Thanks, Ted, for your question.

How people judge a person's age from a headshot is not clear. How the neural network that is called the brain makes these associations is a black-box type of process.

However, when some (bio)marker correlates with calendar age, then it is potentially useful as a biomarker of biological age. Imagine a scatter diagram with calendar age as a function of age estimates. The statistical significance of the correlation is in the slope of the mean that runs through the cloud of data points (using a least-squares method of fitting). An individual's distance to this mean is the distance between biological and calendar age. In reality, the algorithmic approach is a little more complicated than what I described, but that's it in essence.

So, with that background information of the 2009 study, Belsky and colleagues simply asked the question (among others) whether the biomarker "subjective facial age estimate" and the aging biomarker that is their bioage clock correlate. Which was the case. That's how the 2 parts of my post's story connect. Maybe I could have expressed this a bit clearer