfortunately, reliable data on the prevalence of eating disorders, including coprophagy and regurgitation, during the laboratory setting are not available. However, two cases of social transmission of regurgitation occurred. In the first case, the female, Bonnie, was housed together with the male, Max, because both participated in a breeding program in the late 1990s. This co-housing had to be stopped shortly after, since Bonnie started to regurgitate as did Max [36]. In the second case, the male, Blacky, who had been regurgitating already during single caging, was observed to accurately reproduce the sequence and mode of regurgitation of the male, Max, during social housing in 2004 [37]. Moreover, in the social setting, regurgitation was found to be persistent in four out of seven individuals who performed this behavior. That regurgitation is extremely difficult to overcome was shown in a case study where even behavioral training along with a dietary change could not eliminate this unhealthy behavior [38], possibly due its presumed self-stimulatory and comforting effect (reviewed in [39]). However, while the levels were stable or even decreased in the four females, they increased in the three male chimpanzees. Thus, it can be assumed that these males met the criteria for obsessive-compulsive disorder [26].

Self-directed behaviors, including self-patting and hair pulling, contrary to eating disorders, were more widespread among ED compared to LD subjects. All ED chimpanzees exhibited self-patting similar to most human infants reared in restricted environments and lacking maternal stimulation, who perform this self-stimulating abnormal behavior as well [40]. Beyond that, self-patting became more widespread in the whole study population in the two years following re-socialization, possibly by social transmission. Hair pulling is triggered by long-term exposure to anxiety- and boredom-inducing environments in diverse animal species, from mice to non-human and human primates (reviewed in [41]). Accordingly, hair pulling was the most prevalent self-directed behavior in our study population and did not differ between ED and LD chimpanzees, as all of them spent decades in similar adverse environments. Self-clasping, which was more common in our study population than in zoo-living chimpanzees [17], is associated with social deprivation, as it emerges when clinging to the mother is unachievable and, thus, has to be redirected to the own body (reviewed in [42]). Furthermore, the occurrence of self-clasping might provide an indication of generalized anxiety disorder (GAD) in chimpanzees [26]. It occurred in three of four ED females, but in none of our ED males, which gives rise to the suspicion that females might be more vulnerable to the absence of body contact in early infancy and consequently to the development of GAD, which is in line with findings from humans [43].

More severe categories of abnormal behaviors, such as dissociative behaviors, whole-body stereotypies and self-injurious behavior (SIB) were rare in our study population. Three chimpanzees in our study population exhibited dissociative behaviors—in humans associated with traumatic life events [44]—where, e.g., a dissociated limb served as a play partner or a limb was “floating”. Whole-body stereotypies, which are mainly caused by social deprivation [45], disappeared in two of four subjects after the period subsequent to re-socialization, similarly to what Fisher et al. [46] found in institutionalized children years after adoption. That rocking was extremely rare in our chimpanzees might be due to the fact that the cages were freely suspended and swinging in the first years in the laboratory and, thus, seemed to provide the respective stimulation [47]. Additionally, low levels of rocking might also be traced back to the fact that these chimpanzees experienced at least some social companionship in early infancy, and rocking was found to be most prevalent in subjects who were separated from their mother within the first months of life [48]. Adverse early rearing conditions, repeated biomedical experimentation and single housing are supposed to be the main factors linked to the etiology of SIB in non-human primates, which in turn is assumed to serve affect-regulatory functions as it does in humans [49]. SIB occurred only once and was ceased, respectively, in two ED males in the social setting. Briefly summarized, in six out of seven chimpanzees who performed severe abnormal behaviors, the behaviors disappeared in the course of group-living. Thus, altogether, an improvement in the most severe categories of abnormal behavior is recognizable, emphasizing the curative potential of social companionship. Further evidence of social housing serving a therapeutic role is provided by the fact that stereotypic scratching, pacing and swagger and screaming were not observed at all when the chimpanzees were socially housed. Table S1 provides an overview of the occurrence of all seven abnormal behavioral categories per individual and year in the social setting.

Concordant to the findings of Birkett and Newton-Fisher [17], all study subjects exhibited abnormal behaviors with a similar range with respect to the individual diversity (1–12 vs. 2–14), but while all our ED subjects exhibited at least four different abnormal behaviors, LD individuals displayed a broader distribution. However, the overall number of different abnormal behaviors was not affected by any of the predictors, except for group composition. This was caused by the two females in the MS1 group—Martha and Pünktchen—who exhibited the highest variety of abnormal behaviors in this study population. Focusing on the persistence of abnormal behaviors revealed that only a small fraction of the abnormal behaviors occurred at all three sample periods, i.e., solely eight out of a total of 33 different abnormal behaviors. Almost half of the study population exhibited one persistent abnormal behavior. These persistent abnormal behaviors may be so-called self-stimulatory behaviors defined as “learned, operant behaviors for which the reinforcers are the perceptual stimuli automatically produced by the behavior” and which are extremely difficult to eliminate, as the reinforcing properties are also retained in enriched environments [50].

Our study had methodological weaknesses, as we conducted scan sampling and, thus, could not evaluate the absolute frequencies and durations of specific abnormal behaviors. Furthermore, as some rarely occurring abnormal behaviors with short durations may have been missed, results have to be interpreted with caution. Additionally, as our sample size was modest (N = 18), the results may not generalize to a larger population of captive chimpanzees. The strength of this study, however, is the expanded observation time over a period of two years, which gives insight into the development of abnormal behaviors of rehabilitated chimpanzees in the course of group-living.