Screening Mother, Coding Baby: Videography & the Pre-History of Emotion Recognition in the Clinic

• In Event: Total Institutions: Technosciences of Care and Control IV

Abstract

For over a century, the tradition of psychologists and pediatricians performing infant observation has generated norms for diagnosis and screening of a variety of maternal and infant pathologies, along with corollary advice for parents on how to raise emotionally healthy babies. This art of “baby watching” changed radically with the advent of videotaping. Starting in the 1970s, this new medium and new evidential regime greatly enhanced the data-gathering capabilities of clinical psychologists’ clinics without removing bias.
The detail-oriented practice of coding the faces and behaviors of videotaped mothers and children did more than just render so-called invisible disabilities visible: new diagnostic categories and their criteria were developed out of this closer, pausable, and reviewable form of looking. Videography in the clinic allowed for the eventual coding and typing and pathologizing of “bad” mothers and “good” mothers, “sick” mothers and “well” mothers, “secure” infants and “anxious” infants (as in the studies of Ainsworth & Bell, Bebe, and Tronick).
“Screening Mother, Coding Baby” offers a pre-history of facial recognition, data coding, and diagnostic algorithms in the relationship between mother and child. I argue that the new diagnostic categories and screening capabilities resulting from clinical videography work in two ways. At their best, they still preserve definitional problems and offer diagnostic criteria so that caregivers can receive help—either for themselves or their babies. At their worst, they create the ideal image—literally—of mothering or infant states as well as stigmatized distances from that ideal. While beneficial for screening large numbers of mothers for, as just one example, postpartum depression, this way of looking at and recording mothers also creates an atmosphere of surveillance—mothers, once under care, are also at risk for social and state intervention, and unevenly so across intersectional lines. Even after the actual video equipment is put away, a visual comparative framework instructs that a good mother looks like x, not y. The sample populations of these videos are narrow: nearly always white, middle or upper class, neurotypical (a “normal” mother). As this surveillance paradigm has given way to the training of algorithms on mass facial recognition datasets, neuroimaging, and other high-tech solutions to the problem of screening mothers and their children, these formalized diagnostic criteria, and their pathological norming, have extended their reach, becoming a dragnet that further encodes racial and classed biases elaborated nearly fifty years ago.

Author

• Hannah Zeavin, UC Berkeley