We present optical , infrared and radio data of the z = 4.11 radio galaxy TN J1338 - 1942 including an intermediate resolution spectrum obtained with FORS1 on the VLT Antu telescope . TN J1338 - 1942 was the first z > 4 radio galaxy to be discovered in the southern hemisphere and is one of the most luminous Ly \alpha objects in its class . The Ly \alpha and rest–frame optical emission appear co–spatial with the brightest radio hotspot of this very asymmetric radio source , suggesting extremely strong interaction with dense ambient clouds . The Ly \alpha is spatially extended by \sim 4″ ( 30 kpc ) , has an enormous rest–frame equivalent width , W _ { \lambda } ^ { rest } = 210 \pm 50 Å , and has a spectral profile that is very asymmetric with a deficit towards the blue . We interpret this blue-ward asymmetry as being due to absorption of the Ly \alpha photons by cold gas in a turbulent halo surrounding the radio galaxy and show that the required neutral hydrogen column density must be in the range 3.5 - 13 \times 10 ^ { 19 } cm ^ { -2 } . The two-dimensional spectrum indicates that the extent of the absorbing gas is comparable ( or even larger ) than the 4″ ( 30 kpc ) Ly \alpha emitting region . The VLT observations are sufficiently sensitive to detect the continuum flux both blue-ward and red-ward of the Ly \alpha emission , allowing us to measure the Ly \alpha forest continuum break ( Ly \alpha ’ discontinuity ’ , D _ { A } ) and the Lyman limit . We measure a D _ { A } = 0.37 \pm 0.1 , which is \sim 0.2 lower than the values found for quasars at this redshift . We interpret this difference as possibly due to a bias towards large D _ { A } introduced in high–redshift quasar samples that are selected on the basis of specific optical colors . If such a bias would exist in optically selected quasars , – and even in samples of Lyman break galaxies – , then the space density of both classes of object will be underestimated . Furthermore , the average H i column density along cosmological lines of sight as determined using quasar absorption lines would be overestimated . Because of their radio-based selection , we argue that z > 4 radio galaxies are excellent objects for investigating D _ { A } statistics .