Fig. 5. Transitions between the four major classes of stationary patterns in Fig. 4 observed upon variations of model parameters. The patterns are classified according to the number of peaks in the spatial distribution of signaling components. (A-D) The concentration of Spitz in the center of the domain, s(x=0), is plotted as a function of a single model parameter, with all the other parameters fixed at the values corresponding to the transient in Fig. 3. The point corresponding to the ‘wild type’ steady pattern is denoted by ‘’. (A) An increase in the input amplitude, g₀, produces the 0121' sequence of transitions. Each transition is accompanied by a hysteresis: there is a region of inputs where qualitatively different patterns co-exist. This sequence of transitions is used to account for different phenotypes observed upon the variation in the level of Gurken, see Discussion. (B) A qualitatively similar sequence of hysteretic transitions is observed upon a variation of the strength of the positive feedback in the model. Experimentally, this sequence of transitions can be realized by uniform changes in the level of Spitz or Rhomboid. (C) A uniform decrease in the strength of the negative feedback by Argos is predicted to generate a hysteretic transition between the phenotypes with two and one broad appendages. (D) The existence of the two-peaked pattern is conditional on the separation of the length scales of Spitz and Argos. Decreasing the length scale of the diffusing inhibitor, corresponding to increasing in the model, leads to disappearance of the two-peaked patterns; this transition is also accompanied by a hysteresis.