Mistuning a harmonic increases its salience and produces an exaggerated change in its pitch. The effects on component grouping of spectral pattern, global pitch, and local harmonicity were explored using these phenomena. Stimuli were either harmonic (F0=200 Hz) or frequency shifted by 25% of F0. Component 1 or 2 was replaced by one of a set of sinusoidal probes in the same spectral region. Listeners either matched the probe pitch by adjusting the frequency of a pure tone (experiments 1 and 3) or matched the probe loudness by adjusting the level of a tone of identical frequency (experiment 2). Probe positions corresponding to greatest perceptual fusion were estimated from the variations in pitch shift and loudness across frequency. Both measures gave similar estimates. For harmonic stimuli, fusion was greatest at harmonic values. For shifted stimuli, fusion was greatest close to the suboctave (225 Hz) and the frequency (450 Hz) of component 2. The latter value moved downwards to near 433 Hz (2:3 ratio with component 3) when component 1 was removed. Together, these results indicate that the lowest component of a shifted complex is grouped by local harmonicity, whereas the higher components are grouped by common spectral spacing. Global pitch did not influence component grouping.