Spatial audio reproduction systems aim to provide the psychoacoustic cues necessary to create a subjectively convincing portrayal of a sound field. Interaural time and level differences synthesised by ambisonic and amplitude-panned reproduction systems are simulated to assess the ability to accurately reproduce binaural cues. Two binaural models - an analytical sphere and measured HRTF data from the MIT KEMAR database - were used to obtain ITD and ILD cues for ambisonic and amplitude-panned reproduction with three loudspeaker layouts. By comparing results for central and off-centre listening positions, the relationship between listening area 'sweet spot' and binaural cues is studied. For central listening positions, large ITD errors are observed with an amplitude-panned ITU 5.0 layout, with first-order ambisonics showing improved performance. Third-order ambisonics is required to achieve improved ILD performance over the ITU 5.0 system. For off-centre listening, fourth-order ambisonics using an octagon layout is required to achieve improved ITD and ILD performance over the ITU 5.0 system. The analytical limitations of binaural cues are discussed, with suggestions for further work in the area.