在近几十年，神经科学家逐渐认同树突活动对可塑性和神经元运算至关重要。Antoine Madar指出，树突事件多种多样，可出现局部或全域电爆、不同空间范围、不同持续时间，足以让单一神经元具备与深度人工神经网路可比的运算能力。然而过去多数研究在脑切片（in vitro）中进行，神经元虽可被活化却脱离行为中的活体。Jeffrey Magee因此将研究带回真实行为的动物模型，并于2014年聚焦于海马回这一与情景记忆及place cell定位编码密切相关的区域。

在rodent于圆形跑道运行时，Magee团队观察到：某些place cell树突只产生一次plateau potential，该细胞便会在相同位置再度放电；实验诱发后，这一位置放电的命中率达99.5%。这与传统Hebbian模型形成冲突，后者预设编码放置场需重复动作电位共发。BTSP因此改变了以毫秒为核心的直觉：常规Hebbian通常依赖近同步活动，而树突plateau可持续数十到数百毫秒（有时接近一秒），并可增强发生于事件前后六至八秒的突触。这可解释真实行为学习通常以秒为尺度而非毫秒的矛盾。

其他研究者开始将BTSP视为有力的单次学习机制；一次探索即可让动物记住“食物在西北角、威胁在南方”等资讯，对只需一次机会就可能生死攸关的事件尤为关键。候选机制包括eligibility trace在数秒内标记最近活跃突触，接著由树突plateau引发的广域去极化只强化被标记突触。Anant Jain团队在2024年报告，多秒级生化级联可启动关键学习蛋白CaMKII，与树突上受体表面积与数量增加一致，进而放大后续神经传导。BTSP目前证据最强在海马回，对新皮质仅有初步支持，且同一区域并非所有细胞皆显示此效应；因而机制细节仍未定论。多数研究者认为其并非取代Hebbian，而是在成熟大脑记忆形成中与Hebbian并行，后者可能仍主导发育期的初始连结。

Over the past decades, neuroscientists have increasingly recognized dendritic activity as critical for plasticity and neuronal computation. Antoine Madar notes that dendritic events are highly diverse—local or global spikes, varying spatial extent, and variable duration—and can endow a single neuron with computational power comparable to a deep artificial neural network. Yet most evidence was initially obtained in brain slices, where neurons are alive but detached from behavior in a living animal. Jeffrey Magee therefore moved the question into behaving-animal recordings and focused from 2014 on the hippocampus, central to episodic memory and place-cell spatial coding.

In rodents running on a circular track, Magee’s team saw that when place-cell dendrites produced only a single plateau potential, those cells later fired in that same location; induced events drove firing at that location 99.5% of the time. This challenged the classic Hebbian expectation that place-field encoding requires repeated co-activation spikes. BTSP therefore reframes the timescale: conventional Hebbian plasticity is usually tied to near-synchronous activity, while dendritic plateaus can last from tens to hundreds of milliseconds, sometimes near one second, and can potentiate synapses active six to eight seconds before or after the event. This helps resolve a behavioral mismatch because real-world learning episodes commonly last seconds rather than milliseconds.

BTSP is increasingly viewed as a plausible one-shot learning mechanism, where a single exploration can encode “food in the northwest corner” or “danger in the south,” useful for events with only one chance to be learned. Proposed mechanisms include eligibility traces that tag recently active synapses for several seconds, followed by widespread dendritic depolarization that selectively strengthens tagged inputs. In 2024, Anant Jain’s group reported multi-second biochemical cascades ending in activation of CaMKII, consistent with increased dendritic receptor surface area and number, which can amplify future neurotransmitter binding. Evidence remains strongest in the hippocampus, with partial hints in neocortex and not all hippocampal cells expressing the effect. Most researchers treat BTSP as complementary to Hebbian plasticity rather than a replacement: Hebbian effects may dominate developmental wiring, while BTSP appears particularly relevant to memory formation in adults, though molecular mechanisms remain uncertain.