7.1.6, 7.3.1 - Cross-cultural attachment meta-analysis
Attachment research asks whether infants organise relationships with caregivers in similar ways across the world. The classic meta-analysis by van IJzendoorn and Kroonenberg (1988) lets us examine both similarities and differences, while separating a country's average from the varied families within it.
Combining attachment evidence
A cross-cultural comparison needs comparable measurements before differences can be interpreted as differences in attachment.
Attachment is an enduring emotional relationship in which a child seeks closeness and security from a caregiver. Ainsworth's Strange Situation uses a short standardised sequence of play, separation and reunion to observe this relationship. Classification depends especially on reunion behaviour. Secure infants (B) use the caregiver as a secure base and can seek and accept comfort; avoidant infants (A) show little overt proximity seeking on reunion; resistant infants (C) seek contact but also resist it and are difficult to settle. These are patterns in a particular relationship and procedure, not permanent moral labels for children or parents.
Meta-analysis
A meta-analysis combines numerical findings from separate studies to examine a broader pattern than any single sample provides.
Van IJzendoorn and Kroonenberg asked how attachment classification distributions differed between samples, countries and regions, and whether differences within countries were smaller than those between them. They did not bring 1,990 children into a new laboratory. They searched the Lockheed computer files for attachment research and examined an existing multinational dataset, then selected published study samples.
To keep the comparisons meaningful, studies had to use the classical Strange Situation with infant–mother dyads and report A, B and C separately. Nonstandard procedures, other caregiver dyads and reports combining A and C into one insecure category were excluded. They also excluded special groups such as twins and children with Down syndrome, samples below 35, overlapping samples and studies of children older than 24 months. Removing overlaps reduced the risk of counting the same children twice; requiring the same procedure reduced one source of measurement inconsistency. These decisions also narrowed the population represented.
The final database contained 32 samples and 1,990 classifications, grouped under eight country/cultural labels. These included Britain, Germany, the Netherlands, Sweden, Israel, Japan, the United States and a Chinese sample living in the United States. Calling this simply eight national populations would conceal that last sample's setting. This was a sample of available studies, not random sampling of the world's infants.
Within and between countries
The main result concerns how much distributions varied, rather than a ranking of countries from good to bad parenting.
The researchers put the A, B and C frequencies into a contingency table. They compared observed frequencies with those expected from the overall distribution, partitioned variation into within-country and between-country components, and used correspondence analysis to identify similar sample profiles. Students need to understand these purposes; reproducing correspondence-analysis calculations is not required here.
Secure B was the most frequent classification in every aggregated country group. At the individual-sample level it was the most frequent in all but one sample, the Bielefeld sample from Germany. Compared with the global distribution, A was relatively more frequent and C less frequent in the four Western European country groups; the reverse pattern appeared in Israel and Japan. These are relative tendencies in the sampled data, not claims that all infants from either region have that pattern.
The particularly important result was greater variation within countries than between countries. The within-country contribution to total chi-squared variation was 146.2; the between-country contribution was 102.4. Their ratio is approximately 1.43, described by the authors as nearly 1.5 times. Do not reverse this result. A Japanese Tokyo sample resembled some US samples more closely than the Japanese Sapporo sample. The Israeli city sample also resembled a US sample more than the Israeli kibbutz sample.
| Comparison | What is being compared? | What it prevents us assuming |
|---|---|---|
| Within-country | Separate samples assigned to the same country group | One local sample represents everyone in that country |
| Between-country | Aggregated distributions across country groups | Every family difference is caused by nationality |
The US supplied 18 of the 32 samples and 1,230 classifications. Its many samples showed substantial internal variation, but their combined distribution lay close to the global distribution. Importantly, the authors found that combining the non-US samples also produced a distribution close to the US aggregate. US dominance therefore matters for representativeness, but does not by itself erase all evidence of similarity.
Continue this lesson.
You’re previewing NovaLearn without an account. Create a free account to read the rest.
Create free account