Divalent thulium has been doped into CsCaI₃, CsCaBr₃, CsCaCl₃ and RbCaI₃- a series of AMX₃ salts. Using previously published optical spectra, a series of parameterised energy level calculations have been performed. The calculated energy levels, optimised crystal field parameters and simulated optical absorption spectra are presented. Theoretical predictions yield excellent approximation to the experimental data. Temperature dependent fluorescent lifetimes from the (³F₄,t₂g) and (³H₆,t₂g) excited (emitting) states have been measured using a pulsed dye laser. For CsCaBr₃ and RbCaI₃ doped with Tm²⁺, visible emission for the (³F₄,t₂g) state yields 10 K and 28 K lifetimes of 1.7 μs and 0.40 μs respectively. In both cases no emission is observed at room temperature. Considering direct multiphonon relaxation to the lower lying (³H₆,t₂g) levels, a simple energy gap law well accounts for the measured data with effective phonon energies in the range 100-200 cm⁻¹ - consistent with the phonon density of states in these low phonon energy hosts. Monitoring infrared emission from the (³H₆,t₂g) states, 14 K and 10 K lifetimes of 301 μs and 250 μs are found for CsCaBr₃ and CsCaCl₃ respectively. For CsCaBr₃ this value reduces to 270 μs at 200 K and is not quenched until 300 K, whilst for CsCaCl₃ emission is quenched by 170 K. This temperature dependent behavior is interpreted in terms of internal conversion via configurational crossing between the excited and ground state potential energy surfaces. Fitting the fluorescence lifetime data to a modified Mott equation, it is inferred that the potential barrier for non-radiative relaxation is five times larger in CsCaBr₃ compared to CsCaCl₃. This explains the fact that emission is still observable in the bromide host at room temperature.