Introduction

For species in which parents take care of their offspring, a special bond exists between the caretaker and infant. Bowlby (1969) referred to this as “attachment:” the bond to discriminate special individuals, maintain contact with them, and recover this contact when it is destroyed. Among primates, the development of this attachment has been extensively studied in macaques. By exchanging mother–infant combinations of long-tailed macaques, Negayama and Honjo (1986) showed that mothers did not seem to discriminate between their own infants and those of others during the first 0.5 months. However, in another report, pig-tailed macaque mothers did show discriminatory behavior towards their own infants within 1 week after birth (Jensen 1965). In this study, Jensen observed a tendency for mothers to pace around when isolated from their infants. This pacing behavior reduced when presented with their 6-day-old infants through Plexiglas compared with when they were presented with other infants. Jovanovic et al. (2000) reported improved discriminability of infant calls by female rhesus monkeys as their infants became more than a week old. Hence, the time at which mothers can discriminate their own infants differs depending on the species examined, the method employed, or the available modality.

A social preference test, whereby a mother and nonmother adult female are presented simultaneously to an infant, has also been used to investigate attachment in macaques (Sackett et al. 1967; Suomi et al. 1973, 1983). In this test, preference is assessed according to proximity to the stimulus animal. Using this method, it was revealed that Japanese macaque infants selectively approach and maintain close proximity to their mother by the age of 2 months (Nakamichi and Yoshida 1986). However, long-tailed macaque infants under 0.5 months seem to differentiate their mothers’ nipples from those of other females by contact (Negayama and Honjo 1986). Using a habituation–dishabituation method, Masataka (1985) showed that Japanese macaques aged 12–28 days could discriminate between their mothers’ voice and other voices. A study using preferential looking methods revealed that 2-day-old Japanese macaque infants raised by humans looked at their surrogate mother longer than other individuals (Yamaguchi 2000). Thus, one factor in infant attachment, the ability to recognize their caretaker, seems to develop early, while selective approach behavior is observed later in macaques. Since macaque infants scarcely have contact with individuals other than their mothers during the first month after birth, this discrepancy may be rational.

Callitrichid primates, marmosets and tamarins, live together as a large family, with the mother, the father and older siblings all taking care of infants (Ferrari 1992; Price 1992; Rothe et al. 1993; Bales et al. 2002; Kostan and Snowdon 2002; Washabaugh et al. 2002; Mills et al. 2004; Fite et al. 2005; Mota et al. 2006; Yamamoto et al. 2008). There are high costs for infant carrying in callitrichids: twining, a high maternal/infant weight ratio (Leutenegger 1973), metabolic expenses (Sanchez et al. 1999; Achenbach and Snowdon 2002), and reduced mobility (Schradin and Anzenberger 2001). It has been suggested that these energy and ecological demands of rearing two heavy infants are the reason for the existence of a communal breeding system. Infant transfers from one caretaker to the next occur smoothly. Usually, the next caretaker takes infants from the back of the current caretaker. Nonmother members of groups are highly motivated to carry infants (Schradin and Anzenberger 2003; Zahed et al. 2008). Infant transfers are controlled by not only caretakers but also infants themselves: they spontaneously transfer from one caretaker to another (Tardif et al. 2002). Infants therefore have passive and active contact with individuals other than the mother relatively soon after birth.

The division of caregiving behavior is affected by the presence or absence of alloparental resources (McGrew 1988; Ximenes and Sousa 1996; Santos et al. 1997; Nunes et al. 2001). Infants have different relationships with different caretakers, and each relationship changes according to development (Locke-Haydon and Chalmers 1983; Cleveland and Snowdon 1984). In fearful situations, infants seek protection from their father, who is most involved in infant carrying in cotton-top tamarins (Kostan and Snowdon 2002). In common marmosets, however, the removal of the father does not distress an infant as long as its mother remains available (Arruda et al. 1986). In these species, it is suggested that infants can form multiple, varied attachments with their various caretakers (Maestripieri 2003). Differences may exist in the parent–offspring relationship compared with macaques.

Parent marmosets do not seem to discriminate between their own infants and others in the early stage of development. When presented with infants that are less than 2 weeks old, mother marmosets attempted to carry other infants as well as their own (Saltzman and Abbott 2005). [However, Hilario and Ferrari (2010) reported that a mother who had few-day-old infants killed infants of other subordinate females.] Moreover, fathers of 2- to 4-week-old infants did not discriminate between their own infants and others (Zahed et al. 2008); they also approached and tried to carry both equally. On the other hand, Zahed et al. (2008) further reported that infants often rejected carrying attempts by an unfamiliar male. Although they did not include quantitative data on this behavior, if true it suggests that 2- to 4-week-old infants are able to discriminate between their father and other adult males. Thus, an investigation of the approach behavior of marmoset infants is likely to be of interest.

To investigate the approach behavior of infants toward their parents, in the present study we investigated the behavior of common marmoset infants in a social preference test. In this test, mothers and fathers as well as same-sex adults who were from another group and had identical status in that group were presented to infants aged 4, 10, and 15 weeks. Beginning around week 4, infants start to spontaneously leave their caretakers. Although the frequency increases with increasing age, infants are carried until weeks 8–12, and around week 15 are rarely observed on their caretakers (Yamamoto 1993). We therefore predicted that their behavior would change according to their independence.

As in previous studies evaluating mother–infant attachment in macaques using the social preference test (Sackett et al. 1967; Suomi et al. 1973, 1983; Nakamichi and Yoshida 1986), we presented stimulus animals to the subjects in a situation whereby they could watch but not mutually touch. In this situation, since the behavior and vocalization of the stimulus animals can influence the reactions of the test animals, the results include these effects.

Methods

Subjects

We tested 14 common marmoset infants, Callithrix jacchus; seven twins from five pairs (eight males and six females) born and kept at the National Institute of Neuroscience, National Center of Neurology and Psychiatry (NCNP), Japan. The experiments were conducted when the subject infants were 4, 10, and 15 weeks old. They were housed with their parents and older siblings in a cage 100 cm wide × 60 cm deep × 80 cm high in accordance with the Guidelines of NCNP. The cage contained perches, platforms and a nest box (20 × 20 × 20 cm). Although visual and tactile contact was almost completely restricted, olfactory and auditory communication was possible among cages. Feeding of a monkey diet was conducted between 1030 and 1230 hours. From Monday to Saturday, steamed sweet potatoes, bananas, and yogurt were fed between 1500 and 1830 hours. We carried out the experiments from April 2006 to January 2007.

Apparatus and stimuli

Tests were conducted in an I-shaped apparatus (Fig. 1) in a soundproof experimental room. This apparatus consisted of three separate areas: a central holding space and two approach wings. The holding space and approach wings consisted of transparent acrylic boards separated by two opaque sliding doors. A carrying cage for a stimulus animal was placed at the end of each wing. The carrying cage (21 × 24 × 22 cm) consisted of stainless steel with a transparent acrylic door on the wing side so that the subject could inspect the stimulus animal visually but not by touch. When sufficient sight was possible, the subjects were able to compare stimulus animals in the holding space. Auditory and olfactory cues were not completely eliminated; infants could hear the voices of stimulus animals and smell their odors through acrylic doors and openings in the apparatus. The holding space was used to keep a subject in the central space before starting a trial. Opening the two opaque sliding doors permitted the subject access to the two adjacent approach wings. Half of each wing was defined as the choice area (Fig. 1). The stimulus animals were one parent (mother or father) of the subject and one adult animal of the same sex from another group. Care was taken to ensure that parents and adults from other groups presented simultaneously were not related.

Fig. 1
figure 1

I-shaped maze used in this experiment

Full size image

Procedure

One trial was conducted per day for each subject. Before the test trials, two pretest trials were conducted in which no stimulus animals were presented; this was to allow habituation to the apparatus. After the pretest trials, we presented a father and an adult male in two test trials (father trials), and a mother and an adult female in another two trials (mother trials). The order and position of the stimulus animals were pseudo-randomized and counterbalanced.

All trials were recorded using a digital video camera located above the apparatus. The process was conducted as follows. First, the stimulus animals, after voluntarily entering the carrying cage under the lure of a piece of raisin, were respectively placed at the end of each wing. A subject was then manually captured and enclosed in an identically sized carrying cage and placed in the holding space for 30 s to calm it down. The trial was started by opening the two opaque sliding doors. One trial lasted for 180 s. The experimenter, whilst hiding behind a partition, monitored the subject’s movement on a monitor. The subject’s behavior was analyzed from videotape recordings.

Analysis of data

The total time (s) the subject stayed in the choice area (choice time) was counted. The total choice times from the two father trials and the two mother trials were analyzed by repeated-measures ANOVA to assess the effects of sex of the subjects, sex of the stimulus animals, age in weeks, and parent/nonparent. The first factor was a between-subject factor and the latter three were within-subject factors.

To test for parent preference at each age, we summed the choice times of the father trials and those of the mother trials for each age, and then compared those that used parent and nonparent stimuli based on a two-tailed paired t test.

We investigated which choice area the subject entered first (the first choice) and last (the last choice) in each trial. If the subjects showed discrimination and selected one of the stimulus animals in the holding space, it was deemed that the first choice was biased. However, if the subjects selected one side after watching from a close distance, the last choice rather than the first was considered to be biased. Using the first and last choices of four trials, we categorized the subjects into three groups: “parent-preferring,” those who chose the parent in three or four trials; “nonparent-preferring,” those who chose the parent in zero trials or one trial; and “neutral,” those who chose the parent and a nonparent in two trials. We examined whether the subjects chose their parent more in the first or last choice by comparing the frequencies of parent-preferring and nonparent-preferring subjects using a two-tailed binomial test.

Performance in the I-maze task depends on the motor activity of the subjects. Therefore, to determine changes in motor activity with age, we counted the number of turnovers in each trial and calculated the median for each subject. Turnover was defined as when a subject entered one choice area from the other. The difference in the number of turnovers between ages in weeks was analyzed using the Friedman test.

Results

The main effects of age in weeks and parent/nonparent were significant in repeated-measures ANOVA (Table 1). As shown in Fig. 2, the average choice time in response to parents was longer than that in response to nonparents in all cases but one. None of the other main factors and interactions were significant (Table 1). This indicates that the infants preferred their parents to adults from another group, and that preference changed according to age. Figure 3 shows the choice times for parents and nonparents. A comparison revealed that there was no significant difference between choice times for parents and nonparents at the age of 4 weeks (t = 1.35, df = 13, P = 0.20); however, the differences at 10 and 15 weeks were significant (t = 2.15, df = 13, P = 0.05; t = 3.45, df = 13, P < 0.01, respectively).

Table 1 Results of repeated-measures ANOVA for choice time
Full size table
Fig. 2
figure 2

Mean choice times of a male subjects in the father trials, b male subjects in the mother trials, c female subjects in the father trials, and d female subjects in the mother trials. Open bars show the choice times for parents and filled bars the choice times for nonparents. Error bars indicate SEs

Full size image
Fig. 3
figure 3

Mean choice times of the sums of both father and mother trials at each age. Open bars show the choice times for parents and filled bars the choice times for adults from another group. Error bars indicate SDs

Full size image

In addition, because Fig. 2 suggests that the infants might have preferred parents in father trials but not in mother trials at the age of 4 weeks, we compared the choice times for parent stimuli with those for nonparent stimuli in father and mother trials at the age of 4 weeks separately using a two-tailed paired t test. The comparison revealed that in the father trials, the parent choice time (mean = 186.3, SD = 78.5) was almost statistically longer than the nonparent choice time (mean = 114.3, SD = 79.0; t = 1.86, P = 0.085). On the other hand, in the mother trials, the parent choice time (mean = 138.3, SD = 74.7) did not significantly differ from the nonparent choice time (mean = 144.8, SD = 64.7; t = −0.18, P = 0.857).

In the first choice, the number of parent-preferring subjects did not differ from the number of nonparent-preferring subjects at all ages (Table 2). On the other hand, in the last choice, the number of parent-preferring subjects was significantly larger than the number of nonparent-preferring subjects at the age of 15 weeks. These data indicate that 15-week-old marmoset infants ended the trial near their parents more frequently than near nonparents.

Table 2 The nature of the first and last choice responses of the infant marmosets
Full size table

The number of turnovers in each age group is shown in Fig. 4. The difference in the number of turnovers between ages in weeks was significant (Friedman test, χ 2 = 13.11, df = 2, P = 0.0014). The differences between ages at 4 and 10 weeks and between ages at 4 and 15 weeks were also significant (z = 9.42, P = 0.0090; z = 12.01, P = 0.0025, respectively). The difference between ages at 10 and 15 weeks, on the other hand, was not significant (z = 0.17, P = 0.92).

Fig. 4
figure 4

Box plots for the number of turnovers in each age group. The thick horizontal lines represent the medians, the boxes indicate the quartile points, and the vertical lines indicate the ranges

Full size image

Discussion

The attachment of infants to caretakers involves two aspects: one is the ability of infants to recognize their caretakers; the other is selective approach behavior toward caretakers. In macaques, the former develops earlier than the latter. In callitrichid primates, details of the development of this behavior are as yet unknown. We examined the preference of common marmoset infants for their parents over adults from another group at the ages of 4, 10, and 15 weeks. At the age of 4 weeks, preference for a parent over an adult from another group did not significantly differ with regard to choice time, first choice, and last choice. On the other hand, at the ages of 10 and 15 weeks, infants spent significantly more time near their parents than adults from another group. In addition, at the age of 15 weeks, they often chose the parent side by the end of the trial, even though they initially chose the parent and nonparent sides equally. This suggests that they compared the two stimulus animals and actively chose their parents. The present results indicate that common marmoset infants selectively approach their parents by the age of 10 weeks.

Previous studies suggest that common marmoset infants recognize their mother or father within a few days or a few weeks after birth (Tardif et al. 2002; Zahed et al. 2008). In our study, however, 4-week-old infants did not show a clear preference for their parents over nonparent animals. Our study and previous ones differed in terms of whether infants could directly contact the adult animals or not. It is possible that 4-week-old infants cannot discriminate between their parents and others without direct contact. Olfactory cues are more important for mother recognition than visual cues in squirrel monkeys (Kaplan et al. 1977). Although our apparatus did not completely block odor flow from stimulus animals to the subjects, olfactory cues might be insufficient. Moreover, the voice of the mother is recognized by infants of Japanese macaque at less than 1 month (Masataka 1985), but vocal cues from stimulus animals were not presented positively to the subjects in our experiment, as described below. Finally, as the eyes develop rapidly during 4–8 weeks (Graham and Judge 1999), visual maturity was not sufficient at the age of 4 weeks. Therefore, sufficient cues might not be provided to 4-week-old infants to discriminate their parents from others in our experiment.

In addition to insufficiency of perceptual cues for 4-week-old infants, motor maturity might explain the results of our study. Four-week-old infants showed fewer turnovers than older infants. Although 4-week-old infants housed in their own cages have been observed to start climbing off their carriers and locomote independently (Ingram 1977; Yamamoto 1993), our data suggest the possibility that 4-week-old infants are not yet sufficiently developed to perform the I-shaped maze test.

In macaques, the time at which the infants start to selectively approach their mother in the social preference test is the period when they become primarily responsible for both breaking and making contact with mothers (Hinde and Atkinson 1970). At the same time, infants start exploring their surrounding environment using their mothers as a secure base (Maestripieri 2003). They are exposed to danger, such as kidnapping by group members (Maestripieri 1993; Silk 1980). In marmosets, in contrast, all group members are basically caretakers of infants. A rare exception is infanticide by dominant females in polygynous groups (Bezerra et al. 2007; Hilario and Ferrari 2010). Although intergroup encounters are frequent in wild populations of common marmosets (Hubrecht 1985), individuals carrying dependent infants seldom participate in direct intergroup interactions (Lazaro-Perea 2001). Considering these social environments, it is not strange that infants whose locomotor skills are not yet fully developed do not selectively approach their parents from a distance.

In the first choice, no bias towards the parents’ side was observed at any age. This could be explained by two possibilities. First, the subjects could not discriminate the stimulus animals in the holding space because of qualitatively or quantitatively insufficient visual, auditory, or olfactory cues, as described above. Common marmoset infants rarely show head cocking behavior toward stimuli located at a distance of more than 50 cm (Kaplan and Rogers 2006). This behavior is an aspect of visual exploration (Rogers et al. 1993). Therefore, it may be difficult for common marmoset infants to visually recognize objects over a distance of 50 cm, which was the distance from the holding space to the stimulus animals in our experiment. Second, although the subjects were able to discriminate between parents and nonparents, some dared to approach an unfamiliar individual as their first choice. At present, we are not able to examine which of these possibilities is the most likely.

In titi monkeys, monogamous and biparental New World monkeys, infants show attachment bias toward their father. They selectively approach their father when presented with their mother and father simultaneously in a Y-shaped maze at the age of 6 months (Mendoza and Mason 1986), and separation from the father elicited a greater stress response from them than separation from the mother at the age of 3–5 months (Hoffman et al. 1995). In cotton-top tamarins, 9- to 20-week-old infants, who are already independent, also show attachment bias toward their fathers. In a fearful situation, they run to their fathers (Kostan and Snowdon 2002). Separation from the father at the ages of 15, 30, and 45 days did not affect the behavior of common marmoset infants (Arruda et al. 1986). In addition, infant preference bias toward the father has not been shown in the situation of food transfer at the age of about 2–12 months (Vitale and Queyras 1997; Saito et al. 2008). In our results for 4-week-old infants, preference for parents over nonparents tends to be shown for the father but not for the mother. This result suggests the existence of an attachment bias toward the father in marmosets. However, since the methods used and the ages of the subjects differ from those in previous studies, it is difficult to form a conclusion about the attachment bias in marmosets.

It is also possible that the behavior of the stimulus animal could have affected the subject’s behavior, as mentioned in the “Introduction.” As monitored by the experimenter, who remained in a soundproof room throughout the test, the stimulus animals made no clear vocalizations, especially those produced for a distant infant such as a “phee call.” However, in a few cases, the stimulus animals reportedly emitted “chattering given when angry” (Epple 1968) to an approaching subject. To further determine the effect of stimulus animals, simultaneous recording of their behavior is also needed.

In conclusion, the present results show that it is possible to apply a social preference test to infant marmosets at the age of 10–15 weeks in the context of showing preference between parents and adults from another group. As a result, they were shown to selectively approach their parents and stay longer near them. To investigate the development of this behavior in detail, 6- to 7-week-old infants, not used in this experiment, should be used in this method.