1.1. Tectonics and Climate Change of Central Asia
 Two extreme views of continental tectonics can be contrasted by the significance that they attribute to major faults in the crust. One view treats such faults as major lithospheric discontinuities that separate nearly rigid blocks, such that resistance to regional deformation depends on resistance to slip along the faults. The other view treats the upper crust as weak, compared to a stronger, but nevertheless viscous, substratum. In this view, the strength of the continuously deforming upper mantle, and perhaps lower crust, determines the distribution of regional deformation and the strain of the upper crust, manifested by faulting and folding, provides a roughened image of that deeper deformation.
 The difference in these views developed in part from studies of Asian tectonics. Proponents of the first view have worked to quantify rates and amounts of slip along major faults (Figure 1) [Armijo et al., 1986, 1989; Liu, 1993; Peltzer et al., 1989], and have used elements of plate tectonics to describe the kinematics of deformation [Avouac and Tapponnier, 1993]. Others [England and McKenzie, 1982; England and Houseman, 1986] have utilized numerical experiments on a thin viscous sheet that deforms continuously, and implicitly treated the upper crust as merely a collection of passive markers floating on a stronger, deforming, viscous substratum. Although the average kinematic descriptions resulting from the two approaches differ little when smoothed (compare Avouac and Tapponnier  and Peltzer and Saucier  with England and Molnar , Holt and Haines  or any of Holt et al. [2000, 1995, 1991]), the mechanical underpinnings differ profoundly in one respect. The first view suggests that much of India's collision with Eurasia is absorbed by lateral transfer along strike-slip faults penetrating into the upper mantle and carrying material eastward out of India's northward path; the second view considers most of India's penetration into Eurasia to be absorbed by crustal thickening with minor lateral transfer.
 These contrasting points of view have their origins in differing estimates of the slip rates along the major strike-slip faults in Asia. Assuming postglacial ages (i.e., 10 ± 2 ka, or more recently, 13.5 ± 2 ka) for displaced landforms, some have argued that slip along several of the major faults is rapid: ∼30 mm yr−1 on the Altyn Tagh and Karakorum faults [Avouac and Tapponnier, 1993; Liu, 1993; Peltzer et al., 1989] and 10–20 mm yr−1 on right-lateral faults across the middle of Tibet [Armijo et al., 1989] (Figure 1). Others suggest significantly lower rates for right-lateral faults within Tibet [Molnar, 1992]. On the basis of analyses of slip vectors of earthquakes along the Himalayas, Molnar and Lyon-Caen  argued that slip along the Karakorum fault could not exceed ∼10 mm yr−1. England and Molnar , exploiting a method similar to that of Haines and Holt , found that it is not possible to generate a consistent velocity field (i.e., one obeying compatibility of strain) for Asia with a slip rate greater than ∼10 mm yr−1 on the Karakorum fault or greater than ∼15 mm yr−1 for the Altyn Tagh fault.
 Although there is no historical record of a significant earthquake on the Karakorum fault, its importance was noted by Burtman et al.  and Peive et al. , who estimated, on the basis of observed offset units, that 250 km of right-lateral slip had occurred. As Landsat and SPOT imagery became available, the Karakorum fault was recognized as one of Asia's major active faults [Liu, 1993; Molnar and Tapponnier, 1975, 1978]. Right-lateral slip along this fault accommodates part of India's penetration into Eurasia, as the western syntaxis of the Himalaya advances toward stable parts of Eurasia. In the region of the Karakorum fault, India's motion toward Eurasia is nearly due north [e.g., DeMets et al., 1990], but slip vectors of earthquakes show northeastward underthrusting of India beneath the Himalayas in this region [e.g., Molnar and Lyon-Caen, 1989]. Right-lateral slip along the Karakorum fault, thus, absorbs part of the east-west component of movement, so that the oblique slip at the Himalaya is partitioned into pure thrust slip at the Himalaya and pure strike slip farther northeast. Armijo et al.  suggested that the Karakorum fault is the westernmost of a series of en echelon right-lateral faults along which much of Tibet is transferred east with respect to both stable Eurasia and India. Additional studies have produced estimates of total slip along the Karakorum fault ranging from as little as 85–120 km [Searle, 1996] to perhaps as much as 1000 km [Peltzer and Tapponnier, 1988], implying average slip rates that differ by nearly an order of magnitude.
 In the most extensive geologic study of slip rates carried out along the Karakorum fault, Liu  and Liu et al.  exploited satellite (SPOT) imagery as well as field investigations on the Tibetan side that allowed direct examination of offset features along parts of the fault. Liu  studied 24 sites in the southern Karakorum with clear offsets (ranging from ∼50 to ∼350 m) along the fault, suggesting sustained and present-day activity. One third of these sites show offsets of 250–350 m; assuming that the offset landforms date from approximately 10 ± 2 ka, Avouac and Tapponnier  and Liu  inferred slip rates of 30–35 mm yr−1.
 A critical component of the rapid inferred rate is the assumption that the history of expansion and retreat of glaciers in this region corresponds directly to global paleoclimatic variations. If dates of formation of offset geomorphological features were older than the postglacial ages assigned by Avouac and Tapponnier  and Liu , the corresponding slip rates would be proportionally lower. The assumed chronology for formation of major geomorphological features is supported by correlations with lacustrine sedimentary records from western Tibet [Gasse et al., 1991, 1996; Van Campo and Gasse, 1993]. These, as well as sedimentary records from lakes in the central and eastern Tibetan Plateau [Lister et al., 1991] and in northern Xinjiang [Rhodes et al., 1996], provide generally consistent evidence of a humid period during the latest Pleistocene and early Holocene. However, the sediment cores on which these records are based provide little information on climatic variability earlier than 15–20 ka.
 The history of glaciation in much of high Asia remains poorly known, and there are few direct dates on the formation of glacial features. Nevertheless, there is evidence that central Asian glaciation was characterized by relatively small valley glaciers and not a regional ice sheet. In some areas, particularly in the northern and eastern areas of the Tibetan plateau, these valley glaciers reached their maximum positions during the Last Glacial Maximum (∼20 ka), but in areas to the southwest of the Tibetan Plateau maximum expansions of valley glaciers occurred earlier (see reviews by Benn and Owen , Derbyshire et al. , Lehmkuhl and Haselein , and Shroder et al. ). In any case, there is a clear need for additional direct dating of geomorphic features and a better knowledge of timing of glacial expansion in this part of the world.
 Because the Karakorum fault has been assigned one of the highest slip rates, it provides an ideal laboratory to examine the suggestions (1) that major glacial and alluvial landforms in central Asia date from the last glacial period, (2) that rates of slip along major strike-slip faults in Asia can be estimated reliably by assuming that landforms formed during the last deglaciation, (3) that rates of slip along such faults are high, and thus (4) that eastward extrusion of Tibetan crust is rapid and plays a key role in the kinematics of Asian deformation.
1.2. Cosmic Ray Exposure Techniques
 Cosmic ray exposure dating utilizes the accumulation of rare nuclides in rock exposed at the earth's surface. These nuclides are produced through nuclear reactions induced by high-energy cosmic radiation. This method has been applied to a range of problems in climate history and geomorphology (see reviews by Bierman, , Cerling and Craig , or Lal ) including applications to dating of depositional features [Anderson et al., 1996; Brown et al., 1998]. The approach provides a means of determining slip rates along faults by dating offset depositional features (alluvial fans, moraines, debris flows). Here we use in situ-produced 10Be (t1/2 = 1.5 Myr), to date debris flows deposited over active faults, and subsequently displaced by movement on those faults.
 In general, the accumulation of these nuclides in a rock exposed at the earth's surface and undergoing erosion at a constant rate (ε; g cm−2 yr−1) may be described by
where N(t) is the concentration of a cosmogenic nuclide at time t, P represents the production rate at the rock's surface (atom g−1 yr−1); L is the characteristic attenuation length (g cm−2) for cosmic secondary particles (mostly neutrons) producing cosmogenic nuclides, and λ is the decay constant for radioactive cosmogenic nuclides (yr−1). Because the half-lives of the cosmogenic radionuclides commonly used for cosmic ray exposure dating (10Be, 26Al, 36Cl) are long relative to the ages of most landforms in a tectonically active region, little additional information is gained by measurement of multiple nuclides in the same sample. Any date obtained through this method can thus either overestimate (due to exposure prior to the episode of interest such that N(0) is significant) or underestimate (because of decreased accumulation of the cosmogenic nuclide resulting from postdepositional processes such as burial, shielding, or erosion) the time of the deposition or the formation of a feature. These biases may be evaluated by judicious sampling and field observations. The working hypothesis we adopted for sampling was that rocks in our study area had minimal near-surface exposure before deposition, so that their cosmogenic nuclide content would reflect surface exposure in their present positions. Any scatter in the values would be the result of postdepositional processes, all of which decrease the accumulation of cosmogenic nuclides. Under these conditions, the sample with the highest cosmogenic nuclide concentration would be the one least affected by postdepositional processes and would thus yield the most accurate age, ignoring potential errors due to uncertainties in cosmogenic nuclide production rates [Brown et al., 1998]. These assumptions may be reasonable in many high-energy depositional environments and may be evaluated through examination of “geological blanks”: samples of rocks whose exposure histories in source regions and during transport were similar to those of the materials to be dated, but that have been exposed only briefly in their present positions [Brown et al., 1998].