We present measurements of polarization lensing using the 150 GHz maps which include all data taken by the BICEP2 & Keck Array CMB polarization experiments up to and including the 2014 observing season ( BK14 ) . Despite their modest angular resolution ( \sim 0.5 ^ { \circ } ) , the excellent sensitivity ( \sim 3 \mu K-arcmin ) of these maps makes it possible to directly reconstruct the lensing potential using only information at larger angular scales ( \ell \leq 700 ) . From the auto-spectrum of the reconstructed potential we measure an amplitude of the spectrum to be A _ { L } ^ { \phi \phi } = 1.15 \pm 0.36 ( Planck \Lambda CDM prediction corresponds to A _ { L } ^ { \phi \phi } = 1 ) , and reject the no-lensing hypothesis at 5.8 \sigma , which is the highest significance achieved to date using an EB lensing estimator . Taking the cross-spectrum of the reconstructed potential with the Planck 2015 lensing map yields A _ { L } ^ { \phi \phi } = 1.13 \pm 0.20 . These direct measurements of A _ { L } ^ { \phi \phi } are consistent with the \Lambda CDM cosmology , and with that derived from the previously reported BK14 B -mode auto-spectrum ( A _ { L } ^ { BB } = 1.20 \pm 0.17 ) . We perform a series of null tests and consistency checks to show that these results are robust against systematics and are insensitive to analysis choices . These results unambiguously demonstrate that the B -modes previously reported by BICEP / Keck at intermediate angular scales ( 150 \lesssim \ell \lesssim 350 ) are dominated by gravitational lensing . The good agreement between the lensing amplitudes obtained from the lensing reconstruction and B -mode spectrum starts to place constraints on any alternative cosmological sources of B -modes at these angular scales .