Flicker noise is the most crucial issue in an instrumentation amplifier (IA) for neural recordings because low-frequency neural signals overlap with the frequency of the amplifier's flicker noise. A Chopping technique, often used to reduce the flicker noise, is not desirable for high-impedance input sources due to the charge injection and clock feedthrough from the MOSFETs of the input chopper to the signal source, resulting in a significant increase in the total input-referred noise. Whereas MOSFETs have less flicker noise at the moment of turning on, and the intrinsic flicker noise can be then reduced by turning on and off the MOSFETs in the IA. This brief proposes a bulk switching IA, which can avoid the input current noise. A prototype IA is implemented in a 65 nm CMOS occupying 0.053 $\hbox{mm}^{2}$, and it achieves the input-referred noise of 0.74 $\mu\hbox{V}_{\rm rms}$ (local field potential) for 100 k $\Omega$ source impedance, a 3.3 times reduction compared with that of the chopper IAs, while consuming only 3.96 $\mu\hbox{W}$ from a 1.2 V supply.