Attachment is the pull that sends a scared toddler running to one particular person. The bond between mothers and babies is far older than humans: it shows up in virtually all mammals. What may be special about humans is how many people help. One hypothesis says our ancestors raised children with help from fathers, other relatives and other group members, and that babies evolved to win all of them over. But nobody can watch attachment evolve. No fossil records a cry or a cuddle. What we know comes from comparing animals and living families, and from a theory John Bowlby and Mary Ainsworth developed from the 1950s on. Parts of it, especially what counts as “normal” across cultures, are still disputed.
How to read the labels on this page: [Documented] means the evidence shows it directly. [Strongly supported] means strong sources agree. [Plausible] means the evidence allows it but doesn’t prove it. [Disputed] means qualified experts disagree. [Unknown] means the evidence can’t answer the question.
What do scientists mean by “attachment”?
In 1958, the psychiatrist and psychoanalyst John Bowlby (1907–1990) proposed that a baby’s tie to its mother is built from several inborn behaviors, including sucking, clinging, following, crying and smiling. Over the first year they come together and focus on a mother figure, especially in the second six months (Bretherton, 1992, p. 762). [Documented as what Bowlby proposed]
Mary Salter Ainsworth (1913–1999) added the idea of a secure base: a trusted person a child explores away from and comes back to. The phrase already appears in her University of Toronto doctoral dissertation, published in 1940 (Bretherton, pp. 759–760).
Analogy: a secure base works like base camp on a mountain. Climbers go farther because camp is there, and they head straight back when the weather turns.
How old is the bond between mothers and babies?
Very old. A 2016 review of the brain research calls mother–infant bonding “a characteristic of virtually all mammals” and the most common lasting bond mammals form (Numan & Young, 2016, Abstract and Introduction). In most mammals, hormones around birth, including oxytocin, help switch on a mother’s caregiving. [Strongly supported]
The bond isn’t built the same way in every species. A new rat mother will care for any rat pup. A ewe learns her own lamb’s smell in the first hours after birth, then cares only for that lamb and turns other lambs away (Numan & Young, maternal-behavior and sheep sections). [Documented in laboratory studies] This review studies mostly the mother’s side of the bond. When did the first such bond appear? [Unknown] Fossils don’t preserve relationships.
What do other primates show?
In more than 30% of primate species, someone other than the mother helps care for babies. Newborn langur monkeys can spend up to half of their first day being held and carried by females who aren’t their mothers (Hrdy & Burkart, 2020, sections 1 and 2b). [Strongly supported]
Our closest relatives go the other way. Chimpanzee, gorilla and orangutan mothers keep newborns in constant skin-to-skin contact, and Hrdy and Burkart call great ape mothers “exceedingly possessive.” Among primates, heavy help with feeding the young from group members other than parents is known only in humans and in marmosets and tamarins, small monkeys of the Americas (Hrdy & Burkart, sections 1 and 2b). [Strongly supported] That contrast is part of why Hrdy and Burkart think shared care is something new in our branch of the ape family, not an inheritance. [Plausible]
Did early humans raise children together?
Maybe. Anthropologist Sarah Blaffer Hrdy and Judith Burkart of the University of Zurich argue that apes with our slow-growing, big-brained, costly children “could not have evolved” unless others helped mothers. They call this the Cooperative Breeding Hypothesis. Their “conservative” estimate is that mothers relied on helpers by about 2 million years ago, in the time of Homo erectus. Other researchers place it earlier. The authors call their own reconstruction “inevitably speculative” (Hrdy & Burkart, Abstract and section 1). [Plausible as a hypothesis; its timing is Disputed]
In their model, babies best at winning over many caregivers survived more often, so over generations infants got better at reading other people. The model comes from comparing living apes, monkeys and humans, not from fossils. [Plausible]
Living people show how flexible this can be. One study of Aka forager families in Central Africa reported an average of six attachment figures per child (Meehan & Hawks, 2013, as reported by Hrdy & Burkart, section 2b). [Plausible: one study, and we read it secondhand] Are human babies more ready than other apes to attach to many people? No one knows, say Hrdy and Burkart: other ape babies almost never have caregivers besides their mothers. [Unknown]
Homo sapiens shows up about 300,000 years ago (Smithsonian Human Origins Program). Whether attachment had its present form by then is [Unknown].
How did scientists learn love isn’t just about milk?
In 1958, University of Wisconsin psychologist Harry Harlow summed up the common view of his day: basic needs such as hunger, thirst and pain were “primary drives,” and love was a “derived or secondary” one (Harlow, 1958). Bowlby rejected both that idea and the psychoanalytic view that love grows out of the pleasure of nursing (Bretherton, p. 762). [Documented]
Harlow tested it with infant rhesus monkeys. Each baby could reach two artificial “mothers,” one of bare wire and one covered in soft terry cloth. For four infants only the cloth mother gave milk, and for four others only the wire one did. Across 165 days of testing, the babies kept spending their time on the cloth mother, even when their milk came from the wire one. Harlow judged “contact comfort” of “overwhelming importance” to the bond, and nursing of “negligible importance” (Harlow, 1958). [Documented for these monkeys] The monkeys paid for this knowledge. Harlow’s paper records that his lab took infant monkeys from their mothers 6 to 12 hours after birth, more than 60 of them in the three years before these studies began.
Where did Ainsworth first watch attachment up close?
In Uganda. First, some background: in 1951, Bowlby wrote a World Health Organization report on homeless children. As Bretherton’s history reports, it centered a continuing relationship with a mother or mother-substitute, and placed fathers in a supporting role (Bretherton, 1992, p. 761; historical view). [Documented]
Ainsworth, born in Glendale, Ohio, and raised in Toronto (Marvin, 2000), left London in late 1953 for Kampala. Before leaving Uganda in 1955, she observed 26 Ganda families with unweaned babies. With an interpreter, she visited every two weeks for two hours, for up to nine months, in the living room where Ganda women usually hosted afternoon visitors (Bretherton, p. 764). Bretherton calls it the first empirical study of infant–mother attachment patterns; Ainsworth’s obituary says “one of the first” (Marvin, 2000). [Documented]
In Baltimore, starting in 1963, Ainsworth followed 26 more families, with about 72 hours of home visits each. She also used the Strange Situation, first described in 1969: a 20-minute, eight-part laboratory “miniature drama” in which a one-year-old meets a stranger and is twice briefly separated from the mother, once with the stranger and once alone (Bretherton, p. 765). Babies’ reunion behavior was sorted into patterns now called secure, avoidant and ambivalent. Mary Main later added “disorganized.” A 1978 book by Ainsworth and colleagues reported a “standard distribution” of 66% secure, 22% avoidant and 12% ambivalent (Keller, 2018, citing Ainsworth et al., 1978), drawn from 105 children, not only the 26 home-visit families (van IJzendoorn & Kroonenberg, 1988, Table 1). [Documented]
Is secure attachment the same everywhere?
This is the live argument. [Disputed]
A 1988 study pooled 1,990 Strange Situation results from 32 samples in eight countries. Secure was the most common pattern in all but one sample, and differences within countries were nearly 1.5 times as large as differences between them (van IJzendoorn & Kroonenberg, pp. 147–149). But look at who was studied. Eighteen of the 32 samples, and 1,230 of the 1,990 children, were from the United States. One more sample, listed under “China,” was Chinese-American families living in the U.S. The rest came from West Germany, Great Britain, the Netherlands, Sweden, Israel and Japan. None came from Africa, where Ainsworth’s Uganda study was done. The authors themselves called for data from Africa, South America and Eastern Europe (p. 149; Table 1, pp. 150–151; p. 154). [Documented]
In 2016, van IJzendoorn (co-author of the 1988 study) and two colleagues concluded that cross-cultural studies are “remarkably consistent with the theory” (Mesman et al., as quoted by Keller, 2018). Heidi Keller of Osnabrück University disagrees. She argues the theory reflects a Western middle-class model of child-rearing. She cites studies of Nso farming families in Cameroon, where most young children picked up by an unfamiliar woman showed no fear, and of villages in southern Madagascar, where young children, not adults, did almost all the social interacting with infants. She calls large programs pushing Western-style parenting on rural communities without knowledge of local culture unethical, and criticizes child-welfare agencies that judge migrant families by the same yardstick (Keller, 2018). By her account, Margaret Mead criticized Bowlby for ignoring cultural variation as early as 1954.
What about fathers? Bowlby’s 1951 report gave them a supporting role (Bretherton, p. 761). Hrdy and Burkart call fathers’ caregiving responses “facultative,” meaning they depend on circumstances, and Keller notes that in many farming villages grandmothers and children may matter more than fathers in infancy. Who does the caring varies a lot between societies. [Strongly supported]
How we know
What evidence exists: lab and field studies of animals, long home observations of families, the Strange Situation lab procedure, pooled studies, and a hypothesis tying them together. There’s no archaeological record of attachment. Bones can show that someone survived, not who comforted them or how anyone felt.
Whose voices are missing: babies can’t tell us anything in words. The Ganda mothers reach us through Ainsworth’s notes and an interpreter. Aka, Nso and Malagasy families reach us through outside researchers, and we found no published accounts by those communities themselves. In the WHO report, mothers and children are talked about, not heard.
How power shaped the record: In the 1988 pooled study, about 9 in 10 children lived in the United States or Western Europe (Table 1). That is the heart of Keller’s critique: a theory used worldwide to judge what is “normal” drew most of its early data from a few countries.
Then vs. now
- Then (1940s–1950s): love was a side effect of being fed. Now: comfort and contact matter in their own right; feeding is part of the bond, not all of it. [Strongly supported]
- Then (1951): one continuous mother relationship, with fathers in a support role. Now: children often attach to several people, such as fathers, grandparents and siblings (Bretherton, p. 770), and who those people are varies by community. [Strongly supported]
- Then: the evolution story starred a mother and a baby alone. Now: the Cooperative Breeding Hypothesis says many caregivers shaped human babies [Plausible], and its timing is [Disputed].
- Then (1978): a published classification and distribution of attachment patterns. Now: [Disputed]. Keller reports that some attachment researchers themselves see insecure patterns as possible adaptations to particular settings.
- Then and now: when attachment first evolved. [Unknown]