The issues of research into attachment.

by Research Methods & Statistics

Attachment is defined as the emotional bond between a baby and care-giver shown in their behaviour. There are two main explanations of attachment; behavioural theory and Bowlby’s evolutionary perspective. Learning theory focuses on the rewards provided by caregivers to infants in terms of food and warmth and relates to operant and classical conditioning. Bowlby’s evolutionary theory focuses on adaptation advantages of attachment in order to ensure that babies survive. The theories of attachment have been extensively researched however previous studies have relied upon non-human animals, self-report measures, retrospective studies and observations questioning the validity of research within this field.

Harlow and Harlow (1962)1 studied the formation of attachment in infant monkeys. Monkeys were placed into a cage with two wire mesh cylinders. One cylinder was bare and provided the monkey with milk. The second cylinder was covered in towelling acting as a form of comfort. The researchers investigated whether the monkeys would spend more time on the bare cylinder that provided milk (behavioural theory) or the towelled cylinder that provided a secure and comforting base (Bowlby’s theory). Researchers found that the monkeys spent more time on the towel cylinder and therefore concluded that infants need a responsive carer in order to form attachment as food alone is not sufficient. However, the validity of such research is flawed by using a non-human animal sample. Generalising the results from one species to another is extremely difficult; especially from animals to humans. Critics have argued that evolutionary and physiological similarities are not significant enough to justify generalizing results between animals and humans (Koestler, 1970)2. Throughout evolution each species has adapted to gain survival within its own niche by evolving its own unique sets of behaviours that will not be useful to other species. Specific to this experiment, the behaviour shown by monkeys may not generalise to human infants. Also, several ethical issues arise from this research as the methodology caused disturbance to infant monkeys by separating them from their mothers resulting in psychological distress for both the infant monkey and mother.

In other research Schaffer and Emerson (1964)3 reported 39% of infants were not securely attached to the person who brought them food thus contradicting the learning theory and supporting Bowbly. However, this research has low external validity as the sample were Glaswegian working class single mothers and therefore, we cannot be sure that results generalise to other cultures or cases where social economic status and family circumstances are different. Additionally, there is low internal validity as data collection relied on mothers recording information about their babies in diaries. The mothers were likely to be extremely preoccupied with child care duties therefore the data may not have been accurate.

An important aspect of Bowlby’s theory of attachment is the continuity hypothesis; he proposed if a child gained a secure attachment they would be able to form successful adult relationships. Hazan and Shaver (1987)4 found a relationship between early attachment styles and later attitudes towards romantic love. They devised a ‘love quiz’ that assessed participant’s attachment type and style of romantic love. Measure of attachment style was a simple adjective checklist of childhood relationships with parents and the measure of love and experience assessed individual’s beliefs of romantic love. The quiz was printed in a local newspaper and readers were asked to send in their responses. This methodology has many limitations. Firstly, the measure of attachment was collected retrospectively. Research into memory suggests that an individual’s recollection of past events is not reliable due to cognitive biases therefore; it seems unlikely that participant’s memory of childhood experiences will be accurate and this refers to low internal validity. Secondly, the data was collected through questionnaires. Self-report measures are subject to distortion and prone to demand characteristics. Participants may have not responded with truthful answers by presenting themselves in a desirable way rather than telling their true feelings. Research demonstrates participants respond more positively about themselves due to distorted perceptions about the self (John & Robins, 1994)5 and have a predisposition towards self-enhancement and maintaining positive attitudes about themselves at the cost of being unrealistic (Fiske & Taylor, 1999)6. Thirdly, the sample was biased as participants were volunteers. Only certain types of people read newspapers and complete and send in questionnaires therefore this research has low external validity as results may not generalise to the population. Finally, this research shows a correlation between child attachment style and adult relationship experience however causation cannot be shown. There has been a third variable suggested by Kagan (1984)7. He proposed the temperament hypothesis; children with a pleasant disposition are more likely to form positive relationships with parents and later in life, assuming they maintain their ‘niceness’, will form more loving relationships.

Ainsworth set up the Strange Situation (SS) (Ainsworth, Blehar, Waters & Wall, 1978)8; an experimental procedure in order to observe the variety of attachment forms exhibited between mothers and infants. The experiment was conducted in a small room with one way glass so the behaviour of the infant can be observed. The procedure goes as follows;

The researchers investigated four categories;

  1. Exploration – how much infants explored the unfamiliar room
  2. Separation behaviour – how much the infant reacted when the mother left
  3. Stranger anxiety – the response of the infant to the stranger
  4. Reunion behaviour – how the infant reacted when the mother returned

From the study, Ainsworth identified three types of attachment behaviour shown by the infants. The categories are summarised in the table below:-

Ainsworth’s conclusion that the SS can be used to identify the child’s attachment type has been criticized on the grounds that it lacks validity. In terms of lacking internal validity the mothers were likely to have behaved differently with their child in the study than they would have at home as they knew they were being observed, this refers to evaluation apprehension (a type of demand characteristic that frequently occurs in observational research). Secondly, the SS fails to acknowledge the child’s temperament and the effect this has on attachment style. Kagan proposed the temperamental hypothesis; temperaments are described as differences in babies that are inbuilt and visible from birth. Some infants are easy however, some are slow to warm up and this may affect attachment style. Ainsworth’s explanation for differences in attachment depended upon the sensitivity of the mother. However, a meta-analysis conducted by De Wolff et al9 found a weak correlation between sensitivity and attachment style. Additionally, the SS means different things to different children which again question the internal validity. Different childhood experiences prior to the experiment such as attending nursery, separation from the mother, multiple attachments and playing in different environments may affect how the child was classified. For example, a child who has attended day care may have shown signs of insecure avoidant attachment due to previous experience even though they are securely attached.

In terms of low external validity, the study only measures attachment style between the mother and the infant however; other research illustrates that children may have different types of attachment to fathers or grandparents (Lamb, 1977)10. Other research shows children can develop secure sibling attachments (Koluchova 1976)11. This means the SS lacks validity, as it is not measuring general attachment style, but instead an attachment style specific to the mother. Main & Weston (1981)12 found that children behaved differently in the SS depending on whether they were with their mother or father. Furthermore, the sample used in the study was biased as it only used 100 middle class American families. Therefore, it is difficult to generalize the findings outside of America and to working class families. It has been questioned whether the study measures attachment or the response to the SS. Critics have argued that the SS lacks ecological validity due to the unfamiliar nature of the playroom. Bronfenbrenner (1979)13 raised the issue of ecological validity as she found that attachment appeared stronger when the SS procedure was conducted in the family home (a natural environment).

Van Ijzendoorn et al14 conducted a meta-analysis of cultural variations in attachment using the SS. They used Ainsworth’s classifying system and found large differences between cultures. The results are shown in the graph below:-

However, Golding (2002)15 argues child rearing practices within cultures are huge demand characteristics in the research which significantly reduce internal validity. For example, Japanese babies are rarely separated from their mothers and this explains why they reacted violently with tears during separation leading them to be classed as insecure resistant when they are not. Also, babies brought up in Israel are rarely exposed to strangers which explains their violent protests to strangers in the SS. Furthermore, the sample of the meta analysis may not be representative as 27 studies were carried out in individualistic cultures (where independence is encouraged) and only 5 in collectivist cultures (a higher degree of interdependence).

The SS also has serious ethical implications. When testing separation behaviour and stranger anxiety the child was likely to suffer from severe ‘anxiety’ and psychological distress. As a result, anxiety was probably inflicted onto the mother when she saw her child in severe distress. It was even noted that the children classed as insecure resistant cried and screamed violently!

It seems that research into attachment, especially the SS is severely flawed by the validity of such methodologies. Current research lacks external validity due to studies using specific samples (in terms of culture, social economic status and species) and focusing on attachments between a baby and the mother (in the SS) making it extremely difficult to generalise the findings across the population. Research also lacks internal validity due to cognitive biases in retrospective studies and demand characteristics in self-report and observational methods. Specific to the SS a child’s rearing practices or experiences act as demand characteristics which lower the internal validity. Finally, when carrying out investigations within attachment researchers should adopt a sensitive approach to avoid causing any psychological distress to both human and non-human animal samples.


  5. John, O. P., & Robins, R. W. (1993). Determinants of interjudge agreement on
    personality traits: The Big Five domains, observability, evaluativeness, and
    the unique perspective of the self. Journal of Personality, 61, 521-551
  6. Fiske, S. T., & Taylor, S. E. (1991). Social cognition (2nd ed.). New York:
    McGraw Hill.