Files in this Data Supplement:

• Supplemental Figure S1 -

  Fig. S1. Comparison of wing position during leg extension for two different escape take-offs. Fly 1 executes an escape in which the wings are not raised before the start of leg extension. The two different camera views (top and side) show that by mid-leg extension the wings are still closed along the back of the fly, a wing posture that the fly maintains even as its legs leave the substrate. Fly 2 executes an escape in which its wings were fully raised before the start of leg extension (these wing-open escapes are labeled **Esc. in Fig. 10). As the fly extends its legs, the wings flap down in a normal downstroke, similar to the wing motion during a voluntary take-off. The result is that the wings are fully outstretched during the leg extension period.

• Supplemental Figure S2 -

  Fig. S2. Alternative conventions for describing rotation. Any arbitrary attitude of the fly (orientation relative to the lab frame) can be uniquely described by a sequence of three rotations. In a rotation sequence common to aerodynamics, the transformation from the lab fixed coordinate frame to the fly body-centered coordinate frame is accomplished by a ZYX rotation sequence: first, the fly body is rotated by an angle ψ about the lab z_f-axis, then rotated by an angle θ about the new y₁-axis, and, finally, rotated by an angle φ about the x₂-axis, that is the x_f-axis after the first two rotations. The angles ψ, θ, and φ are commonly referred to as Euler angles, and we call them heading, elevation and bank, respectively. In order to uniquely describe an object’s attitude, we must restrict the range of the Euler angles as follows: −180°<φ<180°, −90°<θ<90° and 0°<ψ<360°. Thus while Euler angles are a useful scheme for defining the posture of an animal relative to the lab frame unambiguously, their use is limited when trying to describe large rotations, such as those observed during escape take-off, because singularities occur when the elevation angle reaches ±90° (a phenomenon often called ‘gimbal lock’). For this reason we have chosen to express the rotational kinematics of the fly as roll, pitch and yaw rotations around the orthogonal x_b-, y_b- and z_b-body axes, respectively (Figs 4, 5 and 7). Note that Euler angles do not have a one-to-one relationship with roll, pitch and yaw (see Phillips, 2004). Expressing the kinematics in the body-centered coordinate frame has the advantage that this is the coordinate frame in which the fly itself senses the world. For example, neurons in the lobula plate of blowflies respond to specific self-motion rotations about different body-centered axes (Krapp et al., 1998; Kern et al., 2001). However, for comparison with other work, which has used Euler angles schemes to describe flight kinematics, we present the bank (φ), elevation (θ) and heading (ψ) angles in a ZYX-rotation scheme for the same data as shown in Fig. 5B.

  (A) Both the individual time courses and mean (±s.e.m.) values for bank, elevation and heading. Blue traces correspond to voluntary take-offs and red to escape responses (see Fig. 5 for the number of flies averaged at each time point). These Euler angles are difficult to interpret because when the fly reaches an elevation of 90° (corresponding to the long body axis pointing straight up, perpendicular to the ground), the heading and bank angles are displaced by 180°. Thus a fly flying ‘upsidedown’ is represented by an elevation angle <90°, but heading and bank angles 180° different from their ‘right-side up’ values. (B) For comparison, the individual and mean (±s.e.m.) roll, pitch and yaw values for the same data as shown in A and Fig. 5B. The only difference with the data presented in Fig. 5B is that the initial ‘roll’ and ‘pitch’ orientations for each fly have not been added, so the angles are expressed relative to the fly’s starting posture rather than starting in the context of the lab frame. These ‘position’ values represent the cumulative amount of rotation about each of the body axes over time. The data in both A and B have been adjusted as in Fig. 5B so that initial banking/rolling and yawing/heading movements are to the right.

  Kern, R., Petereit, C. and Egelhaaf, M. (2001). Neural processing of naturalistic optic flow. J Neurosci 21, RC139.

  Krapp, H. G., Hengstenberg, B. and Hengstenberg, R. (1998). Dendritic structure and receptive-field organization of optic flow processing interneurons in the fly. J Neurophysiol 79, 1902-1917.

  Phillips, W. F. (2004). In Mechanics of Flight, pp. 867-890. Hoboken, NJ: John Wiley & Sons, Inc.

• Movie 1 -

  Movie S1. A high-speed movie of a voluntary take-off. The movie was filmed at 6000 frames s⁻¹ and is played back 100 times slower than real time.

• Movie 2 -

  Movie S2. A high-speed movie of an escape response. The movie was filmed at 6000 frames s⁻¹ and is played back 200 times slower than real time.

• Movie 3 -

  Movie S3. A high-speed movie of the first two wing stroke cycles of an escape response. The movie was filmed at 6000 frames s⁻¹ and is played back 3000 times slower than real time. See Fig. 4 for a full description.