Paleomagnetic data collected in the last 30 years indicate that a simple orocline model is not sufficient to describe the complex evolution of the Calabrian Arc. Present-day curvature of the Calabrian Arc is the result of a different tectonic history between the edges of the arc, namely the Southern Apennines and Sicily, and its core, the Calabria-Peloritani Domain. These differences mirror the structural architecture and deep lithospheric configuration of the Calabrian Arc, which are related to the geometry and evolution of the Ionian subduction system. In particular, the presence of lateral heterogeneities in the subducting lithosphere and the subsequent progressive decrease in width of the trench during subduction are likely the main causes of Calabrian Arc formation and of its present-day narrow tight shape.
 Since the pioneering work of Carey  on the concept of oroclines, the origin of curvature in mountain belts has become an important topic in tectonic and geodynamic research. Although in its original meaning, the term “orocline” was intended as “an orogenic system which has been flexed in plan to a horse-shoe shape” [Carey, 1955], through time it has been used to describe any orogen of arcuate shape, regardless of its deformation history (for recent reviews on this topic, see Marshak  and Weil and Sussman ). The improper use of the term orocline mainly derives from difficulties in restoring the original shape of arcuate mountains belts, i.e., in defining the primary or secondary nature of the arc, using classical structural and geologic methods. Paleomagnetism has been increasingly used to reconstruct the tectonic histories of curved mountain belts and to define the timing of the bending by its ability to determine the distribution and magnitude of vertical axis rotations. A fundamental step toward understanding orogenic curvature is documenting the kinematics and timing of the curvature in relation to tectonic elements. Depending on the relationship between the timing of thrusting and vertical axis rotations, salients [see Miser, 1932] can be distinguished as (1) primary curves, where no vertical axis rotations are required; (2) secondary curves, where vertical axis rotations postdate thrusting; and (3) progressive curves, where rotation and thrusting occur simultaneously (among others, Carey , Elliott , Hindle and Burkhard , Marshak , Sanderson , Sussman et al. , Weil and Sussman , and Wilkerson ). In some cases, deformation can also produce a “secondary” curve superposed on a “primary” curve, as documented in the Caledonides where the deformation wrapped accreted units around a preexisting promontory [Mac Niocaill et al., 1998]. This classification of arcuate belts provides two end-members, which can be distinguished by comparing paleomagnetic declinations with structural data. In primary arcs, paleomagnetic declinations remain parallel along the arc and do not correlate with changes in thrust and fold-axis trend. In secondary arcs, paleomagnetic declinations change direction along the arc and follow changes in thrust and fold-axis trend with a one-to-one correlation. In the last few decades, increasing evidence has been found that these two end-members cannot describe the kinematic evolutions of most curved orogens. In fact, detailed paleomagnetic investigations from well-dated syntectonic sediments show that in several orogens, vertical axis rotations occurred neither before nor after, but during thrust activity (among others, Allerton et al. , Barke et al. , Gray and Stamatakos , Mattei et al. , Pueyo et al. , and Sussman et al. ). Furthermore, information from thrust tectonics and stratigraphy of syntectonic sediments shows that fold-thrust belts evolve over a long time span, with complexities in the evolution of deformation [e.g., Elliott, 1976].
 Since the formulation of plate tectonic theory, subduction processes have been the most frequently evoked tectonic cause for the origin of orogenic curvature. Since the first hypothesis that attributed the curvature of arcs in subduction zones to the Earth's sphericity [Frank, 1968], several hypotheses have been proposed that describe the relations between subduction and orogenic arcs. These hypotheses include the role of mantle flow, such as return flow induced by retreat of the trench and rollback of the slab, deformation of the subducting slab [Dvorkin et al., 1993; Garfunkel et al., 1986; Russo and Silver, 1994], and the role of internal heterogeneities within the subducting lithosphere that induce lateral changes in the buoyancy of the lithosphere or its mechanical strength [McKenzie, 1969; Stein and Stein, 1992]. These heterogeneities may derive from along-strike variations in the age of the subducting oceanic lithosphere [Molnar and Atwater, 1978], or the presence of buoyant features such as oceanic plateaus, seamount chains, or continental fragments [Hsui and Youngquist, 1985; McCabe, 1984; McCabe and Uyeda, 1983; Nur and Ben-Avraham, 1982; Vogt et al., 1976]. Secondary induced mantle flow and internal heterogeneities within the subducting lithosphere have been investigated by analog and numerical experiments [Funiciello et al., 2003; Kincaid and Griffiths, 2003; Morra et al., 2006; Schellart et al., 2007], and have been suggested to control formation of most Pacific arcs, such as Papua New Guinea, New Zealand, Tonga, New Hebrides, and the Marianas. In those settings, a transition from collision to subduction has been suggested as responsible for rapid vertical axis rotations and subsequent curvature of the upper plates [Wallace et al., 2005].
 The Mediterranean region has long provided a natural laboratory for arc origin models to developed and tested. One of the principal peculiarities of many Mediterranean arcs is their narrow and tight shape, which in some cases have a change in structural trend of almost 180°. Moreover, some Mediterranean arcs are located on top of narrow slabs, well imaged by deep seismicity and seismic tomography, which offer a unique opportunity to unravel the connection between subduction processes and arc formation. Finally, in Mediterranean arcs a large quantity of paleomagnetic data collected in the last few decades exists, allowing kinematic models for their formation to be vigorously tested (among many others, Channell et al. , Kissel et al. , Platt et al. , Platzman , Platzman et al. , and Speranza et al. [1997, 1999]). Of the different arcs encircling the Mediterranean area, the Calabrian Arc represents one of the most distinctive examples of arc formation related to subduction processes. The most significant peculiarities of this arc are its geodynamic location on top of a very narrow (roughly 200 km) and steeply dipping (70°) slab, which is well defined by seismicity and seismic tomography (among others, Anderson and Jackson , Selvaggi and Chiarabba , Lucente et al. , and Spakman et al. ), and its particularly tight and narrow shape. Starting from the original work by Malinverno and Ryan , many authors have proposed that the Calabrian Arc formed as a consequence of progressive retreat of the subduction zone, controlled by the presence of a narrow Ionian oceanic lithosphere intervening between the Apulian and African continental lithospheres [Faccenna et al., 2004; Gueguen et al., 1998; Lonergan and White, 1997; Rosenbaum and Lister, 2004]. In this geodynamic context, the secondary origin of the curvature of the Calabrian Arc has been demonstrated since the 1980s, on the basis of available paleomagnetic data [e.g., Eldredge et al., 1985]. Most of these data came from Mesozoic rocks from the Southern Apennines and Sicily, and are, therefore not useful for fully unraveling the Neogene kinematics of the Calabrian Arc. In the last 20 years, a large number of paleomagnetic data have been collected in Neogene deposits all along the Calabrian Arc (among others, Butler et al. , Channell et al. [1980, 1990], Jackson , Cifelli et al. [2004, 2007a, 2007b], Gattacceca and Speranza , Mattei et al. [2004, 2007], Scheepers et al. [1993, 1994], and Speranza et al. [1999, 2003]). In this study we reanalyze the available paleomagnetic database for the Calabrian Arc. This new quantitative analysis represents a unique opportunity to constrain the tectonic and kinematic evolution of the Arc during the Neogene. Results from this analysis indicate that a simple orocline model is not sufficient to describe the complex evolution of Calabrian Arc curvature. Alternatively, a progressive arc model for the Calabrian Arc is proposed, which is mechanically related to the geometry and evolution of the Ionian subduction system.
2. Geological Setting of the Calabrian Arc
 The Calabrian Arc defines a curved mountain belt encircling the Tyrrhenian Sea, from the Southern Apennines to the Sicilian Maghrebides (Figure 1). The curvature of the Calabrian Arc is defined by a regional variation in fold axes trend and compressional structures, which range from NW–SE with a NE vergence in the Southern Apennines to E–W with a southern vergence across Sicily (Figure 1).
 A peculiarity of the Calabrian Arc is the existence of a polymetamorphic orogenic domain, the Calabria-Peloritani Domain (CPD), in the core of the arc (Figure 2). This structural domain shows completely different geologic and structural characteristics compared to the Southern Apennine-Maghrebide orogenic system [Alvarez et al., 1974; Bonardi et al., 2001; Faccenna et al., 1997; Malinverno and Ryan, 1986; Mattei et al., 2002]. These differences include (1) an upper Tertiary-to-Quaternary fore arc basin that is very well developed along the Calabrian Ionian side of the CPD, and is missing in the Southern Apennines and Sicily [Bonardi et al., 2001; Cavazza et al., 1997]; (2) an absence of typical Neogene foreland-foredeep system in front of the deforming orogenic structures, which is instead observed in the Maghrebian-Southern Apennines fold-thrust belt; (3) the early occurrence of extensional tectonics along the Tyrrhenian side of the CPD (middle–late Miocene) compared to that in the Southern Apennines and Sicily (Plio-Pleistocene); and (4) the occurrence of Alpine high-pressure units that do not outcrop in the Southern Apennines and Sicily fold-thrust belts.
 The peculiarity of the CPD with respect to the Southern Apennines and Sicily clearly reflects a difference in lithospheric structure and subducting processes. Importantly, the CPD lies on top of a very narrow slab that is well defined by deep seismicity and seismic tomography, whereas there is no trace of a subducting slab beneath the Southern Apennines and Sicily [Piromallo and Morelli, 2003; Wortel and Spakman, 2000]. The across-strike width of the slab in the CPD corresponds to the Ionian realm, a Mesozoic oceanic lithosphere, which intervenes between the Apulian and African continental margins [Dercourt et al., 1986]. Its northeastern and southwestern boundaries match, at the surface, the boundary between the CPD and the Southern Apennines and Sicily, respectively, which conversely do not show any evidence of deep seismicity (Figure 1). Trench retreat of the Ionian slab was particularly fast in the Neogene, as suggested by high extension rates (50–70 mm a−1) recorded in the newly formed Vavilov (Neogene) and Marsili (early middle Pleistocene) oceanic seamounts [Nicolosi et al., 2006; Marani and Trua, 2002], as well as rapid migration of the trench and the progressive movement of the foredeep basins [Patacca et al., 1990]. Conversely, trench retreat velocity drastically reduced since about 1 Ma to the present-day, as shown by GPS and paleomagnetic data [D'Agostino and Selvaggi, 2004; Goes et al., 2004; Mattei et al., 2007].
 The aforementioned arguments all suggest that the present-day shape of the arc is the result of a complex history of subduction, and therefore a simple oroclinal bending model must be reevaluated.
 Generally, the paleomagnetic data show a correlation between changes in paleomagnetic declination and changes in orogenic trend. This suggests that the present-day shape of the Calabrian Arc is a secondary feature, achieved through circa symmetrical opposite rotations of the belt limbs, with counterclockwise (CCW) rotations in the Southern Apennines and clockwise (CW) rotations in the Calabria-Sicilian Maghrebides. Paleomagnetic rotations are confined to the orogenic wedge and do not extend into the Apulia and Hyblean forelands, which did not undergo tectonic rotations. This suggests that tectonic rotations were related to a progressive deformation of the chain, which provides information on the orogenic wedge kinematics. A final important point is that paleomagnetic data indicate that tectonic rotations were almost over by the end of the early Pleistocene [Cifelli et al., 2007b; Mattei et al., 2004, 2007; Scheepers et al., 1993]. Accordingly, in the following analysis we exclude data collected in the Apulian, Hyblean and External Saccense foreland domains (Figure 1). Conversely, we focus on paleomagnetic data collected in orogenic nappes, differentiated according to the ages of sampled rocks and their tectonic positions within the orogenic wedge.
4. Methodology and Results
 Paleomagnetic data from Southern Italy have been compiled and analyzed in order to determine the main mechanism responsible for arc curvature.
 Several methods have been proposed to test the origin of curved mountain belts using paleomagnetic data, which take into account both the geographic distribution of paleomagnetic rotations along the arc, and the relationships between vertical axis rotations and deviation in structural trend (orocline test, according to Lowrie and Hirt  and Schwartz and Van der Voo ). In a hypothetical perfect orocline, if paleomagnetic declinations are plotted against structural marker orientations (fold axes, magnetic lineations, kinematic indicators), a linear regression should exist (Figure 3a) with a slope close to one. A similar pattern is observed if paleomagnetic rotations are plotted against a geographic reference, for example against the distance along the length of the arc (Figure 3b). Classic examples of positive orocline tests come from the Cantabrian Arc in Spain [Hirt et al., 1992; Parés et al., 1994; Weil et al., 2000], the Appalachians [Kent, 1988], the Andes [Allmendinger et al., 2005; Butler et al., 1995; MacFadden et al., 1995], and from the Central and Northern Apennines [Mattei et al., 1995; Speranza et al., 1997]. Conversely, in primary arcs, no correlation should exist between paleomagnetic rotations and geographic or structural references. In this case, the slope of the regression line should be close to zero (Figure 3c). However, a large number of structural and paleomagnetic studies show that many orogens have complex geometries, implying tectonic evolutions that are difficult to fit to an idealized primary arc or an orocline model. Complications may arise from local deviations from the regional trend to form minor arcs (Figure 3d), or from abrupt deviations in structural trend due to major tectonic discontinuities (Figure 3e). In particular, lateral heterogeneities in the stratigraphic successions or the presence of local obstacles to foreland thrust propagation, such as structural highs or intrusive bodies in the foreland, have been recognized to cause variation in the detachment depth, forming salients and recesses that characterize many curved orogenic belts, such as the Appalachians, the Zagros Mountains, and the Northern Apennines [Castellarin et al., 1985; Marshak, 2004; Pieri and Groppi, 1981; Sherkati et al., 2006; Tull and Holm, 2005]. In these cases, the orocline test may show a very different pattern with respect to those tests based on geographic distributions of paleomagnetic rotations. In fact, as the orocline test takes into account the correlation between paleomagnetic rotations and structural fabric, independently from their geographic distribution along the length of the arc, a linear correlation with a slope close to 1 is expected in the presence of minor salients and recesses along the main arc. On the other hand, the distribution of paleomagnetic rotations along the length of the arc will show a different pattern, characterized by a larger scatter of paleomagnetic rotation values, and by a lower correlation coefficient compared to simple arc models (Figures 3d and 3e).
 In addition to the difficulties in testing the kinematics of arcuate orogens given their complex structural architectures, a further complication is that most curved orogenic structures are progressive arcs, implying that rotation and deformation are synchronous. In this case, if paleomagnetic data come from synorogenic sediments, it is expected that the distribution of paleomagnetic rotations against structural axes (or against distance along the length of the arc) plot in a very wide region between the line with zero slope and the line with a slope 1 (Figure 3f) [Allerton, 1998; Weil and Sussman, 2004]. This last possibility leaves the temporal relationships between curvature and orogenic deformation unresolved unless more detailed analysis is completed.
 On the basis of these considerations, we analyze paleomagnetic data from the Calabrian Arc using multiple approaches. First, we analyze the spatial distribution of paleomagnetic declinations along the length of the arc and along three different profiles oriented orthogonal to the main structural axes of the chain. We then compare the amount of rotation displayed by paleomagnetic declinations with the change in strike of the structural trends (oroclinal test according to Schwartz and Van der Voo ). Finally, we investigated the temporal variation of paleomagnetic declinations to constrain the timing of the curvature.
4.2. Distribution of Paleomagnetic Declinations Along the Arc
 We plotted paleomagnetic rotations as a function of their location along the length of the Calabrian Arc in order to analyze their spatial distribution (Figure 4). Starting from the northernmost sector of the Southern Apennines where paleomagnetic data are available, paleomagnetic rotations were analyzed in relation to their distance, in percentage, along the arc. A hypothetical axis of the chain was traced across the arc and tectonic rotations were represented as the projection of the sampled sites on this axis (Figure 1). In the Southern Apennines and Sicily, i.e., at the opposite edges of the arc, sedimentary deposits underwent large magnitude tectonic rotations, CCW and CW, respectively. Within these two areas, however, paleomagnetic rotations show a very large scatter, which could be due either to the presence of minor arcs, or to the large age range of the sampled units, as will be discussed later. The distribution of paleomagnetic rotations along the arc shows that an abrupt change in paleomagnetic directions occurs along a very narrow region. This region is located between the Southern Apennines and Calabria and is characterized by a complex pattern of paleomagnetic rotations [Cifelli et al., 2007b], which change from CCW rotations in the north to CW rotations in the south (Figure 1). This is not expected in a typical orocline, where a gradual change in the magnitude of along-strike paleomagnetic rotations should occur. Finally, no significant differences in paleomagnetic rotations are observed along the entire segment of the arc represented by the CPD, which indicated that this sector behaved as an almost homogeneous and rigid block (Figure 4). The different behavior of the CPD compared to the Southern Apennines and Sicily thrust belts is outlined in Figure 5. In Figure 5, the amount of paleomagnetic rotations is represented along three different transects perpendicular to the coast (in the Southern Apennines, in the CPD and in Sicily). Tectonic rotations are plotted as a function of their tectonic location in the orogenic wedge, making the reasonable assumption that moving away perpendicularly from the Tyrrhenian coast we progressively move from the more internal deformed sectors of the chain to the more external, less deformed sectors [Bigi et al., 1992]. In the Southern Apennines and Sicily, the distribution of paleomagnetic declinations indicates a gradual reduction in the amount of rotation away from internal sectors toward external sectors [Channell et al., 1990; Gattacceca and Speranza, 2002]. This is clearly seen in Sicily, where a decrease in the amount of paleomagnetic rotations from sites of the same age is observed, moving from the internal Panormide-Imerese, Sicanian and M. Judica units (up to 145°CW), to the intermediate paleogeographic units of the Trapanese and Internal Saccense Platform (50° of CW), whereas the external, undeformed units (External Saccense Platform) are unrotated (Figures 1 and 5a). Similarly, in the Southern Apennines we observe a gradual reduction in the amount of paleomagnetic rotation from the internal to external sectors of the chain (Figures 1 and 5b). This paleomagnetic trend may be either due to the different tectonic positions of the sampled sites or to their progressively younger ages moving from the internal to external sectors of the orogenic wedge. In the CPD, there are no significant differences among paleomagnetic rotations measured in the different structural positions in the Ionian fore arc basins and the Tyrrhenian postorogenic extensional basins, which are characterized by an almost homogeneous CW rotation of about 20° [Cifelli et al., 2007b; Scheepers et al., 1994; Speranza et al., 2000], uppermost lower–middle Pleistocene deposits excluded (Figure 5c).
4.3. Oroclinal Test
 The distribution of paleomagnetic rotations along the Calabrian Arc indicates that the limbs of the arc underwent circa symmetrical opposite rotations, with CCW rotations in the Southern Apennines and CW rotations in Sicily. In order to verify whether this distribution correlates with changes in the trend of the orogen supporting the orocline model, we performed an oroclinal test. Paleomagnetic data described above indicate that the CPD had a different tectonic history compared to the other portions of the Calabrian Arc. For this reason, we excluded paleomagnetic data collected in the CPD from the oroclinal test, focusing only on data from the Southern Apennines and Sicily. The relation between paleomagnetic declination and structural directions was investigated using the method originally proposed by Schwartz and Van der Voo  for the Appalachians and later applied in different curved orogenic systems (among others, Eldredge et al. , Lowrie and Hirt , Parés et al. , and Speranza et al. ). In our analysis, we directly compared paleomagnetic rotations with structural strikes determined either from the original publications or from geological maps of appropriate scale (Table 1). Where available, anisotropy of magnetic susceptibility data were used to estimate fold axis directions, representing the in situ site mean direction of the principal maximum susceptibility axis (kmax), a very good indicator of the local fold axis direction [Cifelli et al., 2005; Graham, 1954; Kissel et al., 1986; Kligfield et al., 1983]. We did not consider those data where the relation between paleomagnetic rotation and structural elements was poorly constrained.
Table 1. Paleomagnetic and Structural Directions Used for the Oroclinal Test
Error for Rotations Values (±)
Number of Sites
Strike values obtained from regional map.
Strike values obtained from site strike.
Strike values obtained from anisotropy of magnetic susceptibility.
 In order to quantify the variation in paleomagnetic directions as a function of change in the regional structural trend, we used a linear regression technique [e.g., Eldredge et al., 1985; Schwartz and Van der Voo, 1983]. In Figure 6, paleomagnetic rotations are plotted against structural strike. The distribution of the data indicates that tectonic rotations in the two ends of the Calabrian Arc are well correlated with the structural trend. Calculation of the best fit line of the data gives a correlation coefficient of 0.73, with a line of slope 0.55 (solid line in Figure 6). This value indicates quite a good correlation between changes in declination and the orogen trend, indicating that Calabrian Arc curvature is not a primary feature. However, ideal oroclinal bending would display a line of unit positive slope (dashed line in Figure 6), whereas in Figure 6 the distribution of paleomagnetic rotations ranges between the slope of zero and one, similar to that expected in a progressive arc. A more detailed analysis that takes into account the timing of the Calabrian Arc curvature is therefore needed.
4.4. Temporal Variation of Paleomagnetic Declinations
 The temporal variation of paleomagnetic declinations in the Southern Apennines and Sicily was analyzed considering both their geographic location along the arc and the correlation with structural trend. Paleomagnetic data were grouped into four distinct age intervals, representative of the different stages of Calabrian arc evolution: Mesozoic to middle Miocene, late Miocene–Pliocene, early Pleistocene, and uppermost early Pleistocene–middle Pleistocene. During the Mesozoic to early Tertiary the different paleogeographic units, that presently form the fold and thrust belt of the Southern Apennines and Maghrebian chain, were still part of the passive margin of the African-Apulia plate or represented the foreland of the former Apennine and Maghrebian chain. These units did not rotate with respect to Africa until the middle Miocene, when they were incorporated into the Apennine and Maghrebian fold-thrust belt [Channell et al., 1990; Cifelli et al., 2007b; Gattacceca and Speranza, 2002]. During the late Miocene and up to the early Pleistocene, the Southern Apennines and Maghrebian units were involved in the main phases of compressional tectonics, with progressive migration of the thrust front toward the Apulian and Hyblean forelands. Because of the rapid tectonic evolution of compressional deformation during this time interval, we consider data from upper Miocene–Pliocene deposits separately from lower Pleistocene units. Finally, as compressional phases almost ceased in the Southern Apennines and Western Sicily during the uppermost early Pleistocene [Mattei et al., 2007, and references therein], data from uppermost lower Pleistocene to middle Pleistocene units were also separated.
 In Figure 7 we report results from geographic and oroclinal tests for each time interval. Paleomagnetic rotations show very high correlation coefficients for all time intervals except the uppermost late Pleistocene–middle Pleistocene, both correlating paleomagnetic data with distance along the length of the arc (Figures 7a, 7b, 7c, and 7d), and paleomagnetic rotations with structural trend (Figures 7e, 7f, 7g, and 7h). A clear decrease in the slope correlation is also observed through time. In particular, in the oroclinal test (Figures 7e, 7f, 7g, and 7h) the slope is very close to 1 (0.86) for the Mesozoic–middle Miocene interval, and decreases to 0.33 for the upper Miocene–Pliocene deposits and 0.30 for the lower Pleistocene units, until it reaches values close to zero for the uppermost lower Pleistocene–middle Pleistocene units, which are almost unrotated both in the Southern Apennines and in Sicily.
5.1. Kinematics of the Calabrian Arc: Implications for the Oroclinal Bending Model
 Among the different arcs of the Mediterranean region, the Calabrian Arc has long been considered a classic example of an orocline, which assumes that its curvature was acquired through bending of an originally almost linear orogenic chain [e.g., Eldredge et al., 1985]. Reanalyses of the large paleomagnetic database collected in the last 20 years demonstrate that this model is inappropriate for explaining the spatial distribution and temporal evolution of paleomagnetic rotations along the arc. If we consider the distribution of vertical axis rotations versus time, we note that the paleomagnetic rotations systematically decrease in progressively younger sedimentary deposits in both the Southern Apennines and in Sicily (Figures 4, 5, and 7). This trend is clearly documented in the decrease in slope values of calculated regression lines from oroclinal test for the different age intervals reported in Figure 7 (for both the variation of paleomagnetic rotations along the length of the arc and their correlation with the variation in structural trend). These results clearly indicate that the curvature of the Calabrian Arc was acquired progressively and that, taking into account only the Southern Apennines and Sicily, the Calabrian Arc could be considered a progressive arc, following the kinematic classification proposed by Weil and Sussman .
 The knowledge that the Calabrian Arc acquired its curvature during the progressive deformation of the Southern Apennines and Sicily emphasizes the possibility that other classic examples of oroclines might instead be classified as progressive arcs. This seems the case of the Gibraltar Arc, the Northern Apennines, the Bolivian Orocline, and the Appalachians, where recent paleomagnetic data indicate that such curved belts progressively bent throughout their deformation history (among many others, Barke et al. , Cifelli et al. , Gray and Stamatakos , and Speranza et al. ). It is worth noting that vertical axis rotations have been measured, using GPS data, in the Central Andes [Allmendinger et al., 2005; Bourne et al., 1998; Hall et al., 2004] and the Western Pacific [Wallace et al., 2005], where the curvature of mountain belts is still an active process that accommodates most of the present-day deformation on top of actively subducting slabs. On the basis of these observations, it seems reasonable that the orocline, as related to curved mountains that acquire their curvature after the end of deformation, represents mostly a conceptual model that can be rarely applied to real orogens, where curvature is acquired during deformation (for a recent review, see Weil and Sussman ).
 However, when we include the CPD domain in the entire length of the Calabrian Arc, it is clear that the progressive arc model is also inadequate to explain the measured pattern of paleomagnetic rotations along the arc. In fact, in the progressive arc model a continuous change in the sense and magnitude of paleomagnetic rotations is expected along the length of the arc. This is not the case for the area that marks the boundary between the Southern Apennines and the CPD, where an abrupt change in the sense of paleomagnetic rotations (from CCW in the Southern Apennines to CW in the CPD) has been observed (Figure 4). Furthermore, within the CPD the geographic distribution of paleomagnetic rotations and their progression with time show some notable differences with respect to the Southern Apennine and Sicily thrust belts: (1) vertical axis rotations are almost uniform throughout the region, with any appreciable differences related to the tectonic location of the sampled sites; and (2) there is no evidence for a decrease in paleomagnetic rotation with time, because an almost constant value of about 20°CW rotation was obtained from Serravallian to lower Pleistocene strata. These observations indicate a strong difference in the rotational style between the two arms of the arc (the Southern and Sicilian Apennines) and the arc core (the CPD).
5.2. Calabrian Arc Curvature and the Evolution of the Subduction Process
 Determining the precise distribution and timing of vertical axis rotations during the progressive curvature of the arc is a necessary step in addressing the most important aspect of orogenic arcs: identifying the tectonic or geodynamic mechanism by which curvature is attained.
 Paleomagnetic data support the hypothesis that the Calabrian Arc curvature process was related to the configuration and evolution of the subducting slab. The different rotational domains that characterize the Calabrian Arc clearly mirror the deep configuration of the lithosphere, as the CPD is located on top of the Ionian slab, whereas the Southern Apennines and Sicilian Maghrebides represent the parts of the Calabrian Arc where continental collision processes occurred. This agrees with previous geodynamic models that suggest that the origin of Calabrian Arc curvature comes from rollback of the narrow Ionian oceanic slab, which is laterally confined by the Apulian and African continental lithospheres [Faccenna et al., 2004; Gueguen et al., 1998; Lonergan and White, 1997; Malinverno and Ryan, 1986¸ Rosenbaum and Lister, 2004].
 First, how did the Calabrian Arc evolution result in such a tight arc? Numerical and analog models, compared with natural examples, suggest that arc curvature tends to increase when the downgoing slab is narrow. Curvature is also favored by the presence of strong heterogeneities in the subducting plate and can be enhanced by lateral mantle flow, possibly induced by slab-retreat [Funiciello et al., 2003; Morra et al., 2006; Schellart et al., 2007]. In the case of the Calabrian Arc the subducting plate is strongly heterogeneous, with a stiff and heavy oceanic portion (the Ionian slab) that is more resistant to bending and has a higher subduction velocity compared with the continental part of the slab (Apulian and African lithospheres). The role of lateral slab heterogeneities progressively increased during the Neogene, when the amount of oceanic lithosphere available for subduction was progressively reduced and the Apulian and African continental lithospheres were increasingly involved in the subduction process (see paleogeographic reconstructions in papers by Bonardi et al. , Cifelli et al. [2007b], Faccenna et al. , Gueguen et al. , and Patacca et al. ). The progressive increase in the area affected by continental collision decreased the efficiency of subduction, which became inactive along most of the arc length. Subduction is now confined to the frontal part of the Calabrian Arc, which corresponds to the present-day Ionian oceanic domain (Figure 1). This process is particularly well constrained in the Southern Apennines where structural and stratigraphic data show that foredeep basins progressively shifted southeastward, parallel to the front of the chain during the late Pliocene to Pleistocene, and that thrust activity along the outer front of the chain was progressively sealed, marking the cessation of active subduction along this segment of the arc [Mattei et al., 2004; Patacca and Scandone, 2001; Tropeano et al., 2002].
 Another intriguing aspect of Calabrian Arc evolution is the exceptionally fast rate of rotation [Mattei et al., 2007; Scheepers et al., 1993]. Numerical models and natural examples show that the curvature of arcs strongly increases along the edges of the slab, where rotations can be particularly fast, while the central part of the trench tends to not rotate and remain nearly rectilinear [Morra et al., 2006; Wallace et al., 2005]. In the Calabrian Arc, rapid and large rotations achieved 5° to 20° Ma−1 in the transition area from subduction to continental collision (Southern Apennines and Sicily), where the maximum along-strike gradient in subduction rate is recognized [Mattei et al., 2004]. Conversely, the CPD, which is located in the frontal part of the arc at the top of the Ionian subducting lithosphere, behaved as an almost rigid block and underwent only minor vertical axis rotations with respect to the paleogeographic realms along the edges of the slab. However, the exceptionally fast rotations derived in the Calabrian Arc may be the result of combined mechanisms, and could have been enhanced by lateral mantle flow along the slab edges. This process has been clearly recognized by seismic studies in the Calabrian Arc [Pondrelli et al., 2004], and has been suggested as a possible mechanism for generating arcs on top of subducting slabs, being particularly efficient when the slab is very narrow [Dvorkin et al., 1993; Funiciello et al., 2006; Schellart et al., 2007].
 In summary, we suggest that the presence of lateral heterogeneities in the subducting lithosphere and the subsequent progressive decrease in trench width during subduction is ultimately responsible for Calabrian Arc formation and for its present-day narrow configuration.
 In this paper, we have used paleomagnetic data to document the spatial and temporal evolution of tectonic rotations in the Calabrian Arc. Data show the occurrence of Neogene rotations of opposite sense and variable amount in different sectors of the Calabrian Arc. The widely varying paleomagnetic declinations roughly correlate with the overall structural trend of the arc, suggesting that the sites underwent rotation in a process that involved folding about vertical axes during arc tightening. Paleomagnetic data provide further insights into spatial and temporal heterogeneities and complexity of Neogene deformation in the Calabrian Arc. Integrated with structural and geologic data, these data give a general image of how and why the Calabrian Arc formed with a curved shape, supporting the idea that the rotational pattern indicates a complex tectonic history and that the classic concept of the orocline may not be adequate to fully describe curvature of the Calabrian Arc. The distribution and timing of vertical-axis rotations show that the progressive shaping of the Calabrian Arc from the Neogene to Quaternary reflects the evolution of slab rollback processes of a laterally heterogeneous subducting plate, and its transition from subduction to collision.
 The authors would like to thank associated editor and the two reviewers Arno Brandon Weil and Conall Mac Niocaill for the accurate and constructive reviews of the paper. We would also like to thank Fabio Speranza for his stimulating suggestions.