The widely accepted view that catastrophic flow of liquid water was the dominant process involved in formation of outflow channels on Mars has as part of its foundation the assumption that the flow of lava could not have formed Martian features such as streamlined islands and anastamosing channels. However, lunar and Venusian channels, believed to have formed through volcanic processes in the absence of water, are indeed associated with such features. Additionally, many lunar and Venusian rilles head at topographic depressions, a common characteristic of Martian outflow channels. Lunar rilles typically lack positive relief accumulations of volcanic deposits at their mouths, suggesting that contrary to previous assertions, absence of such accumulations at the mouths of Martian outflow channels is not incompatible with a common mode of formation. Consistent with an igneous origin for outflow channels and certain valley systems on Mars, volcanic processes can produce a wide range of landforms that are similar to those normally associated with the flow of water, including channel terraces and complex channel networks. The simplest interpretation of Martian channels that extend from volcanic source regions onto volcanic plains is as conduits formed by the flow of lava. Channel formation hypotheses requiring major changes in Martian atmospheric properties or repeated catastrophic flow of water from volcanic features appear needlessly exotic alongside the igneous hypothesis for channel formation. In light of these points, future investigations of Martian outflow channels should more actively consider volcanic processes as candidate mechanisms for channel formation.
 Cited problems with the volcanic hypothesis for outflow channel formation include the apparent absence of large accumulations of volcanic materials at the mouths of Martian channels, a belief that the flow of lava cannot form anastamosing channels, and a perceived lack of volcanic sources at the heads of channels [Carr, 1974, 1996; Baker et al., 1992a]. Rejection of a channel-forming role for lava flows has rested most heavily on the assumption that the dominance of accretionary and constructive processes in the flow of lava excludes the possible action of erosive processes necessary for formation of channel features such as streamlined islands [e.g., Baker, 1978a; Baker and Kochel, 1978, 1979; Mars Channel Working Group, 1983].
 On the Moon, large volumes of basalt are known to have been emplaced by low-viscosity lava flows [e.g., Head, 1976], and although many lunar flow units lack obvious channels and source regions [e.g., Whitford-Stark, 1982; Wilhelms, 1987], others are associated with clear sources and associated distributary channels [e.g., Greeley, 1971a; Zisk et al., 1977; Strain and El-Baz, 1977]. Some of these lunar channels have been used in the past in the assessment of possible roles for flowing lava in formation of Martian channels [Masursky, 1973; Carr, 1974], resulting in determinations that at least some Martian channels of relatively simple morphology were likely formed through volcanic processes. Recent studies have suggested that some channels associated with sites of purported ancient Martian lakes show morphological and contextual affinities not with terrestrial fluvial landforms but with lunar volcanic rilles [Leverington and Maxwell, 2004; Leverington, 2004].
 Venus is characterized by a wide variety of volcanic features, extensive volcanic plains, and several classes of channels [e.g., Head et al., 1992; Guest et al., 1992; Baker et al., 1992b]. The flow of low-viscosity lava on Venus is believed to have produced both simple sinuous rilles and complex channel networks, including a prominent outflow channel associated with large streamlined islands [e.g., Baker et al., 1992b; Head et al., 1992; Baker et al., 1997]. Venusian lava channels share many attributes with Martian outflow channels [Baker et al., 1992b, 1997], and although it has been speculated that these channels may owe their existence to a combination of surface conditions and materials that is unique to Venus [e.g., Baker et al., 1992b; Kargel et al., 1994], their value as possible analogs for Martian channels remains clear. In this paper, lunar and Venusian landforms are used to suggest that limitations previously placed on the competence of lava flows to form Martian outflow channels may not be valid and that future studies of Martian outflow channels should more actively consider the flow of lava as a candidate mechanism for channel formation.
2. Outflow Channels on Mars
 Martian outflow channels may be broadly defined as relatively large channels that head in regions of chaotic or otherwise disturbed terrain, merge downstream with plains units, and have characteristics suggestive of fluid erosion [Mars Channel Working Group, 1983], although some features commonly designated as outflow channels lack one or more of these characteristics. For the purposes of this paper this definition will be taken to include channels such as Mangala Valles, Ma'adim Vallis, and the large channels of Elysium and eastern Hellas [e.g., see Carr, 1995] in addition to major outflow channels such as Ares Vallis. The outflow channels of Mars are generally located equatorward of ±30° latitude [Mars Channel Working Group, 1983] and are characterized by a variety of sizes, morphologies, ages, and relationships with surrounding terrain features [e.g., Masursky et al., 1977]. The largest outflow channel system, Kasei Valles, has a length of ∼3000 km and maximum width of ∼230 km.
 The view that the flow of lava could not have formed the features of outflow channels of Mars has ultimately been founded upon the assumption that the nature of these features is indicative of [Milton, 1973, p. 4040] “active erosion of material, which would not be accomplished by lava” [see also Mars Channel Working Group, 1983; Baker et al., 1992a]. Additional assumptions said to invalidate the volcanic hypothesis for formation of outflow channels include the following: (1) Lava channels should have obvious depositional features at their mouths, (2) lava channels are unlikely to form as parts of complex flow networks, and (3) Martian outflow channels head in regions that lack obvious volcanic sources [e.g., Mars Channel Working Group, 1983; Gulick and Baker, 1990; Baker et al., 1992a; Carr, 1996]. In sections 3 and 4, examples of lunar and Venusian features formed by the flow of lava are used to suggest that the above assumptions may not be valid.
 Many workers have concluded that the flow of lava can only act to construct landforms and not to incise or otherwise erode them, and this view has been an important factor in the elimination of lava flow as a candidate mechanism by which Martian outflow channels might have formed. However, incision and erosion by the flow of lava are suggested by valleys such as those in the region of lunar crater Plato, where lava once flowed from local sources across highlands onto the surrounding basalt plains of Mare Imbrium and Mare Frigoris [e.g., M'Gonigle and Schleicher, 1972]. This flow resulted in formation of valleys up to ∼3 km across and formation of nested channels in valley fill deposits (Figure 6). It is apparent that the valleys formed as sinuous lava channels after formation of surrounding uplands, and although a constructive origin for parts of these channels is conceivable (e.g., where lava has flooded relatively broad topographic lows), the action of erosive processes is also suggested. Channels nested within the larger valleys may have formed by incision, although a dominantly constructive origin related to emplacement of volcanic valley deposits is possible. Numerous similar examples of channels incised across lunar uplands are located in, e.g., the Aristarchus Plateau, Prinz crater, and Mare Orientale regions [e.g., Hulme, 1973; Greeley, 1976; Zisk et al., 1977; Strain and El-Baz, 1977].
 Most workers have concluded that the flow of lava is incapable of forming Martian features such as streamlined islands and anastamosing channels [e.g., Mars Channel Working Group, 1983; Baker et al., 1992a]. Yet immediately north of the lunar Aristarchus Plateau, there are examples of streamlined islands that formed by lava flow (Figure 7). Importantly, it is uncertain if these features formed during the same event in which surrounding basalt units were emplaced or if erosion of preexisting terrain took place by confined flow; nevertheless, the final products are streamlined islands that developed by lava flow in the complete absence of water. Streamlined islands and anastamosing rilles exist as part of a large network of lava channels (possibly formed through dominantly constructive processes related to emplacement of broad lava flows) in Mare Imbrium (Figure 8), demonstrating that lava conduits need not be restricted to simple sinuous channels that lack tributaries and that do not branch and recombine in a complex manner.
 Erosion by lava is apparently not restricted to the simple vertical incision implied at some lunar sites: Lateral erosion by the sinuous flow of lava is suggested at Hadley Rille, where some meanders appear to cut into adjacent highland massifs [Howard et al., 1972]. High sinuosity is a characteristic of numerous lunar rilles [e.g., Young et al., 1973; Wilhelms, 1987], and although many such rilles may have formed at least partly through constructional processes, they may also have evolved over time through lateral erosion during eruptive events [e.g., Hulme, 1973]. Examples exist of lunar channels with terraces that may have been modified by later lava flows (Figure 9), suggesting that volcanic processes can both form and modify such features.
 Absence of large positive relief lava deposits at the mouths of Martian channels has been cited as key evidence against formation of these channels by the flow of lava [Baker et al., 1992a]. Yet it is typical for lunar sinuous rilles to entirely lack such distinct accumulations of lava flows at their terminations [El-Baz et al., 1972], a characteristic that can be attributed to processes such as burial of channel mouths by subsequent lava flows. For example, Vallis Schröteri, which extends from the Aristarchus Plateau onto the basaltic plains of Oceanus Procellarum, lacks association with a clear terminal volcanic deposit (Figure 10). Any perceived lack of volcanic deposits at the mouth of this or other lunar rilles (e.g., Figure 6a) would be incorrect: Terminal accumulations of materials that once flowed through lunar rilles must, in fact, be present in large volumes and, in combination with flow materials derived from other sources, must collectively comprise the materials of both the distal channels and the volcanic plains onto which they flow. In this case, as with many others on the Moon, the large volumes and extremely low viscosity of lunar lava flows [e.g., Head, 1976; Hörz et al., 1991] appear to have inhibited preservation of morphologically distinct local accumulations of terminal deposits, favoring instead very broad volcanic deposits that collectively fill substantial volumes of basins and bury any such accumulations that may have once been exposed at the surface [El-Baz and Worden, 1972]. The supposed absence of deposits at the mouths of many Martian outflow channels may similarly be illusory. Most Martian outflow channels extend onto the extensive volcanic plains of the northern lowlands [e.g., Masursky et al., 1977; Carr, 1995], a relation consistent with volcanic origins for these channels. On the basis of lunar analogs such as Vallis Schröteri the typical transition of Martian outflow channels from sharp channel margins in highland regions to indistinct margins on terminal volcanic plains [Carr, 1995] is also consistent with volcanic origins.
 Many Martian outflow channels commence abruptly at topographic depressions such as pits, regions of chaotic terrain, and volcanic fissures. Such regions are often interpreted as areas where volcanic materials and water interacted, in some cases explosively, and from which large volumes of water were released to form associated channels [e.g., Masursky et al., 1977; Baker et al., 1992a; Tanaka and Chapman, 1990; Burr et al., 2002; Wilson and Head, 2002; Wilson and Mouginis-Mark, 2003; Head et al., 2003; Manga, 2004]. However, many lunar sinuous rilles similarly commence abruptly at large topographic depressions that clearly mark the source areas of volcanic flows that formed the rilles [e.g., Greeley, 1971a; Zisk et al., 1977; Strain and El-Baz, 1977] (Figures 6 and 11). If, in the absence of water, volcanism on the Moon can form lava channels that head at large topographic depressions, there is no requirement to invoke the action of water in the formation of either Martian depressions or associated channels. Concentrations of surface and near-surface volatiles are evident on Mars [e.g., Kieffer et al., 1976; Boynton et al., 2002; Feldman et al., 2003], and it follows that significant and sometimes explosive interactions between such volatiles and igneous materials are likely to have taken place during the planet's history. However, on the basis of lunar analogs, large topographic depressions at the heads of channels need not represent the signature of phreatomagmatic eruptions that involved catastrophic releases of large volumes of water from cryosphere-confined aquifers.
 Many Venusian channels are simple in form, being characterized by long channels with few or no tributaries and having general morphologies that are similar to those of simple sinuous lunar rilles [Baker et al., 1992b, 1997; Head et al., 1992; Komatsu and Baker, 1994]. Simple channels on Venus generally head at topographic depressions, have relatively constant widths of ∼1–2 km and lengths of tens to hundreds of kilometers, and lack clear associations with lava flow margins [Baker et al., 1992b, 1997; Head et al., 1992]. The longest recognized channel on Venus, located mostly on lava plains located north of Rusalka Planitia, has a total length of 6800 km [Baker et al., 1992b]. Features associated with simple channels on Venus include meander cutoffs, abandoned channel segments, and levees [Baker et al., 1992b; Head et al., 1992]. The meanders of simple Venusian channels are suggestive of modification of original flow patterns by erosional channel widening during lava flow [Komatsu and Baker, 1994]. As with their lunar counterparts, there are numerous examples of simple Venusian channels that cut across irregular topography, suggesting that formation of these channels involved erosive processes [e.g., Komatsu and Baker, 1994] (Figure 12).
 Complex channels, which in numerous cases develop large braided patterns that define streamlined islands (Figure 13), are also found on Venus [Baker et al., 1992b]. Many of these channels head at topographic depressions, and a minority of channels have large delta-like distributary systems at their mouths [Baker et al., 1992b]. A prominent complex Venusian channel, Kallistos Vallis, is located in the northern region of Lada Terra, east of Lavinia Planitia. Kallistos Vallis is over 1200 km long and up to 30 km wide and heads at a collapse feature on the flank of a large volcanic construct [Baker et al., 1992b; Komatsu et al., 1993]. This collapse feature is morphologically similar to regions of chaotic terrain that acted as sources for many outflow channels on Mars [Baker et al., 1992b]. The features of Kallistos Vallis, which include large streamlined islands (Figure 14), are strongly suggestive of formation by flood waters, although a volcanic origin is widely favored on the basis of the nature of Venusian landforms and environmental conditions [Baker et al., 1992b, 1997].
 The existence and properties of complex volcanic channels on Venus are relevant to the search for the origins of outflow channels on Mars. Complex Venusian channels represent evidence that volcanic processes can form chaotic terrain and that the flow of lava from such terrain can form large channels and streamlined landforms; these attributes are consistent with those of smaller but analogous lunar features. Importantly, Venusian channels suggest that the flow of lava can produce landforms that mimic those known to have been formed on Earth by aqueous flood events.
 Some workers have hypothesized that formation of complex channels on Venus may have been a consequence of the planet's high surface temperatures and of particular fluid chemistries such as liquid sulfur or ultramafic compositions [e.g., Baker et al., 1992b]. Appeals to special rock chemistries and environmental conditions to justify the existence of Venusian channels may be unnecessary, however. A simpler interpretation of Venusian channels involves a recognition that basaltic lava, erupted in sufficiently large volumes and at sufficiently low viscosities, is capable of formation of large channels through processes that include erosion. This interpretation is consistent with measurements returned by the Venera and Vega landers [e.g., Surkov, 1983; Surkov et al., 1987; Kargel et al., 1993; Fegley et al., 1997], the seemingly basaltic nature of volcanic landforms on Venus [Head et al., 1992], and the nature of lunar basaltic landforms such as those described above.
 Although the basic processes and materials involved in the formation of Venusian channels remain poorly understood, the channels themselves are nevertheless very likely to have been formed by the flow of low-viscosity lava. Some Venusian channels formed as distributary features of lava flows and do not appear to incise surrounding terrain and, as such, seem to have constructional origins [e.g., Baker et al., 1997]. Other channels lack the characteristics of constructional features (such as association with adjacent flow deposits) and are suggestive of incision [e.g., Komatsu and Baker, 1994; Baker et al., 1997]. Complex Venusian channels have characteristics and features that are highly suggestive of formation by erosion. The large outflow channel Kallistos Vallis is suggestive of erosion on a massive scale, and the further association of this channel with a source at chaotic terrain is significant to the interpretation of corresponding Martian landforms.
 Although the scales of lunar volcanic channels do not approach those of the largest Venusian and Martian channels, lunar analogs nevertheless provide an important and independent perspective on the capacity of lava to form features such as those of Martian channels. Lunar analogs suggest that (1) basaltic lava flows can form large channels through processes that likely include incision and erosion; (2) basaltic lava flows can form anastamosing drainage systems as well as streamlined islands and channel terraces; (3) volcanic channels need not have obvious positive relief accumulations at their mouths; and (4) the association between volcanic channels and sources at local topographic depressions is common and existence of such channels and depressions need not imply the action of phreatomagmatic processes involving catastrophic releases of large volumes of water. These points are fully consistent with the nature of Venusian volcanic channels and similarly suggest that flow of lava should be more actively considered as a viable candidate process for formation of Martian outflow channels.
 With typical widths of meters to tens of meters and lengths of hundreds of meters to tens of kilometers, terrestrial lava channels are generally much smaller than their lunar and Venusian counterparts [e.g., Wentworth and Macdonald, 1953; Ollier and Brown, 1965; Carr and Greeley, 1980; Hon et al., 1994]. Terrestrial lava channels nevertheless share a number of characteristics with sinuous lunar and Venusian rilles. For example, terrestrial channels are frequently simple in morphology and often head at volcanic sources that are (or that ultimately develop into) distinct topographic depressions [e.g., Greeley, 1977]. Importantly, terrestrial channels generally form as constructive distributary channels that emplace lava flows [e.g., Hon et al., 1994], and there is a distinct paucity of clear evidence in support of the action of significant erosive processes in modern flows [e.g., Greeley et al., 1998]. No recognized terrestrial lava channels share the erosive characteristics or dimensions of the most intriguing lunar and Venusian channels discussed above. Modern terrestrial eruptive events [e.g., Hon et al., 1994] do not involve the lava volumes believed to have been associated with large lunar and Venusian events [e.g., Head, 1976; Crumpler et al., 1997], and thus the absence today of large erosive terrestrial channels should perhaps not be surprising. While the capacity of ancient terrestrial lava flows to incise channels is poorly understood [e.g., Fagents and Greeley, 2001], the preserved features of some ancient channels are suggestive of a past competence for erosion of bedrock by terrestrial lava flows of various compositions [e.g., Huppert et al., 1984; Huppert and Sparks, 1985; Barnes and Barnes, 1990; see also Kerr, 2001]. Even modern basaltic lava flows hint at capacities for formation of anastamosing systems and incision of channels. For example, some terrestrial lava channels, such as tubes at Mount Etna [Calvari and Pinkerton, 1998] and lava streams in Hawaii [Carr and Greeley, 1980], can form distinctly complex networks with anastamosing and braided reaches. Some recent and modern terrestrial channels are believed to have formed at least in part through thermal or mechanical erosion by lava flow [e.g., Peterson and Swanson, 1974; Greeley et al., 1998; Kauahikaua et al., 1998, 2002, 2003; Williams et al., 2004]. Most notably, a small Hawaiian lava stream, with a width of only 4 m, is documented to have incised 6 m into bedrock over a period of 60 days [Kauahikaua et al., 1998]. In light of terrestrial measurements such as these a capacity for substantial vertical incision may not be unrealistic for hypothetical Martian high-volume and low-viscosity lava flows with channelized widths of kilometers to tens of kilometers.
 The landing sites of the Viking 1 [e.g., Binder et al., 1977; Greeley et al., 1977], Pathfinder [e.g., McSween et al., 1999; Chapman and Kargel, 1999], and Spirit [e.g., McSween et al., 2004] spacecraft, all of which are located near the mouths of prominent Martian outflow channels, are plains with properties that are entirely consistent with volcanic origins. In a manner similar to that of lunar volcanic plains visited by the Luna, Surveyor, and Apollo landers [e.g., Shoemaker et al., 1966, 1969, 1970; Schmitt and Cernan, 1973; Vaniman et al., 1991; Hörz et al., 1991], surface materials at the three Martian landing sites appear to consist primarily of the remnants of volcanic units that have been subjected to extensive reworking by impacts. As on the Moon, formation of a surface regolith layer at these sites has apparently reduced original volcanic units to deposits of rock fragments and fines and eliminated the outcrop-scale volcanic textures and structures that were likely once expressed at the surface.
 Consistent with a volcanic origin for the largest outflow channels, smaller outflow channels that head in highland regions outside of Tharsis and Elysium (e.g., those in Memnonia, Aeolis, and Hellas) are typically associated with major rift zones (e.g., Memnonia Fossae [Tanaka and Chapman, 1990; Zimbelman et al., 1992]), volcanic rises (e.g., Tyrrhena Patera [Greeley and Crown, 1990]), and large volcanically emplaced plains units [e.g., see Scott and Tanaka, 1986; Greeley and Guest, 1987; Smith et al., 2003]. As with larger outflow features, channels in these regions ultimately flow onto extensive volcanic plains. Even relatively small sinuous valleys and valley networks located in the Martian highlands [e.g., Goldspiel et al., 1993; Carr, 1995; Carr and Malin, 2000] appear in some cases to have characteristics that are more consistent with volcanic than fluvial or sapping origins [e.g., Leverington and Maxwell, 2004; Leverington, 2004], suggesting a possible consistency in the origins of certain Martian channels across a wide range of scales. A minority of outflow channels on Mars (e.g., Ma'adim Vallis) may not have originally formed through incision by the flow of lava, instead forming through other processes and later acting as conduits for large highland lava flows; such a sequence of events would be similar to that hypothesized by Leverington and Maxwell  for a smaller north sloping valley in western Memnonia.
 The simplest interpretation of Martian channels that extend from volcanic source regions onto volcanic plains is as conduits formed by volcanic flows. Volcanic processes operate under an extremely wide range of environmental conditions and thus can account for channel formation without appeals to more complex scenarios involving substantial past changes in Martian climate or concepts such as hemispheric aquifers confined under a global cryosphere. Channel formation hypotheses requiring major changes in the properties of the Martian atmosphere or involving repeated catastrophic flow of water from features such as volcanic rifts appear needlessly exotic alongside the igneous hypothesis for channel formation. This hypothesis requires only that large flows of lava have a capacity for erosion and for formation of streamlined landforms and anastamosing channels, a capacity that is implied by the nature of lunar and Venusian volcanic landforms.
 Processes related to the flow of lava have been previously rejected as possible mechanisms by which Martian outflow channels might have formed. This rejection has mainly been based on assumptions that lava flows could not have eroded channels and formed Martian features such as streamlined islands and anastamosing reaches, that formation by lava flows should have resulted in accumulation of large positive relief deposits at the mouths of Martian channels, and that volcanic sources are absent at the heads of Martian channels. However, there are examples on Venus and the Moon in which the flow of lava appears to have resulted in formation of anastamosing channels and streamlined islands. As with many Martian channels it is typical for lunar rilles to appear to lack large accumulations of volcanic materials at their mouths, though in reality such deposits must certainly be present among the voluminous volcanic plains present at the termini of these channels. Many lunar and Venusian volcanic channels commence abruptly at large topographic depressions, a characteristic shared with most Martian outflow channels. Large lava plains are widespread in both the highlands and lowlands of Mars, representing evidence that massive effusive volcanic events have taken place on that planet. Volcanic processes operate under an extremely wide range of environmental conditions and thus can account for channel formation on Mars without appeals to complex scenarios involving dramatic changes in atmospheric properties or concepts such as hemispheric aquifers confined by a global cryosphere. Formation of outflow channels by aqueous floods remains an important hypothesis for which there is at least partial support from terrestrial analogs. However, on the basis of lunar and Venusian analogs the flow of lava should also be treated in future research as a viable and realistic mechanism by which Martian outflow channels might have formed.
 I thank Devon Burr and Goro Komatsu for their helpful and constructive reviews. This work was supported by the Smithsonian Institution and by grant NAG5-13300 from the NASA Mars Data Analysis Program.