The observed deficit of strongly Lyman-α emitting galaxies at z > 6.5 is attributed to either increasing neutral hydrogen in the intergalactic medium (IGM) and/or to the evolving galaxy properties. To investigate this, we have performed very deep near-IR spectroscopy of z ≳7 galaxies using MOSFIRE on the Keck-I Telescope. We measure the Lyman-α fraction at z ∼ 8 using two methods. First, we derived NLy_/Ntot directly using extensive simulations to correct for incompleteness. Second, we used a Bayesian formalism (introduced by Treu et al. 2012) that compares the z > 7 galaxy spectra to models of the Lyman-α equivalent width (WLy_) distribution at z ∼ 6. We explored two simple evolutionary scenarios: pure number evolution where Lyman-α is blocked in some fraction of galaxies (perhaps due to the IGM being opaque along only some fraction of sightlines) and uniform dimming evolution where Lyman-α is attenuated in all galaxies by a constant factor (perhaps owing to processes from galaxy evolution or a slowly increasing IGM opacity). The Bayesian formalism places stronger constraints compared with the direct method. Combining our data with that in the literature we find that at z ∼ 8 the Lyman-α fraction has dropped by a factor >3 (84% confidence interval) using both the dimming and number evolution scenarios, compared to the z ∼ 6 values. Furthermore, we find a tentative “positive” Bayesian evidence favoring the number evolution scenario over dimming evolution, extending trends observed at z . 7 to higher redshift. Comparison of our results with theoretical models imply the IGM volume averaged neutral hydrogen fraction & 0.3 suggesting that we are likely witnessing reionization in progress at z ∼ 8.