If the state of polarization of a monochromatic light beam is changed in a cyclical manner, the beam acquires-in addition to the usual dynamic phase-a geometric phase. This geometric or Pancharatnam-Berry phase equals half the solid angle of the contour traced out on the Poincaré sphere. We show that such a geometric interpretation also exists for the Pancharatnam connection, the criterion according to which two beams with different polarization states are said to be in phase. This interpretation offers what is to our knowledge a new and intuitive method to calculate the geometric phase that accompanies non-cyclic polarization changes. © 2010 Optical Society of America OCIS codes: 350.1370, 260.5430, 260.6042. In 1984 Berry pointed out that a quantum system whose parameters are cyclically altered does not return to its original state but acquires, in addition to the usual dynamic phase, a so-called geometric phase [1]. It was soon realized that such a phase is not just restricted to quantum systems, but also occurs in contexts such as Foucault's pendulum [2]. Also the polarization phenomena described by Pancharatnam [3] represent one of its manifestations. The polarization properties of a monochromatic light beam can be represented by a point on the Poincaré sphere [4]. When, with the help of optical elements such as polarizers and retarders, the state of polarization is made to trace out a closed contour on the sphere, the beam acquires a geometric phase. This Pancharatnam-Berry phase, as it is nowadays called, is equal to half the solid angle of the contour subtended at the origin of the sphere [5][6][7][8][9][10].In this work we show that such a geometric relation also exists for the so-called Pancharatnam connection, the criterion according to which two beams with different polarization states are in phase, i.e., their superposition produces a maximal intensity. This relation can be extended to arbitrary (e.g., non-closed) paths on the Poincaré sphere and allows us to study how the phase builds up for such non-cyclic polarization changes. Our work offers a geometry-based alternative to the algebraic work presented in [11,12].The state of polarization of a monochromatic beam can be represented as a two-dimensional Jones vector [13] with respect to an orthonormal basis ͕ê 1 , ê 2 ͖ as E = cos ␣ê 1 + sin ␣ exp͑i͒ê 2 , ͑1͒with 0 Յ ␣ Յ /2, − Յ Յ , and ê i · ê j = ␦ ij ͑i , j =1,2͒. The angle ␣ is a measure of the relative amplitudes of the two components of the electric vector E, and the angle denotes their phase difference. Two different states of polarization, A and B, can hence be written asSince only relative phase differences are of concern, the overall phase of E A in Eq. (2) is taken to be zero. According to Pancharatnam's connection [5] these two states are in phase when their superposition yields a maximal intensity, i.e., whenreaches its greatest value, implying thatThese two conditions uniquely determine the phase ␥ AB , except when the states A and B are orthogonal.Let us now consider a sequence o...