Chapter Five

Experiments in Separation

Like experiments in the deprivation category, separation experiments involve removing infants from their mothers. However, the timing of this removal differs between the two types of experiments. In deprivation experiments, the removal occurs before the infant becomes psychologically attached to its mother. In separation experiments, the removal occurs when the infant is older, after is has become attached to its mother. One other difference between separations and deprivations is that the former usually are temporary, whereas the latter are permanent.

There are several types of studies that are related to mother-infant separation. These involve separating infants from other infants, confining infants in vertical chambers (also known as “pits” and “wells of despair”), or rearing infants with punitive artificial mothers. These studies are far outnumbered by experiments involving mother-infant separation (Table Ill). All of these miscellaneous studies have been conducted on rhesus monkeys at the University of Wisconsin. These studies are considered here because of their historical and conceptual association with mother-infant separation studies; maternal separation experiments led to the other types of experiments and all of these studies ‘are considered to be animal models of depression.


Researchers have offered several reasons for conducting separation experiments on animals. Mother-infant separation has been considered to be an animal model of mother-infant separation of humans, as well as a model of any depression or anxiety that results from the separation.

‘Infant-infant’ separation studies have similar rationales. Studies involving vertical chambers or punitive surrogate mothers were attempts to create animal models of psychopathology.

Not all separation experiments on animals have been conducted as animal models of human phenomena. Specifically, a few of the early animal studies of mother-infant separation are best characterized as basic research.

Human Studies of Mother-Infant Separation

Primate research on mother-infant separation is based on human studies that began in the 1940s. Many of these pioneering human studies were the same ones that formed the basis of animal studies of maternal deprivation (see Ch. 4). Psychiatrists such as Bowlby, Spitz, and Goldfarb had been studying children in orphanages, hospitals, and other settings. They found that children sometimes protested being separated from their mothers and later lapsed into despair. Bowlby termed this the “protest-despair reaction” to separation. Spitz (1946) coined the term “anaclitic depression” for this form of despair. Some children also showed a third phase, detachment, in which they were indifferent, if not hostile, to their mothers during brief visits or when reunited.

Early human studies not only identified the protest-despair-detachment reaction to separation, but also began identifying factors that influence the reaction’s severity. The influence of factors such as the child’s age at separation, the availability of alternate care givers during the separation, and the duration of separation, were reviewed in two World Health Organization monographs (Bowlby 1952, WHO 1962). The early human research also began addressing the longer-term consequences of maternal separation on personality development.

Animal models of maternal separation came into use in the 1960s, two decades after the early human studies. The animal modelers advanced their usual criticism of human studies, namely, that these studies were not rigidly controlled experiments. The models purportedly were going to provide definitive answers using monkeys. The models were used to assess (1) the short- and long-term effects of a temporary separation from the mother, and (2) the factors that influence the severity of these effects. Both behavioral and physiological effects of separation were studied.

Summary of Animal Separation Experiments

Separation experiments have been conducted on nonhuman primates, dogs, cats, and rodents, with the primate studies accounting for about three-quarters of the total number of studies (Table III).

Cat and Dog Experiments

Three separation studies have been conducted on cats. These involved prematurely separating kittens from their mothers and sometimes their littermates as well (Seitz 1958, 1959; Rosenblatt et al. 1963; Rosenblatt 1974; Guyot et al. 1980). These studies apparently were designed as basic research, and are rarely (if ever) mentioned in the literature on separation models of human phenomena. Seitz’s study is noteworthy for its harmful effects. All the cats were stressed, some became “catatonic,” and some died.

Puppies were the subjects of several clinically oriented separation studies. Most of these studies were conducted by J.P. Scott and coworkers at the Jackson Laboratories. The separations in Scott’s experiments usually lasted only a matter of minutes, but this was enough to induce acute distress. Longer separations had far more damaging consequences.

Two of Scott’s most damaging experiments involved repeated, overnight separations or repeated 22-hour separations. The former was described as a “drastic emotional experience” accompanied by “serious losses in weight” (Scott & Bronson 1964, p. 183). In the latter, “the animals were . . . severely upset. . . . [They] ate little . . . and one . . . even died. . . . ” (Scott et al. 1973, p. 13).

The most noteworthy findings from Scott’s experiments were: (1) Separation from social companions induces fear or loneliness. (2) Separations are most damaging when they occur during the ‘critical period for socialization.’ (3) Various psychotropic drugs are not “effective in reducing distress vocalizations that are elicited by separation, except at doses that produce sedation or reach near-toxic levels” (Scott et al. 1973, p. 17). Scott and Bronson thought that their “most important result [was] the discovery of a general social drive,” (p. 192) based on fear or loneliness. According to the researchers, the existence of such a drive should be investigated in humans.

Scott and coworkers were not the only researchers to conduct separation studies on dogs. Senay’s (1966) study had an unusual twist: laboratory-reared puppies were separated from the experimenter, not their mother. (Mothers had been removed earlier). This study was an unsuccessful attempt to create an animal model of depression.

Rodent Experiments

Myron Hofer, a psychiatrist at a medical college, has been prematurely separating rat pups from their mothers since the late 1960s. His experiments have emphasized physiological as well as behavioral variables. He gave two rationales for his experiments (Hofer 1970). First, little attention had been given to effects of mother-infant separation in “lower” mammals such as rats, and second, few studies examined the physiological effects of separation.

Hofer attempted to systematically analyze the pups’ separation experience into its component parts. Specifically, he viewed the separation experience as consisting of the loss of a source of nutrition, warmth, tactile stimulation, olfactory stimulation, etc., as well as the psychological loss of the mother. Hofer attempted to relate these aspects of the separation experience to the physiological and behavioral responses to separation. Once a particular cause and effect relationship was indicated, Hofer attempted to determine the biological mechanism underlying this relationship.

In a typical experiment, Hofer would “cull” his rat litters to 8-10 pups each, implant electrodes in the pups, remove the pups from their mothers, and monitor certain physiological and behavioral changes during the next 24 hours. The fate of the pups after this period is unclear from Hofer’s reports.

The separated pups usually were divided into groups that were subjected to different treatments. Some treatments functioned to control for certain variables. For example, one separated group might receive milk through a stomach tube as a control for maternal nutrition. Another group might be housed with an unrelated adult female who potentially could provide all forms of stimulation that a mother could (e.g. tactile stimulation), except providing milk. The same goal was sometimes achieved by returning the mother to the separated pups after her teats were ligated.

Hofer’s experiments had two themes. First, different aspects of the maternal separation can have different effects, each mediated by a different biological mechanism (Hofer 1973a). Second, rat mothers are regulators of physiological processes in their pups (Hofer 1978).

In Hofer’s (1970) first experiment, maternal separation caused a drop in heart rate and respiratory rate. These responses were related to deprivation of nursing (Hofer & Weiner 1971; Hofer 1971, 1973b). Mother’s milk regulated infants’ heart rate via a neural pathway, not the circulatory pathway (Hofer 1973a). The experiment that showed this involved cutting the pups’ spinal cord in various locations. Follow-up experiments (Hofer & Weiner 1975) involved a host of surgical and biochemical procedures.

Separated pups were hyperactive when placed in a novel environment (Hofer 1973c). This was related to the loss of mothers’ warmth, as well as mothers’ tactile and olfactory stimulation (Hofer 1973a, b, c, d; 1975a, 1976a). These experiments involved (1) simulating tactile stimulation with electric shock and (2) destroying the pups’ sense of smell with chemicals. Not surprisingly, the hyperactive pups were also slow in falling asleep (Hofer 1973b). Deprivation of maternal nutrition was implicated in this response.

The pups’ sleep-wake cycle was radically altered by maternal separation (Hofer 1976b). This experiment was highly invasive, involving implantation of cranial electrodes and head plates. Hofer noted that other researchers had obtained similar results in rats, as well as in guinea pigs and monkeys.

Fourteen day-old pups did not survive permanent maternal separation unless supplied with heat (Hofer 1975b). Even with heat, over 50% of the pups died. All separated pups initially showed decreased physiological functioning. The survivors showed a physiological recovery that was attributed to physical maturation.

Hofer and coworkers (Stone, Bonnett & Hofer 1976) performed one of the most damaging series of separation experiments. In survival experiments, pups were permanently deprived of their mothers and some were also denied food and water for a week. Time of death was noted. In metabolic experiments, pups were killed and their bodies “were minced and dried . . . in [an] oven” (p. 243). In developmental experiments, pups were decapitated before or after a starvation period.

The topic in this study was the role of body temperature in survival and development. One separated group was supplied with warmth and another was not. No pups lived through the survival experiments when denied food, though the warm pups survived a little longer than the pups without warmth. When food was available, no pups denied warmth survived whereas most pups with warmth survived. After summarizing their results, the experimenters noted that “similar consequences of prolonged hypothermia have been observed in . . . human infants, young rabbits, and piglets . . . ” (p. 246).

In another experiment, Hofer and coworkers subjected pups to stressful procedures in order to induce stomach “erosions,” It was already known that maternal separation increases pups’ susceptibility to these erosions. The follow-up experiment (Ackerman, Hofer & Weiner 1975) found that this susceptibility was age-dependent. The procedure involved subjecting starved rats to restraint. Some pups developed “massive hemorrhage.” The mechanism for the susceptibility to gastric erosions was later found to be faulty temperature-regulation, induced by maternal separation (Ackerman, Hofer & Weiner 1978).

Researchers other than Hofer conducted separation experiments on rodents. Ader and co-workers conducted three separation studies that were part of a series of experiments linking early social experiences to later susceptibility to stressful or pathogenic procedures. The first of these three experiments was the basis for Hofer’s experiments on gastric erosions (see above).

Ader’s experiment linked early maternal separation to susceptibility to gastric erosions (Ader et al. 1960) All pups were killed either by the experimenters or by the procedures themselves. Ader’s second experiment suggested that the increased susceptibility to gastric erosions does not result from loss of mother’s milk (Ader 1962). In his third experiment, Ader subjected pups to injections of cancerous cells (Ader & Friedman 1965). The separated pups showed higher mortality than non-separated ones.

Newell (1967) subjected mice to early maternal separation and then tested their ’emotional reactivity’ and learning ability. Based on human studies, he expected ’emotionality’ to be high and learning ability to be low. No such effects were found. Not only did the mice respond differently than humans, the two strains of mice reacted differently from each other.

Plaut and Davis (1972), two psychiatrists working at a hospital, conducted a series of mother-litter separations in rats. A high percentage of the separated pups died. A lower percentage died if the pups were kept with a non-lactating adult female (“aunt”). The same improvement in survival was also achieved by returning mothers to their pups after the mothers’ nipples had been burned shut. The only important difference between these findings and Hofer’s earlier studies (see above) was that Plaut and Davis showed that the aunts interacted with the pups.

Another experiment by Plaut and Davis (1972) concerned the effect of premature weaning on body weight and brain chemistry. This experiment also overlapped heavily with previous experiments cited by the authors.

Plaut later repeated some of his experiments using “uncles” as well as “aunts” (Plaut et al. 1984). Uncles, like aunts, reduced the high death toll of early weaning. The mere presence of an aunt behind a screen also had a positive effect on pup survival, though not as great as when aunts could interact with pups. Post-mortem examinations linked death to high urine retention, although experimental stimulation of urination failed to reduce pup mortality. Koch and Arnold (1976) performed a follow-up experiment to Hofer’s demonstration that heart rates in rats drop following maternal separation (see above). The researchers knew from Hofer’s work that artifical feeding would ameliorate this change in heart rate and that the presence of a non-lactating foster mother would not. The researchers wanted to know if foster mothers would have an ameliorating effect while pups were artificially fed. It did. To show this, the researchers surgically excised the teats of foster mothers to preclude nursing.

Primate Experiments

Separating primate infants from their mothers can have profound consequences. As we shall see, it can induce anxiety, despair, and even death. The act of separation and as well as the period of maternal absence can be stressful. Separation is often accomplished by brute force. One study noted that separation required “a trained team of three men” (Seay & Harlow 1965). The following quote describes the immediate response of pigtailed monkeys to the separation procedure:

Separation of mother and infant monkeys is an extremely stressful event for both mother and infant, as well as . . . for all other monkeys [nearby]. The mother becomes ferocious toward attendants and extremely protective of her infant. The infant’s screams can be heard almost over the entire building. The mother struggles and attacks the separators. The baby clings tightly to the mother and to any object to which it can grasp to avoid being held or removed by the attendant. With the baby gone, the mother paces the cage almost constantly, charges the cage occasionally, bites at it, and makes continual attempts to escape. . . . The infant emits . . . shrill screams intermittently and almost continuously for the period of separation (Jensen & Tolman 1962a, p. 132-133).

Another quote in the same report describes the reunion of mother and infant:

The reuniting of mother and infant is a remarkable event. Both return to each other immediately. The mother sits quietly holding her baby, and if no attendants are present, she very quickly seems content and relaxed. All is quiet in the room. No more piercing screeches of the baby or sounds from the mother are heard. . . . [After 15 to 40 minutes,] the infant would venture away from its mother and appear to be free to leave and explore as it did in the control situation, or, it would go to sleep being held by the mother . . . (p. 133).

University of Wisconsin and Wisconsin Regional Primate Center

Separation experiments at Wisconsin initially involved mother-infant separation, but later also involved infant-infant separation, vertical chamber incarceration, and miscellaneous procedures. All of these experiments were conducted on rhesus monkeys, as were the Wisconsin experiments on affection and deprivation.

Seay, Hansen, and Harlow (1962) began the Wisconsin separation experiments to see if the human “protest-despair-detachment” reaction (see above) also occurred in rhesus monkeys. For unexplained reasons, Seay et aL separated mothers and infants by putting a transparent partition between them. Therefore, mothers and infants could still see and hear each other, unlike the human situation. One infant in this study “suffered severely from anorexia and sleeplessness” (Harlow and Harlow 1971, p. 218). In a follow-up experiment (Seay & Harlow 1965), mothers and infants were separated visually as well as physically.

Similar results were obtained in both experiments. Not surprisingly, the infants initially were highly agitated, crying and struggling to reach their mothers. When their agitation subsided, they cried and played little with other infants in an adjoining cage. Mothers and infants eagerly embraced one another after their three-week separation was terminated. Within a few weeks their relationship seemed to return to normal.

Seay et al. interpreted their subjects’ reaction to separation as one of protest and despair; they therefore concluded that their results were in “general accord” with human data. They admittedly found little evidence in rhesus monkeys for the detachment phase.

Mitchell, Harlow, and coworkers (Mitchell et al. 1967) separated infants from foster mothers as well their own mothers. Infants were briefly but repeatedly separated from one mother until transferred to another. These manipulations left the infants psychologically disturbed. The disturbance was evident a year later when these youngsters, as well as others in control groups, were paired with an aggressive, fear-provoking adult male.

Suomi et al. (1970b) introduced infant-infant separations as an alternative procedure to mother-infant separations. They noted that separating infants from other infants had certain methodological advantages over separating them from their mothers. The act of separating infants from each other was easier to accomplish. Because of this, infants could be easily separated from one another repeatedly, and more data on transient responses to separation could thereby be obtained. These advantages brought one disadvantage, however, in that infant-infant separation has less relevance to mother-infant separation in humans.Suomi et al. therefore proposed infant-infant separation as a general model for studying loss of a loved one or ‘attachment object.’

Suomi et al. separated a group of infants from one another no less than twenty times. During their four-day separations, the infants were kept in physical — but not visual — isolation. No explanation for this research design was provided. The reaction to separation was similar to the protest-despair response to mother-infant separation; however, the repeated separations were so traumatic that they stunted the infant’s behavioral maturation.

Separation studies at Wisconsin and elsewhere were viewed in a new light after McKinney and Bunney (1969) called for development of animal models of depression. More attention was paid to the depression-like despair phase of the response to separation. Thus, mother-infant separation and infant-infant separation were considered animal models of depression.

Harlow created the infamous vertical chamber as another means of modeling depression (Harlow & Suomi 1971b). Harlow himself described this chamber as “a modified form of sadism” (Harlow, in Tavris 1973, p. 76). This tiny chamber had sides of stainless steel that sloped inward (Photo 10). This design frustrated the attempts of incarcerated monkeys to climb up the sides and cling to the wire mesh roof. After a few days of scrambling around, incarcerated monkeys spend most of their time huddled on the floor of the chamber (Suomi & Harlow 1972b, see Photo 11).


Photo 10
Vertical chambers, also known as ‘pits’ and ‘wells of despair,’ used to induce psychological disturbance in monkey infants. Reproduced from Harlow (1974) with permission.

The devastating effects of vertical chamber confinement were revealed in two studies (Harlow & Suomi 1971b, Suomi & Harlow 1972b). Thirty and 45 days of incarceration both produced “severe psychopathology.” After confinement, the infants spent most of their time huddling and clasping themselves, and performing other self-directed “disturbance behaviors.” “Most striking was the virtual absence of social activity” during months of social testing (Suomi & Harlow 1972b, p. 14). Suomi and Harlow saw “enormous potential” for the vertical chamber, in that its severe effects were produced rapidly and lasted indefinitely. Harlow had managed to create a device that was more devastating than isolation chambers (McKinney et al. 1971).


Photo 11
Typical behavior of a rhesus monkey infant during long-term confinement in vertical chambers. The view is from the top of a chamber. From Harlow (1976) with permission.

Harlow and Harlow (1971) summarized some vertical chamber experiments that remain unpublished in detailed form.

McKinney, who had called for development of animal models of depression (see above), later joined forces with Harlow and Suomi at Wisconsin. McKinney and his new coworkers presented their studies of infant-infant separation and vertical chamber confinement as attempts to “go beyond the mother-infant separation model” of depression (McKinney et al. 1971, p. 134). McKinney’s praise of vertical chamber studies is puzzling. These studies radically violated one of his proposed criteria for an animal model to be valid; namely, that a syndrome be produced in the animal model by the same or similar cause of the syndrome in humans (McKinney & Bunney 1969). No one would argue that depression in humans is caused by vertical chamber confinement.

The Wisconsin researchers devised another odd method of inducing depression. Infants were reared with cloth surrogate mothers whose bodies could be made ice cold, too cold for the “punished” infants to cling to (Harlow et al. 1973). The cold surrogate was considered an advance over previous attempts to build a rejecting surrogate mother. These earlier versions were either made of uncomfortable material (e.g. sandpaper), or could dislodge infants with hidden spikes, catapults, compressed air or vigorous shaking (Baysinger et al. 1973). These earlier versions of rejecting mothers all failed to produce psychopathology in infants (Harlow & Harlow, 1971).

In the Harlow et al. study, brief drops in the surrogate’s temperature failed to induce depression. Harlow then kept the surrogate cold throughout the night, when infants invariably cling to their surrogates. The infants’ responses were so severe that one died. This was termed “psychological death.” All infants were severely dehydrated. “Although we feared for their lives, we did not discontinue the nocturnal cold . . . period until . . . we received the necropsy report on the dead infant” (p. 305).

According to Harlow’s article, he continued to subject infant monkeys to this procedure; however, no further publications resulted from these experiments. He considered this cold-surrogate procedure to be a model of extreme depression, though no mention is made of this model in reviews of animal models of depression.

The separation and vertical chamber models of depression had a longer lifespan at Wisconsin than did the cold-surrogate model. In fact, the separation and chamber procedures were some-times combined in the same experiment (Suomi 1973b, Suomi & Harlow 1975b).

Lewis et al. (1976) abandoned mother-infant separation as a model of depression after obtaining disappointing results. The results were from a series of five experiments that varied in the degree of access that infants had to their mother (and others) after separation. Most infants were killed after several days of separation, and no infant was returned to its mother. The results were highly variable and did not support the notion of a clear-cut protest-despair reaction. Lewis et al. thought that “future research in separation models of depression should concentrate on the peer separation paradigm” (p. 705). It should be emphasized that the researchers’ comments pertain to models of human depression, not models of mother-infant separation per se.

Two Wisconsin experiments (McKinney et al. 1973b, Young et al. 1973) used “well-controlled experiments” to test a well-known finding of human clinical psychology. The clinical finding was that the effects of stressful situations in adulthood are more severe if such situations were experienced during childhood. McKinney et al. (1973b) employed vertical chamber incarceration as their stressful procedure. Monkeys were profoundly disturbed by vertical chamber incarceration, but their disturbance was no greater if they had been subjected to similar treatment earlier in life.

The early stressful experience in the second study was separation from peers, coupled with vertical chamber confinement. The later stressful experience was no less than 28 separations from peers, one per day and 23 hours each. The experimental group did have a more severe reaction to the multiple separations than did a control group that was not subjected to early peer separation and chamber confinement. “Our data indirectly support previous studies in human subjects … ” but these data “are best viewed as tentative . . . and suggestive of further research” (p. 404). The mixed results of these two studies shed little light on the question of early behavioral sensitization in rhesus monkeys, let alone humans.

Most Wisconsin separation studies examined the effects of temporary separations from mothers or peers. Suomi et al. (1973) examined the effects of permanent separations from mothers. The severity of the short- and long-term responses to separation depended on (1) the infants’ age at separation and (2) whether the infant was housed alone or with a peer following separation. “One is tempted to conclude . . . that the reaction to separation is intensified by subsequent individual housing but mollified by subsequent social housing, a conclusion reached long ago by Spitz (1946) from study of human infants” (p. 382, emphasis added). The researchers admittedly could not even draw this modest conclusion, however. They pointed out that the disturbance of the solitary-housed monkeys could have resulted from the restrictive housing and not the trauma of separation.

The number of novel separation experiments that can be designed is limited only by the imagination of the researchers. This point is well illustrated by a Suomi et al. (1976) study in which four infants were first separated from their mother by wire mesh, then by a plastic panel, and then were separated from each other by a transparent panel. A period of reunion with mothers and peers followed, and this, in turn, was followed by permanent separation from the mother. Infants in a control group were subjected only to the final, permanent separation. Reaction to the wire separation was relatively mild, but the two other temporary separations were “debilitating.” However, the behavior of these individuals eventually resembled that of the non-separated controls during the reunion period. Both groups reacted similarly, and relatively mildly, to the final, permanent separation, despite their different early histories.

The results were contrary to the researchers’ expectations. Suomi et al. expected increasingly severe reactions to the first three separations, and major differences between experimentals and controls in reaction to the final separation. The results are at odds with Young et al. (1973), who reported that early separations tend to exaggerate reactions to later separations. Numerous uncontrolled factors and defects in experimental design were offered to account for the discrepancies between studies.

In their 1977 review, Suomi and Harlow conceded that much of their depression research had not gotten beyond what was already known for humans. A decade and a half of their experiments had been devoted to validating the rhesus monkey model of depression. Later studies were also attempts to further validate the model. For example, Suomi et al. (1978) wanted to see if a particular anti-depressant drug already in use for humans also worked in depressed monkeys. The drug imiprimine lessened the severity of the response to peer separations, but its effectiveness waned rapidly after treatment was discontinued.

Lewis (1978) conducted a series of experiments that apparently have not been published except in abstract form. Three groups of infants were separated from their mothers and temporarily housed alone in wire cages or isolation chambers, or housed together in a peer group. Individuals in the group had the most severe reactions, which included “failure to thrive” (p. 4509). Lewis, following Cairns, attributed this response to the complexity of the peer group’s separation-environment, in which the loss of mothers as social facilitators was most acutely felt.

A 1983 report by Suomi et al. noted that no previous primate study had repeatedly separated mothers and infants over short periods of time. Their study accomplished this. They were interested in the short- and the long-term effects of repeated maternal separations on infants. Mothers and infants were separated eight times, reunited for several weeks, and then the whole process was repeated. A second group of infants was not separated from their mothers. Eventually both groups were permanently separated from their mothers and housed in peer groups. Infants protested after each separation but this waned with repeated separation. Signs of the despair reaction were minimal. However, repeated separations appeared to somewhat retard development of the mother-infant relationship. Suomi et al. discussed the similarities and differences between these findings and those of earlier Wisconsin studies. The patterns in protest and despair reactions were consistent with social learning theory.

Wisconsin researchers focused on the behavioral response to separation until the 1970s; thereafter, physiological as well as behavioral responses were examined. Most of these physiological/behavioral experiments were conducted by McKinney and coworkers in the Department of Psychiatry. These experiments sometimes involved invasive surgical procedures, even death, in addition to the usual psychologically traumatic procedures of earlier studies.

Physiological and behavioral data gave conflicting evidence as to whether infants formed stronger attachments to their mothers or to inanimate surrogates (Meyer et al. 1975). Strength of attachment was inferred from the infants’ response to stress during maternal separation. Stress was induced by placing infants with “mechanical monsters” that made loud noises and sudden movements.

Another physiological study centered on the protest phase of the protest-despair response to separation. Breese et al. (1973) wanted to know if the anxiety characteristic of the “protest response” was associated with the production of certain chemicals in the brain (catecholamines). Human and animal studies had already suggested a link between anxiety and catecholamine production. Breese et al. separated infants from their mothers, observed them in the protest stage, and then killed them to analyze their brains. As expected, catecholamine-producing chemicals were elevated.

A related study focused on the despair phase of the protest-despair response to separation (Kraemer & McKinney 1979). AMPT, a drug that inhibits catecholamine production, was administered to infants that were repeatedly separated from peers or mothers, or infants that were reared in isolation. The resulting dose-response relationships revealed that AMPT induced depressive effects in all groups, but that the peer-separated group was more sensitive to this effect than was the mother-separated group.

This experiment was based on clinical reports linking catecholamine inhibitors to depression in a small percentage of human patients. Clinical data suggested that these patients shared certain predisposing factors. The rationale for the animal study was as follows:

While some of these factors, such as previous history of depression, ongoing psychosocial and environmental stress, can be intuitively identified, there are practical as well as ethical problems involved in actually testing the relative contribution of these factors in precipitating human depression and thereby determining their importance in a quantitative fashion (p. 33).

According to the researchers, the contribution of their animal experiment was in demonstrating that catecholamine inhibitors do interact with developmental and social stressors, such as peer separations.

In a recent study by McKinney’s group, cerebrospinal fluid (CSF) was extracted from monkeys undergoing repeated peer separations (Kraemer et al. 1983). The results indicated that monkeys with relatively low levels of norepinephrine in their fluid are: (1) “more likely to show a severe despair response to . . . separation” and (2) “more sensitive to the [catecholamine] depleting effects of AMPT and show greater depressant effects of the drug. Repeated separations do not affect the level of CSF [norepinephrine], but reductions in [other CSF components do occur]” (p. 141). These results suggest that low CSF norepinephrine is a stable characteristic of certain individuals, and predisposes them to severe responses to separation. In contrast, levels of other CSF components fluctuate in response to repeated separations.

The researchers viewed depression as an outcome of “life events,” such as separation, and constitutional factors, such as CSF norepinephrine levels. The existence of constitutional factors is one reason for individual differences in response to separations.

Recently, individual differences in response to separation also were emphasized by Suomi and coworkers. However, Suomi’s group focused on anxiety, whereas McKinney’s group focused on depression. Suomi et al. (1981) gave four reasons in arguing that anxiety was well-modeled by social separations: (1) the protest phase in the protest-despair response to separation resembles an anxiety-driven panic attack in humans; (2) rhesus and human infants display similar ‘anxious attachments’ to their mothers after reunion; (3) both display similar anxiety in anticipation of predictable separations; and (4) both display similar recovery after administration of the drug imipramine (Suomi et al. 1981).

The Wisconsin monkeys were considered to be anxious when they displayed behavior normally associated with fear, but in the absence of fear-provoking circumstances. This behavior was highly variable depending on the monkey’s age, sex, background, and current circumstances, but it included self-directed disturbance behavior (see Ch. 4) and excessive clinging to the mother. Monkeys with a history of separations displayed a chronically high level of anxious behavior. They appeared “edgy” or “uptight.” This high baseline level of anxiety was increased during “life threatening” events, such as an ongoing separation. However, a monkey’s background and current circumstances did not always permit an accurate assessment of its anxiety level. Constitutional factors, inborn or genetic, also appeared to be involved.

Following the human literature, Suomi’s group began looking for links between anxious behavior and the autonomic nervous system, which governs physiological variables such as heart rate and breathing rate. The responsiveness of the autonomic system to stressful situations was compared to levels of anxious behavior. The first measure of ‘autonomous reactivity’ to be investigated was heart rate (HR).

Month-old monkeys were repeatedly exposed to a mild stressor (a sound stimulus) and their change in heart rate (ΔHR) was monitored. Throughout one year of surrogate rearing, those infants with the highest ΔHR had the highest levels of anxious behavior. In contrast, ΔHR was not correlated with other behavior (Baysinger et al. 1978).

The same monkeys were monitored until puberty (Suomi et al. 1981). After their surrogate mothers were removed they were housed in partial isolation and subjected to various stressful procedures (e.g. daily handling). ΔHR values taken early in life continued to predict levels of anxious behavior. Hence, the researchers succeeded in linking individual variation in anxious behavior to a constitutional factor (ΔHR). Moreover, there was “suggestive” evidence that this factor was heritable.

Suomi (1981) conducted a “replication study,” this time using monkeys reared with peers rather than surrogates. The infants were subjected to repeated separations from peers during their second, fourth and tenth months. The findings were similar to those in the earlier studies. ΔHR was correlated with anxiety levels, but only during stressful conditions (during and shortly after separations).

The most recent study in this series (Suomi 1983) involved HR and a second measure of autonomic reactivity: plasma cortisol (a hormone associated with stress). Both measures were correlated with one another when the monkeys were exposed to stressful conditions (e.g. peer separations), but not under baseline conditions. Both physiological variables were correlated with anxious behavior measured months before or after the physiological recordings; again, however, the correlations were apparent only under stressful conditions.

The results suggested that individuals who have extreme behavioral reactions to stress also have extreme physiological reactions. These extreme behavioral and physiological reactions are stable over time, but highly reactive individuals appear “normal” in the absence of stress.

Suomi has suggested that excessive anxiety can be associated with two other problems, excessive depression and retarded social development. The same individuals who are highly anxious during the protest phase of separations are also the ones who become depressed if the separation endures (Suomi 1986). This claim awaits published supporting evidence, however. Suomi also argues that chronic anxiety inhibits exploration and positive social interaction, and thereby retards social development (see also Suomi & Harlow 1976).

University of Cambridge, England

Several mother-infant separation experiments were conducted at the University of Cambridge. The subjects were rhesus monkeys, as in the Wisconsin studies. However, the monkeys at Cambridge lived more natural lives than those at Wisconsin. They were housed as groups in large outdoor enclosures. The Cambridge researchers, under the direction of ethologist Robert Hinde, were therefore able to study aspects of mother-infant interactions, such as spatial relationships, that were largely precluded by the restrictive housing in the Wisconsin experiments.

In a typical Cambridge experiment, mothers were taken from their six-month old infants and returned six to 13 days later.

Infants in the Cambridge studies clearly were disturbed by separations (Hinde & Spencer-Booth 1966, Spencer-Booth & Hinde 1967). The separated infants usually sat in a hunched posture, crept about on their hind legs, and otherwise acted depressed. They gave “distress calls,” apparently to signal their mothers. The separated infants also were more susceptible than non-separated infants to mildly frightening tests, and paid less attention to new objects in their environment. The separated infants sought care from other individuals in the group, but received far less than their mother would have provided.

These infants were emotionally unstable when reunited with their mothers. They clung to their mothers and remained near them to an exaggerated degree. They threw “violent tantrums” when their mothers started to reject them. They were characterized as “anxiously attached” to their mothers.

The seriousness of the infants’ disturbance was evident in other ways. The infants sometimes became very upset for no apparent reason, or they awoke “terrified” and clung to their mothers. Also, signs of the disturbance persisted at least two years (Hinde & Spencer-Booth 1968; Spencer-Booth & Hinde 1971c, Hinde et al. 1978). Some of the testing for lingering disturbance involved subjecting infants to “mildly frustrating” situations, such as having food placed just outside of their reach or being threatened by a researcher wearing a “grotesque” mask.

A Cambridge study (Spencer-Booth & Hinde 1971a) had alarming results. Nineteen infants of various ages were separated. Of these, seven died during the experiment. For the youngest ones, the researchers were “unable to say how far the deprivation experience contributed to the deaths . . . ” (p. 179, emphasis added). Two other infants collapsed and were prematurely reunited with their mothers. The survivors responded to separation as in the first study, with depression during separation and excessive clinging and trailing of the mother during reunion.

The researchers concluded that the responses of rhesus monkeys and humans to maternal separation were similar. Their conclusion is convincing only for the despair stage of the three-stage human response (protest, despair, detachment). The evidence for protest in the Cambridge experiment was distress calling accompanied by much idle sitting. This idleness contrasts with the frantic scampering that other researchers liken to the protest reaction in humans. Moreover, there was no evidence of detachment in these studies.

These early Cambridge experiments all involved six-day separations. Spencer-Booth and Hinde (1971b) wanted to determine if rhesus infants would be more disturbed by longer separations. The researchers separated infants for thirteen days and compared the results to those from their earlier studies using one or two six-day separations. After reunion (but not before), “the severity and durability of the depression [were] related to the length of the separation” (p. 604). One infant died during this experiment.

One finding from the early Cambridge experiments was that rhesus monkey infants differed markedly from one another in the severity of their response to separation. The infants’ age, sex, and associations with other individuals accounted for little of this variation. A follow-up experiment related this variability to tension in the mother-infant relationship (Hinde & Spencer-Booth 1970). Infants who had the tensest relationships with mothers during reunion were the most disturbed. Tension was measured by the degree to which infants were rejected by their mothers and were responsible for maintaining proximity to their mothers. Tension after reunion was correlated with tension before separation. This indicated to Hinde and Spencer-Booth that rhesus mothers make only limited adjustments to individual differences in the demands of their infants after separation. The researchers conceded that in other species, such as humans, mothers might be more flexible in meeting the post-separation demands of their infants.

This experiment suggested that rhesus infants are disturbed not only by the separation itself, but also by the resulting disruption in their relationship with their mother. The same conclusion was reached by a later study (Hinde & Davies 1972a, b). This study, unlike previous Cambridge studies, involved removing infants, not mothers, to a strange environment during the separation period. It was expected that infants would be more disturbed in the “infant-removed” situation than in the “mother-removed” situation. The reverse occurred. This was attributed to the mothers’ behavior. Mothers were less disturbed when left in their home environment during separation, and therefore were better able to tend to their infants at reunion. In the mother-removed experiments, mothers at reunion had to divide their time between reestablishing dominance relations with other group members, and tending their infants. The researchers thought that there were no human parallels to these findings (Hinde & Davies 1972a, b).

A follow-up study (McGinnis 1978) supported the conclusions of the previous study. Both infants and mothers were removed from their social group and then separated them from each other. The results from this “mother and infant removed” study indicated that removal of mothers disturbs mothers, which in turn, causes them to be less responsive to their infants upon reunion. Mothers that ranked lowest in their group’s dominance hierarchy were most disturbed by removal from their group. Consequently, their relationship with their infants took longest to recover.

In another study, McGinnis (1979) removed both mothers and infants from their groups but not from each other. The results indicated that maternal separation, and not removal from group companions, was responsible for infant disturbance.

Downstate Medical Center, New York

Kaufman and Rosenblum conducted several mother-infant separation studies at Downstate Medical Center. Rosenblum is a former student of Harlow. His subjects were two species of monkeys closely related to rhesus monkeys: pigtailed monkeys and bonnet monkeys. The adult pigtails, and probably the bonnets as well, had been born in the wild. The monkeys were housed in groups in relatively large indoor cages. During the separation experiments, mothers were removed from these groups for up to several weeks. Their infants were several months old at the time of separation.


Photo 12
An agitated pigtailed monkey after its mother was taken away as part of a separation experiment. The infant is grimacing in fear while fleeing from an adult female that had just rebuffed its approach. Reproduced from Kaufman & Rosenblum (1969) with permission.

The first Downstate study (Kaufman & Rosenblum 1967a, b) showed that pigtail infants were severely disturbed by removal of their mothers. The act of separation elicited loud screams and involved a “massive struggle.” Infants were “acutely distressed” for the rest of the day on which their mothers were removed (Photo 12). The next day, three of the four infants were deeply depressed (Photo 13).

Each infant sat hunched over, almost rolled into a ball. . . . Movement was rare. . . . The movement that did occur [usually] appeared to be in slow motion. . . . The infant rarely responded to a social invitation or made a social gesture, and play virtually ceased. . . . Occasionally he would look up and coo (1967a, p. 1020).

The researchers interpreted these responses as fitting the protest-despair framework, only they referred to these phases as “agitation” and “depression”, respectively. A third phase, not seen in the Rhesus monkey studies, was “recovery.” Recovery was gradual and not complete even at the end of a month of separation. One infant showed only the agitation and recovery phases. It had turned to other adults for comfort during the separation period, whereas the other infants tried to cope alone. Only the latter became depressed.


Photo 13
A pigtailed monkey after its mother was taken away as part of a separation experiment. The infant is “showing the characteristic posture and the face of ‘grief.’ ” Reproduced from Kaufman & Rosenblum (1969, p. 687) with permission.

Mothers and infants showed a resurgence of interaction when reunited (Photo 14). Some aspects of the mother-infant relationship remained altered for at least three months, as in the Cambridge studies. These lingering effects contrast “dramatically” with the results of Seay and Harlow.

The most significant aspect of this study is that pigtails responded to separation in a qualitatively different manner from rhesus monkeys. As mentioned above, rhesus did not show the pigtails’ recovery stage. Moreover, neither species showed the detachment stage found in humans.

A more dramatic example of species differences was provided by the next Downstate study (Rosenblum & Kaufman 1968, Kaufman & Rosenblum 1969). The researchers repeated their earlier study, but used bonnet macaques instead of pigtails. The striking result was that the depression phase was virtually absent. The lack of depression was attributed to the care that separated infants received from other adults. Indeed, some infants were fully adopted, although all but one returned to its mother at reunion. These studies showed that data from one animal species do not necessarily generalize to a closely related species—let alone to humans.

The Downstate researchers subsequently made various attempts to induce depression in bonnet infants. One scheme was to remove mothers and leave infants with unresponsive foster mothers. Another scheme was to remove infants from their mothers and place them alone in a cage. Both procedures failed to induce noticeable depression. Kaufman attributed these negative results to the infant’s age and maturity. These results were briefly summarized by Kaufman (1973) but have yet to be published in detail.


Photo 14
A pigtailed monkey mother and infant reunited after a separation experiment. Reproduced from Kaufman & Rosenblum (1969) with permission.

Kaufman eventually succeeded in making bonnet infants depressed. His experiment (Kaufman & Stynes 1978) was conducted after he moved from Downstate to the University of Colorado Medical Center. A bonnet infant was reared in a group that contained pigtails as well as bonnets. When the bonnet infant reached five months of age, the researchers removed its mother and all other bonnets from the group. The infant exhibited pigtail-like severe depression.

The same report also detailed two other sets of experiments; these duplicated Kaufman’s earlier studies of pigtails and bonnets. The only noteworthy difference in the follow-up experiments was that bonnets showed more indications of depression than in the earlier studies, though the reaction still was not severe.

The most recent report of separation experiments at Downstate was in 1980 (Rosenblum & Plimpton ‘1980). Pigtailed infants looked depressed during the first separation night, but not during the daylight observations the next few days. Rosenblum and Plimpton viewed the infants’ behavior both as a response to loss of mother and as an attempt to cope without the mother. According to this view, infants are seeking to restore the emotional equilibrium that the mother’s presence provided.

One of the Downstate separation experiments was physically as well as psychologically damaging to the monkey subjects. It was conducted over a decade ago but remains unpublished. A few details of the experiment were mentioned by Schlottmann and Seay (1972), who cited Kaufman’s data. Kaufman separated bonnet and pigtail infants and placed them with unfamiliar adult females. The infants “barely survived a single day” given the “brutality” of the females (Schlottmann & Seay 1972, p. 339).

University of Colorado Medical Center

Separation experiments have been conducted at the University of Colorado since the early 1970s. The subjects were pigtailed monkeys and, to a lesser extent, bonnet monkeys. The research was directed by psychiatrist Martin Reite, who focused on several physiological factors in the infant monkeys’ response to separation. These experiments were part of a trend begun in the early 1970s to add physiological data to the mass of behavioral data on infants’ responses to separation. These experiments were also viewed more generally as animal models of the increased susceptibility to disease and death following separation or loss of a loved one in humans. Reite claimed that the physiological responses to separation in monkey infants might throw light on the mechanism for this increased susceptibility in humans (Reite & Short 1983).

In the Colorado experiments, infant monkeys were housed with their mothers and other group members until they were five months old. At that time, mothers were removed for several days. The infants’ physiological patterns were monitored before, during and after separation via telemetry. A telemetry transmitter and power supply were surgically implanted in the infants’ abdominal walls, and electrode leads were “tunneled subcutaneously to their recording sites” (Reite & Short 1978, p. 1248).

The physiological responses to maternal separation included alterations in heart rate and rhythm, body temperature, sleep patterns, EEG patterns, and circadian rhythms. These results are summarized by Reite and Short (1983). The general conclusion from these studies is that maternal separation impairs physiological functions in infants by altering regulatory mechanisms.

The Colorado data on sleep disturbance add to the pathetic portrait of monkey infants whose mothers have been taken away. These infants take longer to fall asleep, awaken more frequently from sleep, and spend less total time asleep, than infants left with their mothers (Reite & Short 1978)

Reite et al. (1978) repeated some of their earlier experiments using cloth surrogate mothers instead of real mothers. For unexplained reasons, infants were reared with surrogates in social isolation, inside wooden crates. The physiological and behavioral responses to separation were less intense than those of infants reared with real mothers in social groups. The weak conclusion was that attachment bonds to cloth surrogates “may well be different than the bond to a living mother . . .” (p 427).

Reite et al. (1981a) also repeated some of their earlier studies, this time using a longer separation period (10 days). The longer separation was an attempt to observe the “recovery stage” noted in the Downstate experiments on pigtails. Little behavioral recovery occurred, however, providing another example of conflicting results from studies on the same species.

Two other studies focused on the apparent increase in disease and death following a separation or loss of a loved one in humans. One hypothesis for this increase is that separation or loss impairs the immune system. Two studies of human widows, cited by the Colorado team, strongly supported this hypothesis. Reite and coworkers tested and supported the same hypothesis. The experiments involved infant-infant separation (Reite et al. 1981 b) and mother-infant separation (Laudenslager et al. 1982).

In their most recent experiment, the Colorado researchers went to extraordinary lengths to induce depression in bonnet infants. Following Kaufman and Stynes (see above), Reite and Snyder (1982) separated a bonnet infant from its mother and all other bonnets in the group. This left the infant with unresponsive pigtails, and it became depressed. Its physiological responses to separation were similar to those seen in pigtails, notwithstanding the difference between separation from the mother (pigtails) and separation from all species members (bonnets).

The Kaufman and Stynes (1978) experiment discussed under Downstate experiments, for convenience, actually was carried out at Colorado, after Kaufman had moved from Downstate to Colorado.

Reite and Short (1983) recently found evidence that the physiological responses to maternal separation may result from factors other than the disruption of attachment bonds

. . . if a nursing infant [monkey] is deprived of the ability to nurse by putting a . . . vest on the mother, physiological changes similar to those seen during separation can be produced . . . (p. 247).

This experiment calls in question the interpretation of all the Colorado separation research. It also opens the door to a host of analytic experiments along the lines of Hofer’s rat studies.

Stanford University and Stanford Research Institute

Researchers at Stanford conducted several mother-infant separation experiments on primates since the late 1960s. Most of their subjects were squirrel monkeys. These monkeys inhabit forests in Central and South America; the adults used in the initial Stanford experiments had been purchased from an importer. A few Stanford separation experiments involved rhesus monkeys.

Most of these studies involved relatively brief (30-minute) separations and therefore focused on the panic that characterizes the immediate response to separation. Both hormonal and behavioral responses to separation were examined. The hormonal analyses were restricted to cortisol, which is associated with stress. Blood for these analyses was usually obtained by puncturing the heart of an anesthetized animal with a syringe. The same animals were sometimes subjected to this procedure more than once during a given experiment. When this happened, the monkeys were subjected to a maternal separation for each blood sample.

In the first Stanford separation experiment (Kaplan 1970), only behavioral responses to separation were investigated. Mother-infant pairs of squirrel monkeys were housed alone to preclude adoption of separated infants by other females. The infants were relatively independent of their mothers when separated and, not surprisingly, showed a relatively mild reaction to both separation and reunion. (One infant did die during separation, but apparently for unrelated reasons.) Kaplan attributed the mild reactions to the squirrel monkey’s permissive maternal style.

A cortisol analysis by Mendoza et al. (1978) indicated that a separation lasting 30 minutes was a “potent disturbance” to mother and infant squirrel monkeys. Rough handling (a part of the separation procedure) by itself elevated cortisol levels, but not if mother and infant were left together to comfort one another. The latter suggested that mother-infant interactions can suppress the hormonal response to stressful procedures.

This conclusion was also supported by a follow-up experiment involving the same separation procedure but a different blood sampling schedule (Levine et al. 1978). The opportunity for mutual comfort not only suppressed a cortisol response, but also reduced the cortisol response if the latter had already been activated.

In contrast to the Stanford studies summarized so far, most of the following studies involved mothers and infants living in groups, not in isolated pairs. For example, Coe et al. (1978) separated group-living mothers and infants to determine if adult females (“aunts”) could reduce the disturbance of separated infants. Disturbance levels were inferred from the infants’ hormonal and behavioral responses to separation. The behavioral response, but not the hormonal response, suggested that aunts could reduce the disturbance of separated infants. This decoupling of behavior and physiology suggested that behavioral quietude may mask shifts in internal state (Coe & Levine 1981). An earlier study also showed that aunts could reduce behavioral signs of disturbance in squirrel monkeys (Rosenblum 1971).

The Coe et al. study also suggested, along with a later study (Hennessy et al. 1979), that separation from a cloth surrogate mother is less disturbing than separation from a real mother. An earlier study (Mendoza et al. 1978) suggested that the response to separation from surrogates and real mothers was similar, but that study had weaker controls. Hence, the later experiments suggested that squirrel monkeys form greater attachment bonds to real mothers than to inanimate surrogates. Separation studies on rhesus monkeys and pigtailed macaques yielded similar results (see above).

Vogt et al. (1980) separated mother and infant squirrel monkeys and recorded the responses of nearby adults. Infants were removed from their group and caged either in or out of view of the adults. Only females showed an elevated cortisol response, with lactating females showing the highest response. The fact that non-lactating females responded at all was in sharp contrast with findings on rats . .” (p. 187). Among males and females, dominance ranks and cortisol levels were inversely related.

Vogt and Levine (1980) compared the responses of squirrel monkeys to partial separation (physical but not visual separation) and complete separation. The degree of separation did not affect the mothers’ or infants’ behavioral or hormonal responses. One reason offered for this result was that “. . brief separation seems so distressful for the squirrel monkey mother that the presence of her infant does not intensify her behavioral agitation” (p. 831).

This study raises several disturbing points regarding the value of this research. First, the results contrast with data from rhesus monkeys. The researchers noted that “there are several major differences between the present study and the previous research” on rhesus monkeys that confound comparisons (p. 831). One wonders how many studies would have to be done to reach a consensus of what the relationship is between degree of separation and response to separation. In the absence of a consensus, studies could be conducted indefinitely.

A second disturbing point is that cortisol levels are the lynchpin of this research, yet their interpretation is unclear. For example, an elevated cortisol level in one aspect of the present experiment may have resulted from “disturbance,” “fear,” “heightened arousal,” or “curiosity” (p. 831). Why bother measuring cortisol levels if their interpretation is so unclear? Marks (1981) was also skeptical of the Stanford researchers’ reliance on cortisol levels. In discussing these experiments, he noted:

. . . I am always a bit puzzled about the meaning of the cortisol response, especially in view of findings, for example, that comic films can produce as much increase in cortisol level in humans as can frightening films . . . (p. 175).

A third point concerns the degree to which the cortisol response system of squirrel monkeys is representative of primates. In regard to large differences between mothers’ and infants’ cortisol responses, the researchers wrote: “Other primate species apparently do not show these differences. . . . Under [baseline] and separation conditions rhesus mothers and infants exhibit comparable levels of plasma cortisol, as do human mothers and infants . ” (p. 832). The admitted aberrancy of the squirrel monkey’s cortisol response system calls into question the value of these separation studies.

In another Stanford study, Gunner et al. (1981) subjected squirrel monkey infants to a two-week separation. Separation induced brief cortisol elevations that returned to near-baseline levels within two hours. This suggested that the infants’ agitation was brief; however, some behavioral indications suggested that infants were agitated throughout the separation period. “Despair” responses were infrequently observed. Individual differences in cortisol responses to both separation and reunion were related to the degree of tension in the mother-infant relationship prior to separation. The researchers suggested that “for some infants, adapting to maternal loss may be less stressful than adapting to reunion with the mother” (p. 74).

A recent report summarized several new experiments (Coe et al. 1 983). In one experiment, mothers and infants were separated for 6 to 24 hours to provide data on a prolonged separation. Gunner et al. (see above) had already studied two-week separations, but apparently more data were deemed necessary. At any rate, the new study provided more evidence that behavioral and hormonal indicators of the protest response to separation could provide conflicting results.

A second experiment indicated that separated infants were more disturbed when housed in a strange environment than when left in the familiar environment. This experiment involved repeated four-hour separations; earlier experiments involving half-hour separations, but otherwise similar to the present experiment, were deemed too brief to assess the significance of strange vs. familiar separation environments. The present study agrees with a Cambridge study on rhesus monkeys (see above).

Another experiment summarized by Coe et al. examined the effects of repeated separations on the protest response. The infants’ behavioral responses (but not cortisol responses) attenuated over repeated separations. This result differed “radically” from Wisconsin data, but this discrepancy could have resulted from a multitude of factors, including the more benign procedures in the present experiment. The influence of procedural variation was inadvertently shown in a second experiment, which yielded increasing levels of behavioral response across separation.

The second study was a comparison of repeated separations with a fixed or variable duration. Fixed-interval separations elicited an increasingly severe behavioral disturbance, whereas variable-interval separations elicited a more stable and mild response. The researchers pointed out that these results are somewhat counterintuitive, because predictability is usually associated with less disturbance. However, the results are consistent with Levine’s hypothesis that in stressful situations, predictability without control over events may exacerbate, not ameliorate, the response.

Rhesus monkeys, not squirrel monkeys, were the subjects in a study by Smotherman et al. (1979). Both infants and mothers were removed from their group and then separated from one another (separation condition) or housed together (control condition). The same procedure on the same species was conducted at Cambridge but cortisol levels were not monitored in that study (see above). The cortisol levels of infants, but not of mothers, was higher in the separated condition. The rhesus mothers’ response contrasts with the response of squirrel monkey mothers summarized earlier. The infants’ cortisol response was related to maternal dominance rank. These conclusions were admittedly tentative owing to possible flaws in experimental design.

Miscellaneous Institutions

Many separation experiments were conducted at various institutions in addition to those mentioned above. Some of these studies involved the same primate species (rhesus, pigtailed, bonnet and squirrel monkeys) whereas other studies involved different species ( java monkey, paths monkey, langur, orangutan and gorilla).

Several studies added to the massive amount of data on rhesus monkeys accumulated at Wisconsin, Cambridge and Stanford. Three of these studies were conducted at the California Primate Center and involved Harlow’s former students, Mitchell and Mason.

  • Brandt et al. (1972) attempted to compare infant responses to separation from the home cage and separation from the mother in the home cage. Infants in one group were reared alone and those in another group were reared with their mothers. The former were separated from their home cage, whereas the latter were left in the home cage but separated from their mother by a partition. Unfortunately, the home cages in the two groups were not the same; the isolates were reared in isolation chambers whereas the mother-reared infants were housed in wire cages. Given that the researchers were comparing apples to oranges, it is not surprising that they found several differences between the two groups in response to separation. Numerous conclusions were drawn from the experiment despite the authors’ recognition of this and other defects in experimental design.
  • Several primate separation studies have shown that the presence of mothers reduces behavioral signs of insecurity in infants. Hill et al. (1973) showed the same effect on a physiological sign of insecurity. A high level of the hormone cortisol was the indicator of insecurity or stress. For unspecified reasons, the researchers used cloth surrogates instead of actual mothers. Moreover, all of these infants were reared in isolation, some in totally enclosed cages. A similar study with similar results was conducted at the Delta Primate Center (Candland & Mason 1969).
  • Erwin (1974) observed the reaction of adult rhesus monkeys to recordings of a rhesus infant calling. The calls were taped immediately after the infant was separated from its mother. Females responded to the calls as if an infant were calling from their lap. Erwin did not use a neutral sound as a control, so he was forced to conclude that “there is no way of knowing from this study whether [the] subjects would have responded similarly to some other auditory stimulus . .” (p. 184).

Two other studies of rhesus monkeys were conducted at institutions other than the California Primate Center. In one of these studies, rhesus monkeys were subjected to a series of four mother-infant separations (Chappell & Meier 1975). Two involved separating infants from mothers (“infant-removed”) and two involved separating mothers from infants (“mother-removed”). The Cambridge researchers had already used the infant-removed versus mother-removed comparison, but they did not subject the same infants to both procedures. Chappell and Meier noted similarities and differences between their findings and those of the Cambridge studies. Also, it was noted that the repeated separations had diminishing effects, which contrasts with results from Suomi et al. (1970).

The other rhesus study extended separation experiments to the natural environment (Singh 1 975). Singh removed some mothers from two free-ranging groups in India and observed the response of their infants. The separated infants had to procure their own food during the 3-day separation period. The separated infants clearly were upset and restless. Singh mentions frequent “crying.” The results indicated that free-ranging rhesus infants became agitated by maternal separation. The separation period was too short to conclude that the same could be said for the despair phase.

Four separation studies of pigtailed monkeys were conducted at the University of Washington. These studies preceded the Colorado studies of pigtails, and emphasized mothers’ behavioral responses to separation, not the infants’ physiological responses. These studies are most note-worthy for their poignant descriptions of the monkeys’ behavior during the separation procedure and at reunion (see p. 78 above).

A separation study of squirrel monkeys was conducted at the University of Arizona (Jones & Clark 1973). These monkeys were living in groups, whereas the squirrel monkeys in the Stanford experiments that preceded this study were housed as solitary mother-infant pairs. Mothers were drugged to reduce the trauma of the separation procedure. The results were interpreted as consistent with the human protest-despair responses, although strong signs of despair (e.g. huddling) were not seen. Jones and Clark noted differences from the Stanford studies and offered post-hoc explanations for these differences.

Several new species were added to the list of primates involved in mother-infant separation studies. These species included patas monkey (Preston et al. 1970), java monkey (Schlottman & Seay 1972), Japanese monkey (Minami 1977), langur (Dolhinow 1980; Dolhinow & Murphy 1983), gorilla (Nadler & Green 1975), and orangutan (Nadler & Codner 1983). The experiments on patas and java monkeys were directed by Seay, a former student of Harlow. These studies provide further evidence of apparent species differences in response to separation.

Dolhinow studied langurs because infants receive substantial care from adult females other than their mothers, which is unusual among primates. When mothers of target infants were removed from the group, their infants showed signs of agitation followed by depression, but the response was extremely variable among infants. Two infants died. “Most infants adopted substitute caregivers in their mother’s absence, and a majority of infants elected to remain with [their] adopted female . . . when the mother returned to the group” (1983, p. 122). The two infants that died received too little substitute care. Dolhinow emphasized the importance of two variables in influencing a langur infant’s response to maternal separation: maternal restrictiveness and the availability of substitute caregivers. Those infants that reacted severely to maternal loss, including those that died, revealed that langur infants are strongly attached to their mother, despite contact with other adults.


The Nature and Extent of Suffering

Separation experiments exact a heavy toll of animal suffering. Maternal separation causes acute distress in primates, dogs and cats. This distress gives way to despair in primates, and probably dogs and cats as well. Maternal separation can cause long-term psychological damage even if infants are reunited with their mothers. Long-term effects have been studied in primates; they include anxiety, anxious attachment to loved ones, and retarded social development. These effects are especially pronounced after repeated separations.

The damage from maternal separation experiments goes beyond psychological trauma. Physical damage, even death, was inflicted in scores of experiments. Rodents have borne the brunt of this treatment. They have been subjected to various surgical procedures and killed either during or after experiments. Primates suffered the same fate, but to a lesser extent. And primates, cats, and rodents have died from the separation experience alone.

Animals subjected to infant-infant separation, vertical chamber confinement, and cold-surrogate exposure also suffered greatly. Primates were the unwilling participants in these experiments. Those subjected to infant-infant separation seemed as seriously affected as infants separated from their mothers. Those subjected to vertical chamber confinement were psychologically crippled. Monkey infants reared with cold surrogate mothers were so devastated that some died.

Over 5,600 animals were subjected to these experiments. Rodents (4,847) and primates (579) were the workhorses of this research, with dogs (172) and cats (76) accounting for the balance. The last three figures are underestimates, as numerous studies did not report the numbers of animals used.

Financial Cost

Separation experiments cost over $34.0 million to conduct (Table IV). Most of this money was provided by the National Institutes of Health and the National Institute of Mental Health, as was the case with experiments in the deprivation category. Hence, U.S. taxpayers ultimately paid for most of this research.


Benefits to Humans

How have the results from separation experiments benefited humans? This is a fair question to ask given the clinical orientation of the majority of these experiments, not to mention the cost of the research to both animals and taxpayers.

Animal Model of Mother-Infant Separation

What have over two decades of animal experiments told us about mother-infant separation in humans? Judging from encouraging statements made by the animal modelers themselves (e.g. Brandt et al. 1972, p. 194; Harlow and Novak 1973, p. 464) one could easily assume that the animal studies have had considerable practical applications. But what does the evidence suggest?

Let us first consider the impact of the behavioral studies and then turn to the physiological studies. The behavioral studies apparently have failed to make any novel contribution to existing human data. The Wisconsin researchers themselves conceded this. After a strained attempt to establish the validity of their separation model, Suomi and Harlow (1977b) turned to the practical applications of the model: “We have a considerably more difficult time establishing a strong case [for practical applications], since so much monkey work to date has been based on existing human data and theories” (p. 173). Kraemer and McKinney (1979) reached the same conclusion: “it must be admitted that the primary flow of information, to date, has been from human clinical research into the animal modeling arena,” not vice versa (p. 38).

A comparison of animal and human studies suggests that, in general, the factors that influence the response to maternal separation are similar in human and nonhuman infants (Mineka & Suomi 1978; Mineka 1982). Although these animal studies did not make a novel contribution to the human situation, the rough correspondence between the animal and human data does lend support to the human data (Hinde 1974, Ainsworth 1976).

Animal models of the physiological responses to maternal separation have not fared any better than the behavioral models. A recent extensive review of the physiological experiments (Reite & Short 1983) did not mention a single clinical impact.

Separation studies of cats, dogs, and rodents seem to be even further removed from the human situation than studies of nonhuman primates. However, a few of these studies do make explicit reference to human applications. None of these applications seems particularly enlightening. For example, Elliot and Scott (1961) thought that their results on emotional distress in puppies suggested that “acute crying and fretting in human infants observed in day nurseries and hospitals may have more complex causes than simple separation from mother. Such infants are likewise on strange ground . . .” (p. 21). This patently obvious consideration appears in the early clinical literature (e.g. Yarrow 1961).

Some of the dog separation studies, like many of those on primates, appear to be reinventing the wheel; they go little beyond existing human data and theories. For example, Scott (1971) developed a theory of “attachment formation” on the basis of studies of dogs and humans. He wrote:

Bowlby and his colleagues [studying humans] have already done a great deal to develop ways of avoiding traumatic separation experiences in . . . infancy. The theory which I have developed here does not lead to any striking differences in procedures, but rather strengthens the reasons for them (p. 241).

Further evidence of the meager clinical impact that animal models of separation have had on our understanding of mother-infant separation in humans comes from several recent books on child development and early-childhood issues (Reference Note, p. 69). The books do not even mention any of these animal studies. Only human separation studies are discussed.

Animal models of mother-infant separation not only have failed to have any clinical impact, they also appear to have limited potential to make clinical contributions. The animal modelers have suggested two major aims for current separation studies (Hinde & McGinnis 1977, Dolhinow 1980, Reite & Short 1983): (1) identifying factors that increase susceptibility to the adverse effects of separation, and (2) discovering ways to ameliorate the adverse effects. Consider the procedures already known to ameliorate the response to maternal separation. First, human mothers have the ability to explain to the child the temporary nature and importance of the separation. Second, substitute care can be provided. That the effects of a separation experience in human children can be ameliorated or eliminated by sensitive fostering is well established” (Hinde & McGinnis 1977, p. 209). Third, mothers can attempt to meet the child’s increased emotional needs at reunion. Yarrow (1964) discusses numerous other ameliorating procedures.

Reite and Short (1983) speculate that animal models of mother-infant separation may be used for testing drugs to ameliorate the emotional distress of separated infants.

The evidence indicates that animal models of mother-infant separation in humans have had no clinical impact. At best, some of these experiments provided additional support of conclusions drawn from human studies. At worst, most of these experiments were trivial and redundant.

Separation Models of Human Depression

How have the prevention, diagnosis, treatment, and understanding of human depression benefited from animal studies of mother-infant separation, infant-infant separation, vertical chamber confinement, and other procedures?

One of the easiest ways to determine the clinical impact of animal models is to read the research reports of the animal modelers, who would not miss an opportunity to tout the importance of their work. None of the separation reports mentions any clinical impact of these studies.

Kraemer, Ebert, and McKinney (1983) suggested that their neurobiological studies of separation-induced despair have the potential to lead to clinical advances. They suggest that amine levels in cerebrospinal fluid could be used to identify individuals who are particularly vulnerable to separation-induced depression. These individuals could then be treated during separation to prevent depression. CSF amine levels may also be used to identify neurobiological targets for anti-depression drug therapy.

Time will tell whether or not these suggestions are correct; they don’t appear to be practical. It is difficult to imagine people seeking preventive therapy before they experience a disruptive separation. Moreover, CSF amine-directed drug therapy appears to be a red herring. Any of a number of pre-existing anti-depressants could be prescribed, if drug therapy is deemed necessary.

Separation Model of Anxiety

Suomi’s experiments on anxiety are too recent to have led to any clinical advances. Recall that his monkeys have temperamental differences that are apparent in stressful situations. These behavioral differences among individuals, and their associated physiological differences, are apparent early in development and are stable. In addition, they appear to be genetically based.

This research is highly regarded by the National Institute of Mental Health, which recently chose Suomi to head a new primate center (Cordes 1984). Moreover, this research apparently inspired parallel research on human infants that is currently underway (see Suomi 1986). What remains to be seen is whether or not these animal studies lead to clinical advances, and if so, whether or not these advances required animal studies.

Quality and Duplication

So far, separation experiments have a worse record on clinical application than do deprivation experiments. There is no suggestion that separation experiments on animals have had any clinical impact, while one deprivation experiment has had limited clinical impact.

The lack of clinical applications from separation studies would not be so alarming if few such studies were conducted. However, at least one hundred fifty scientific articles are devoted to these studies. Of these, about one hundred and ten correspond to separate studies. Most of these studies involved maternal separations in primates (Table III).

Why have so many experiments yielded so little?

One obvious reason is that the aim of many mother-infant separation experiments was to see if human responses, or factors that influence these responses, also occurred in this or that animal species. This was candidly acknowledged by Chappell and Meier (1975) who conducted a primate separation study themselves.

Another reason for the meager contribution of separation models is procedural variation from study to study. Studies differ in factors such as infants’ age at separation, degree of force used in the separation procedure, and degree of isolation imposed by the separation environment. These procedural variations can inadvertently affect the outcome of experiments. Moreover, if studies involve different animal species, then their outcomes can be affected by true species differences. While some of these variables are of interest in themselves, their effects are difficult to gauge when comparing studies that differ in more than one variable.

The problem of uncontrolled variation from study to study was the basis for Chappell and Meier’s (1975) pointed criticism:

Whereas these findings [from animal studies] are consistent, little resolution of the various issues central to maternal deprivation via separation can be gained until methodical variations are resolved. Variations in conditions in the supposedly controlled experimental work are as myriad as those in the naturally occurring experiments (p. 644).


The problems created by variations in design not only prevent comparisons across studies (only the grossest comparisons and conclusions can be drawn in this manner, i.e., that separation does result in agitation, at least, and depression if the period is long enough), but within a single study the findings are unclear (p. 645).

This criticism strikes at the heart of these experiments. Recall that the animal modelers originally exaggerated the inadequacies in human studies and touted the potential of their “well-controlled” experiments. Because of the unsystematic variation across studies, Mineka (1982) believes that the importance of variables such as infants’ age and sex has yet to be adequately investigated. This is a sad commentary given that scores of separation experiments have already been conducted.

What accounts for the lack of clinical applications of the various primate models of depression, namely, mother-infant separation, infant-infant separation, vertical chamber confinement, and cold surrogate exposure? One important factor is that virtually all of these experiments involve infants. Although separation models involving infants are thought to be scientifically valid models of human depression (Willner 1984), there are at least three reasons why infants are not good subjects for understanding human depression. One is that human depression is most prevalent in adults, not infants (Suomi & Harlow 1977b). Second, anaclitic depression (the despair reaction to maternal separation) in human children is not representative of adult depression (human data cited in McKinney 1977a and Suomi & Harlow 1977b). If this is true, then certainly anaclitic depression in monkeys is not representative of adult depression in humans. Third, an infant’s reaction to maternal separation is complicated by two factors (McKinney & Bunney 1969, Hofer 1978). One is the infant’s developmental progress, such as its increasing independence from its mother. The other complicating factor is the physiological as well as the psychological dependence of infants on their mothers. In fact, primate infants are so dependent on their mother that maternal separation can be a life-threatening event.

These shortcomings of using infants were apparent to the Wisconsin researchers themselves (Suomi & Harlow 1977b, McKinney & Bunney 1969, McKinney 1977a). Yet they justified their research at least partly—and in some cases wholly—as models of human depression. Consider Mineka’s (1982) justification: “depression in primates may best model childhood depression, which in turn may be thought of as a prototype for adult depression” (1982, p. 203). This view has two major problems. First, it ignores the human data that suggests anaclitic depression is not representative of adult depression. Second, it views the primate studies as models of models —twice removed from human adult depression.

The first of the maternal deprivation techniques to be explicitly viewed as a model of depression was maternal separation. However, the use of this model was challenged by Lewis et al. (1976) on the basis of unpromising results. “It becomes clear that mother-infant separation in rhesus monkeys does not fulfill the requirements of predictability and consistency, which are essential to establish a behaviorally-induced model of human depression” (p. 704, emphasis added). This criticism was not mentioned in Mineka’s (1982) extensive review of primate separation models of depression.

Lewis et al. advocated infant-infant separation as an alternative model of depression. However, they noted that researchers hesitated to use peer separation as a model of depression. According to Lewis et al., this hesitation might be related to the “absence of analogous human deprivations .” (Lewis et al. 1976, p. 705). If so, the hesitation seems justified. The use of peer separations violates one of McKinney and Bunney’s (1969) criteria for an animal model to be valid. Specifically, the procedure for inducing a syndrome in the model species should resemble the cause of the syndrome in humans. No one would argue that infant-infant separation, coupled with maternal deprivation, is a major cause of human depression.

Aside from maternal separation and peer separation, at least two other techniques were considered to be models of depression. One was confinement in vertical chambers and the other, exposure to cold surrogate mothers. Both were used exclusively at Wisconsin and were dismal failures.

The vertical chamber was initially heralded as having “enormous potential” as a model for human depression (Harlow & Suomi 1971b). Yet McKinney et al. (1975) later conceded that “there is considerable question as to what the behavioral changes induced by vertical chamber confinement represent” (p. 319). The vertical chamber was not even mentioned in Mineka’s (1982) extensive review of primate depression studies.

Mineka’s review also does not list exposure to cold surrogates as a model of depression. Yet this lethal technique was initially heralded as a model of “extreme depression” (Harlow et al. 1973). Harlow and company apparently were the only researchers to consider it as such. Actually, Harlow originally thought this procedure would be an animal model of schizophrenia (Harlow, in Tavris 1973), but that also was wishful thinking.

The Wisconsin researchers applied clinical labels so carelessly to their animal models that even isolation chamber confinement was eventually labeled a model of depression (Harlow 1974). Alas, this was also wishful thinking.

One of the most unsettling criticisms of primate models of depression is that depression in monkeys lacks salient cognitive features of depression in humans. These cognitive features include negative thoughts about oneself, the world, and the future.

Many researchers have questioned the usefulness of primate models of depression partly because, they argue, many of the cognitive symptoms which have been claimed to be cardinal symptoms in humans may either not exist in primates, or at a minimum may be very difficult to model in primates (Mineka 1982, p. 201).

Mineka, a Wisconsin researcher, gets around this criticism by arguing that these cognitive features are not cardinal symptoms of depression, and therefore are not needed to diagnose depression.

Researchers have other grounds for doubting that separation in monkeys is even a model of human depression, let alone a good model (e.g. Engel & Schmale, cited in Kaufmann 1977). Ainsworth (1976), for example, took issue with labeling the despair following separation as “depression.” She observed that the use of this label “implies that [the despair response] is akin to depressive illness in adults.. . . There is nothing essentially pathological in grief and mourning” following loss of a loved one (p. 41). This potentially lethal criticism remains unanswered. It and other criticisms go unheeded as researchers continue to tout the applicability of their work to human depression.


Most of the animal separation studies concerned the response to separation and the factors that influenced this response. The irony in this is that the “alternative,” human studies, preceded the animal studies, as was the case with studies of maternal deprivation per se (Ch. 4).

Not all separation studies were attempted replications of human research. Two noteworthy examples are the recent studies by McKinney’s group and Suomi’s group. If both lines of research prove to be valuable, their value will be in engendering parallel human studies. Hence, the animal studies would be facilitative, not essential.

Hinde’s group at Cambridge showed that primate separation studies could be conclusive without being excessively restrictive. Monkey infants could be housed with species members in large outdoor enclosures, not in tiny cages with or without their mother. Monkey infants did not have to be confined in vertical chambers to show the importance of the separation environment. They did not have to be separated up to 28 times to show the effects of more than one separation.


Separation experiments involved separating infants from their mothers or peers after the infants had developed attachment bonds. Related procedures involved incarcerating infants in vertical chambers or exposing them to cold surrogate mothers. These studies were animal models of mother-infant separation, depression, and anxiety.

Over 5,600 animals, mostly primates and rodents, were subjected to these experiments. The primates suffered acute distress and despair, with signs of chronic anxiety. Some were killed. The rodents suffered more physical damage; many were killed.

These experiments cost over $34 million, supplied primarily by federal institutions.

Although over 100 separation experiments have been conducted, none has yet had an impact on clinical practice. Many experiments were trivial extensions of past research, including human research, or were misconceived or misdesigned. Moreover, alternatives to many of these experiments were either already in use or could have been devised.


Please cite as:

Stephens, Martin L. (1986) Maternal Deprivation Experiments in Psychology: A Critique of Animal Models.
Retrieved from