The classification of colours into “alarming and serious” as opposed to “reassuring and playful” is applied for experimentation on architectural model perception. The distinction was corroborated by studies on physiognomic perception (Bonaiuto, 1978), on fairy-tale illustrations (Biasi & Bonaiuto, 2006), and in the “drawing recalling” of personal stress or comfort situations (Biasi & Bonaiuto, 1991, 1997). The colours of the first category mainly include black, grey, purple and olive green; while the opposite colours are mainly pink, orange, light green, sky blue and other pastel hues. Another basic concept is that conflict overload, which triggers “negative” emotions, can lead to active perceptual defence against other conflicts such as architectural incongruities. In comfort situations, which trigger “positive” emotions, the perception of the same incongruities may instead be facilitated and emphasised. Neutral hues lead to intermediate evaluations. Experiments were thus coherently designed in which models of buildings leaning seven degrees from the vertical were painted (or printed on light cardboards) with “alarming” colourings (grey-purple, olive green) or “reassuring” ones (bright pink, orange), or with intermediate shades. In the first case, the conflict overload, due to both the incongruous position and to the “alarming” colouring, reduces the perception of the inclination, while in the second case the architectural anomaly is accepted and significantly emphasised. The research paradigm has been applied with parallel criteria and results also to different architectural incongruities: such as the silhouette of a bayonet shape building, presenting a horizontal fracture line halfway up its body, with partial sliding of the upper part forward, over the lower part. Again, seven-box comparison scales were used to evaluate the degree of apparent sliding. With the same colourings as before, the working hypotheses were again comfirmed: the anomaly was underestimated (perceptual defence) or overestimated (emphatization) depending on the alarming or reassuring appearance. Neutral hues lead to intemediate evaluations. Appropriate comparison scales were used (“limits method”). Control situations were also prepared and studied, for which there is no rigid mental schema with regard to structural qualities (a crane, a cross-level bar). With the latter models, none of the aforesaid defence or emphasising effects, as functions of colours, were found.