What survives
Attachment, which is the best-evidenced material in this series by a wide margin — reliably measured, cross-culturally replicated, predictive across generations, and moderately stable across a life. When this volume talks about anxious and avoidant patterns it is standing on sixty years of data.
What does not
Love languages, on all three of their central claims. And discrete attachment types, which the research field replaced with two continuous dimensions in 1998 — a person is not an anxious attacher, they have a position on an anxiety dimension that varies by relationship.
Ainsworth, Blehar, Waters and Wall (1978)
Patterns of Attachment
Tested Whether infant attachment behaviour falls into reliable classifications
Found Yes — secure, avoidant and resistant, reliably coded from a standardised procedure
Since Extensively, across cultures; van IJzendoorn and Kroonenberg 1988 covered 32 samples in 8 countries
van IJzendoorn (1995)
Psychological Bulletin 117(3) · 854 parent-infant pairs across 18 samples
Tested Whether a parent's own attachment representation predicts their infant's attachment classification
Found Combined effect size 1.06 for the secure-insecure split; 75 per cent correspondence between parent representation and infant security
Since Yes. This is one of the larger reliable effects in developmental psychology.
Fraley (2002)
Personality and Social Psychology Review 6(2), 123-151
Tested Stability of attachment from infancy to adulthood, meta-analysis with dynamic modelling
Found Moderate stability consistent with a prototype model: early representations persist and continue to exert influence, but are revisable
Since The prototype account is now the standard reading
Impett, Park and Muise (2024)
Current Directions in Psychological Science 33(2)
Tested The three central claims of the five love languages
Found People value all five rather than one; the five overlap substantially, which undermines their distinctness; and matching partners on preferred language does not reliably predict satisfaction. All three fail.
Since This is the field's synthesis of the small existing literature
What would change this rating
For love languages: a pre-registered dyadic study showing that matching on preferred language predicts satisfaction after controlling for how much affection is expressed at all. That confound has never been removed.
A rating that no possible finding could
alter is an opinion. Every volume names the result that
would move it.