We analyze subway arrival times in the New York City subway system. We find regimes where the gaps between trains exhibit both (unitarily invariant) random matrix statistics and Poisson statistics. The departure from random matrix statistics is captured by the value of the Coulomb potential along the subway route. This departure becomes more pronounced as trains make more stops.The bus system in Cuernavaca, Mexico in the late 1990s has become a canonical physical system that is well-modeled by random matrix theory (RMT) [1][2][3][4]. This bus system has a built-in, yet naturally arising, mechanism to prevent buses from arriving in rapid succession. If a driver arrives at a stop just after another bus on the same route, there will be few fares to collect so the self-employed drivers introduced a scheme, using a cadre of observers along each route, to space themselves apart so as to maximize the number of fares they collect. Without this interaction, and mutual competition, one should expect that bus arrivals would be Poissonian [5]. While the New York City subway (MTA) system has a different, globally controlled, mechanism to space trains to eliminate collisions, much of the MTA system remains under manual control [6]. In this letter, we compare the predictions and results from Cuernavaca, Mexico with the MTA system.In particular, the authors in [1] noted that if one stood at bus stop in Cuernavaca, Mexico, near the city center, and recorded the set T of times between successive buses then for τ = T / Twhere · represents the sample mean and the function ρ(s) is known as the (β = 2) Wigner surmise (WS) [7]. This is the approximation of Eugene Wigner for the asymptotic (N → ∞) gap distribution for successive eigenvalues in the bulk of an N × N GUE (Gaussian Unitary Ensemble) matrix [8]. This is computed by considering the 2 × 2 case. This approximation of Wigner agrees surprisingly well with the true limiting distribution as N → ∞ [9]. The authors in [1] consider another statistic called the number variance. Fix a time T 0 and consider the time interval, [T 0 , T ], for T 0 ≤ T ≤ T 1 . Let n(T ) be the number of buses (or subway trains) that arrive in this time interval. Once one has made many statistically independent observations of n(T ), the number variance is computed byThis normalization is made so that n(T ) ≈ t. The aysmptotic prediction from RMT iswhere γ is the Euler constant [7]. This prediction is verified for the Cuernavaca bus system in [1]. A physicallymotivated model for the bus system was presented in [10] for which (1) and (3) hold.In this letter, we observe that (1) and (3) hold on a subset of the MTA system. We also find Poisson statistics within the MTA (which are also found in Puebla, Mexico [3]). For example, the southbound #1 train in northern Manhattan exhibits RMT statistics but the northbound #6 train exhibits Poisson statistics in the middle of its route. We also show that the train gap statistics tend to deviate more from RMT statistics as more stops are made. To quantitatively determin...