She opened it. Carol’s voice, transcribed verbatim: “People think grief is a straight line. It’s not. It’s a knot. And SPSS can’t untie knots, Doctor. Only hearts can.”
The climax came on a Tuesday night—or was it Wednesday morning? The line had blurred. Alena decided to run a binary logistic regression to predict which caregivers would develop complicated grief. The dependent variable: Complicated_Grief_YN (1=Yes, 0=No). Predictors: age, years caregiving, cortisol AUC, and—her gamble—the interaction between fMRI_Activation_LeftInsula and a new dummy code for the inverted grief pattern.
She had named the trial file as a safeguard. A sandbox. But somewhere between the third cup of cold coffee and the 2:00 AM wall, the sandbox had become the real world.
SPSS suggested, in its quiet, algorithmic way, that she should exclude the case. “Listwise deletion,” the textbooks called it. A common practice. Just click the button. No one would know.
“Probably.”
He leaned back, tapping the sketch. “But you’ve just done something more important than a tidy p-value, Alena. You’ve proven that the trial—the trial of running the numbers, of testing the limits of the tool—is itself the method. SPSS is a hammer. But you’ve learned that not every problem is a nail.”
The next morning, she walked into Dr. Mbeki’s office and placed a printed draft on his desk. The first page was a graph—not a bar chart or a boxplot, but a hand-drawn sketch of a tangled loop, labeled Carol’s Grief . Underneath, in bold: “Significant at the level of lived experience. p = irreducible.”