The work contained in this thesis is concerned with the design and construction of a near-infrared spectrometer for astronomical applications, Cooled Grating Spectrometer 4 (CGS4). The results of two observational projects carried out with the instrument on the United Kingdom Infrared Telescope are presented. CGS4 is one of the first infrared spectrometers to incorporate a 2D detector array, and the design of the spectrometer is driven by the desire to maximise the improvements in sensitivity which can be obtained with such an array. The need for high throughput and good image quality is discussed, and the way in which they have been achieved outlined. Other factors which affect the sensitivity of the instrument are the sky-background emission, thermal background from the instrument and telescope and detector noise. The method by which each of these is minimised is outlined. CGS4 was designed as an instrument which could be applied to the disparate projects which benefit from observations in the NIR waveband range. These include observations of molecules in star-forming regions, studies of emission lines from active galactic nuclei, and studies of gas dynamics. Two aspects of optimising observations in the NIR are discussed in detail. For observations in the 1-2,3μm region, emission from hydroxyl in the Earth's atmosphere dominates the sky background. The intensity of the line emission from OH varies by ~10% on timescales of ~10mins. As a result, on-source exposure times should be restricted to 60s, setting a fundamental limit to the sensitivity achievable for observations of extended sources. Flat-fielding a 2D array spectrometer, to calibrate the relative gains of the pixels, is another area which is explored in this work. The reasons for excluding the methods used for optical astronomy and NIR imaging are discussed. The solution for CGS4 was to provide a tailor-made 'calibration unit'. The design of this unit and the resulting success in flat-fielding observations is detailed.