Same average, more risk. An Evidence Press briefing. People learn from one another in two broad ways. Sometimes one long encounter, such as an apprenticeship, passes on many things at once. Sometimes knowledge travels piece by piece, through many brief contacts. Does it matter which, if every single piece of knowledge is passed on at the same average rate either way? This paper answers exactly, in a standard mathematical model of how a population builds up cultural traits: people invent things, learn from one another and die. It holds fixed the rate at which each trait is passed on, and changes only how bundled the learning is. The averages do not change. The expected number of traits, and the expected number known to any given number of people, are exactly the same. But the risks do change. Traits that travel through the same encounters share their fortunes: they spread together, and they are lost together when their few carriers die. So across the possible histories of a population, the number of traits becomes more spread out around the same average. Bundled learning makes big shortfalls more likely, and makes losing everything more likely. The same is true when forgetting comes in bundles, and, with one extra assumption, in the discrete-generation model where Wakano and Aoki asked this question in twenty twenty-four. How big is the effect? Where learning only just keeps pace with loss, it is modest. Where learning outpaces loss, it is dramatic. In a population of one hundred, at the setting of the original simulations, copying ninety-three per cent of a teacher's traits at each encounter, instead of all of them, leaves only a third of the variability. There are limits. This is a neutral, idealised model with no data. The chance of falling below one particular threshold is not always higher, and total loss is very rare in large populations. Telling how bundled real learning is would need many comparable populations. This is an anonymous, AI-assisted candidate. It has not been peer reviewed. Its paper and evidence are linked from the Evidence Press release page. This briefing uses an AI-generated voice and is not additional scientific evidence.