There is growing interest in heat pumps based on materials that show thermal changes when phase transitions are driven by changes of electric, magnetic or stress field. Importantly, regeneration permits sinks and loads to be thermally separated by many times the changes of temperature that can arise in the materials themselves. However, performance and parameterization are compromised by net heat transfer between caloric working bodies and heat-transfer fluids. Here we show that this net transfer can be avoided-resulting in true, balanced regeneration-if one varies the applied electric field while an electrocaloric (EC) working body dumps heat on traversing a passive fluid regenerator. Our EC working body is represented by bulk PbSc 0.5 Ta 0.5 O 3 (PST) near its first-order ferroelectric phase transition, where we record directly measured adiabatic temperature changes of up to 2.2 K. Indirectly measured adiabatic temperature changes of similar magnitude were identified, unlike normal, from adiabatic measurements of polarization, at nearby starting passive regeneration, this net transfer has been noted explicitly in a seminal paper 1 , and elsewhere 22 .The key figure of merit for any cooling device is its coefficient of performance COP = Q/W, where in each cooling cycle, work W is done to pump heat Q from a cold load at T c towards a hot sink at T h . The effect of the net-transfer problem is to partially subsume regenerators into sinks or loads, with two consequences. First and foremost, it becomes impossible to realize intended cooling cycles, thus compromising device COPs. Second, the active material undergoes net heat exchange at temperatures lying away from T c and T h , thus compromising the calculation of the Carnot limit T c /(T h -T c ) from measured values of T c and T h . Given that COPs and Carnot limits are thus compromised, it immediately follows that device efficiencies are neither optimized nor accurate (efficiency = COP/Carnot). To solve the nettransfer problem, we will construct valid cooling cycles on detailed field-temperature-entropy maps, and we will accurately evaluate the resulting COPs, Carnot limits and efficiencies.These parameters contrast with COPs for materials in isothermal cycles 23 , and the related parameter of materials efficiency 24 that describes reversible caloric transitions driven isothermally in just one direction.The net-transfer problem has been hitherto masked in EC cooling cycles because the zerofield and finite-field isofields are rendered parallel by assuming peak performance at all points in any given cycle [25][26][27] . However, even for relaxors that show large EC effects over wide ranges of temperatures 6,20,21 , this assumption is only reasonable for small temperature spans T h -T c that would only be useful when exploiting converse EC effects for pyroelectric energy harvesting 28 . The net-transfer problem has also been masked in cooling cycles based on mC materials, either by likewise assuming peak performance 29 or by assuming isothermal conditions 23 . In t...