Counterfactuals and revisionism in historical explanation

This article addresses the role of counterfactuals in historical and ethnohistorical explanation. Drawing primarily on examples from the Conquest of Mexico, it argues (1) for a useful role for counterfactual analysis, not in writing fiction, but in assessing pivotal causation and proofing causal arguments; (2) for a clearer understanding of causation in historical records, especially of ethnohistorical subjects; and (3) for a way in which this perspective can be employed to argue for solidly grounded revisionist interpretations of events.

anthropologist? The growing historicization of anthropology (e.g., Carmack, 1972: 228;Faubion, 1993: 35-44;Krech, 1991;Schwerin, 1976: 323) has made such issues increasingly significant for the discipline, yet at the same time, there is all too often a tendency to treat history as simply an anecdotal illustration of a larger theoretical point, as just another variable, or as a convenient backdrop against which to situate more conventional ethnography. There are, however, critical issues in and about the past for which an anthropological perspective can be particularly useful, especially in studies of non-Western peoples (Simmons, 1988: 2;Trigger, 1975Trigger, : 51, 1982. But attempts so to use it have often been marked by inadequate methodological training in carrying out historical inquiry, thereby frequently yielding unsatisfactory results (Carmack, 1972: 232;DeMallie, 1977: 78-9;Euler, 1972: 202;Lévi-Strauss, 1963: 1). And I would suggest that this failure has been one of the reasons that many anthropologists, dissatisfied with the prevailing interpretations, have turned away from causal analysis and sought answers in meanings, even though they do acknowledge that actions have causes (e.g., Geertz, 1983: 34). But whether meaning yields a better understanding of actions than causal forces, both approaches -meaning and cause -rest on an initial historical assessment that, if fundamentally flawed, will invariably yield an inadequate interpretation, and the meaning approach has little internal basis for correcting such errors. So the fundamental concern here is, how can better historical analyses be achieved, especially from an anthropological perspective?
Though typically ignored and by no means a complete panacea, counterfactual analysis provides an excellent basis for critically examining historical causes, regardless of the subsequent theoretical stance adopted. And the approach advocated here is, first, the use of counterfactual analysis, both of itself to strengthen causal arguments and, second, to delineate explicitly its elements as the basis for revisionist interpretations.

I. COUNTERFACTUALS
The fundamental question underlying counterfactual speculation is whether past events could have taken a different course than the one they did, with the answer having implications for both historical and ethnographic research. While the response to this question is necessarily hypothetical, a counterfactual approach attempts to show plausible instances in which the road not taken very well could have been, with significant consequences for the subsequent course of history, whether the ending point of interest is the past or today. One goal, then, is to use counterfactual analysis to 'proof ' one's interpretation by altering the presumed cause and attempting to determine whether or not the outcome would be changed as a result.
Attempting to write counterfactual history essentially adopts the realist position that the past is fixed and certain (Kuzminski, 1979;Walsh, 1977) though its interpretation is not, nor need it be considered as determined (Ferguson, 1997). A constructionist position, in which the past is taken to be uncertain and wholly dependent upon one's interpretative stance (Goldstein, 1977;Nowell-Smith, 1977), leaves little room for counterfactual analysis, or perhaps, more cynically, it could be argued that these positions are inherently counterfactual and substantiate the plausibility of investigating such alternatives. While events can be taken as fixed after the fact from a realist perspective, history as it is being made is inherently contingent. When people are actually faced with course-altering decisions, many factors influence which one is chosen, but the fact that one is ultimately selected over the others does not obscure the reality that many different alternatives are or were available. After selecting one course of action, the alternatives effectively become counterfactual, but at the time the choice is made, a world of possibilities is open. That is, life as it is lived is contingent and not nearly as determined as either ex post facto ethnographic accounts or the historical record would suggest.
When faced with a choice, one of many alternatives is selected and so becomes historical reality. The alternative courses of action were, nevertheless, to varying degrees, plausible and viable; they were not simply fanciful events, but potential realities that did not come to be. And it is precisely that the paths not trodden are considered that makes them reasonable counterfactuals. Counterfactuals are plausibilities before the action was taken that fixed the subsequent sequence of events. Counterfactuals are thus not fantasy alternatives but considered choices in light of probable consequences, rather than unknowable ones.
Although the explicit writing of counterfactual history is not common, the intellectual exercise is. In the here-and-now, we all think counterfactually -not in the sense of projected possible courses of action in the past, but rather in the present weighing of alternatives, all except one of which will ultimately become counterfactuals. Were it otherwise, we would never recognize choices.

II. CAUSALITY
However entertaining the construction of counterfactuals may be, the goal of the exercise is not simply to create alternative fictional scenarios, but to focus more closely on the nature of causality in both ethnographic and historical explanation. Constructing a compelling counterfactual account depends on identifying the proximate cause in the traditional narrative, the pivotal event without which the known overall sequence of events would not have occurred. For example, if I consider the Spaniards' possession of firearms as crucial to their conquest of Mexico, imagining a plausible way to neutralize these weapons should reverse the outcome. Thus, if the battle at Tlaxcallan had continued for a few more days, the Spaniards' entire supply of gunpowder would have been exhausted, effectively depriving them of their firearms thereafter and insuring their defeat. But altering an event that then does not substantially change the outcome illustrates an erroneous choice, as what was identified as the cause was then clearly not crucial. For instance, since the Spaniards abandoned their cannons during the flight from Tenochtitlan on 30 June 1520, yet returned to attack again, and since they also ran out of powder in the final weeks of the Conquest, but pursued their campaign to a successful conclusion, it does not appear that firearms were, in fact, essential. So while the initial theory of the pivotal significance of firearms seemed reasonable, other data suggest that reconceiving subsequent events in the absence of guns would not crucially alter the course of events as we know them. This, then, forces us to reconsider what is actually explanatory in the historical account. And while this counterfactual experiment failed, having considered it has sharpened our awareness of what was actually causal and significant.
Given the many possible counterfactual alternatives, on what basis can one be argued to be more plausible than another since none in fact actually happened? The answer is that the best counterfactual is the one that requires the fewest changes to produce an altered sequence of events. Altering more than one variable introduces so many alternative consequences that the projected causal chain becomes so complex that it is difficult to establish a single plausible line of reasoning.
Among those that are not compelling counterfactuals are changes that fundamentally reorder the reality that is historically known. For instance, a counterfactual Mexico in which the Aztecs had organized their imperial alliance system in an altogether different manner that would have allowed them to repulse the Spaniards permanently would produce a result contrary to that which is historically known. But it would also fundamentally alter Aztec society in ways that could not easily be foreseen, and so Aztec actions in the face of a Spanish threat could not be convincingly predicted. Altering fundamental conditions yields circumstances that are known, or knowable, to all and affect everyone's actions accordingly. 1 To illustrate, it might be said that the D-day invasion would have failed had there been clouds over coastal France because this would have hindered allied air support and allowed the Germans to move their armor to counter the landings. And this may be a cogent analysis, if the presence of overcast skies were the only variable changed from actual D-day events. However, weather is a condition that is generally known, and both sides would have adjusted their behavior accordingly, so this example actually changes multiple variables, which makes its use in analysis vastly more difficult.
Another inadequate counterfactual is one that changes a circumstance so that it alters the outcome of a particular situation, yet does not fundamentally alter the course of events as we know it. For instance, if Cortés had been killed during the Conquest of Mexico, his entire campaign would probably have been thwarted. But if only Cortés had been killed, leaving the bulk of his force intact, and it had returned reinforced under alternative leadership, intent on conquest, eliminating Cortés would not have achieved a permanent shift in circumstances. A counterfactual change that cannot achieve more than an incidental shift in the course of events is not worth consideration except to demonstrate the improbability of one possible course of action, or to illustrate that this particular point in history was not pivotal. The type of counterfactual that allows the most plausible alternative argument is one that alters a decision or changes an event in a way that would have been unpredictable by all of the participants. In the case of a counterfactual that rests on the decision of a single leader, the general alternatives may be known or suspected by everyone, but until a choice is made, its consequences have little or no effect on others. For instance, at one point in World War II, invading Normandy was merely one of several alternative courses of action. But once the Allies made their choice, it affected all subsequent behavior. Had they chosen another site, that decision would likewise have altered allied efforts in a manner tailored to that location. But in such a case, where a single person or small group of people are entrusted to make the decision and empowered to carry it into effect, only the deciding actor(s) know(s) the course of action to be followed until it is revealed or carried out, and other parties, such as the enemy, will not be aware of it and adjust their own actions accordingly.
Counterfactuals that depend on altering an historically known course of action allow the construction of more convincing arguments than those that require opening up heretofore unconsidered possibilities. For example, because Cortés personally held his band of Spaniards together, killing him in a plausible manner and circumstance would most likely have led to consequences that would have thwarted the Conquest as we know it by bringing about the death of many of his followers and the end of his expedition. The example that best illustrates this in the case of the Conquest, and best fits the minimal rewrite rule, occurred during the Tlacopan causeway battle.
Six weeks into the siege of the Aztec capital of Tenochtitlan, Cortés led his forces across one of the causeways leading from the lake shore to the city. The Spaniards had been fighting their way along this causeway since late May, 1521, in a series of seizures and withdrawals, assailed by troops in front and soldiers in canoes on each side. Much of the fighting centered on large gaps created by the removal of wooden bridges, opposite which the Aztecs built barricades. On 30 June, Cortés and his men fought their way across several such breaches and, appearing to push the Aztecs back, crossed another. In pursuit of the fleeing Aztecs, the Spaniards neglected to fill in the gap, despite Cortés's earlier orders to do so. The flight was a feint and once Cortés and his allies crossed the breach, the Aztecs counterattacked. The Spaniards were pushed back against the gap in disarray, Cortés was wounded in the leg and seized by four Aztecs who were dragging him off when Cristóbal de Olea came to the rescue, killed the four who held Cortés, and freed him, though at the cost of his own life. Cortés escaped, but 68 other Spaniards were taken captive and many more were killed (Aguilar, 1977: 96;Díaz del Castillo, 1977, Vol. 3: 33, 53;Durán, 1967, Vol. 3: 565-6;Ixtlilxóchitl, 1975-7, Vol. 1: 472-3;López de Gómara, 1965-6, Vol. 2: 262-3;Oviedo y Valdés, 1959, Vol. 4: 133;Sahagún, 1975: 104;1989: 121).
Ten of the Spaniards were immediately sacrificed at the Great Temple and their severed heads were sent to the battlefront and thrown into the Spanish lines. And that night, from their camps, Cortés's men could hear the drums from the Great Temple and see the other captured Spaniards being made to dance in front of the god Huitzilopochtli before their hearts were cut out in divine sacrifice. The faces were flayed, tanned, andbeards attached -sent to the Aztecs' tributary towns as a warning (Cortés, 1963: 241-2;Díaz del Castillo, 1977, Vol. 3: 34-6;Ixtlilxóchitl, 1975-7, Vol. 1: 472-3;López de Gómara, 1965-6, Vol. 2: 262-3;Oviedo y Valdés, 1959, Vol. 4: 133).
In the wake of this success, most of the Indians allied with the Spaniards left, and the Aztecs besieged the Spanish camps for the next four days. But Cortés held his men together, adopted a defensive posture where his weapons could be used to best effect, weathered the onslaught, and, when his Indian allies saw that the Spaniards had not been defeated, they came back, the tide of battle turned, and the Conquest continued (Díaz del Castillo, 1977, Vol. 3: 41-5, 47-9, 51-2). But this need not have been the case. Without Olea, Cortés would have been captured and killed along with the other 68 Spaniards, the Indian allies would still have fled, and they would have had no reason to return. In actual fact, they did return, but only because Cortés prevailed. Had he been killed, factionalism among the remaining Spaniards would likely have split the camp, forcing a withdrawal to the coast and a retreat to Cuba.
A counterfactual that opens up new possibilities, such as supplying Richard III with his much sought-after horse, offers so many potential developments that causal sequences rapidly multiply beyond anyone's ability to predict with much assurance. In the case of Cortés's death, the counterfactual cuts off a known causal chain and, while it gives rise to many new lines of action, it is definitive concerning the course of action stopped, whereas altering Richard III's circumstances creates new, unknown, and largely unpredictable possibilities without necessarily effecting any significant change in the known course of action. But where does this leave us in counterfactual analysis beyond sharpening our focus on what is or is not pivotally causal?
The counterfactual experiment alters events or circumstances; the focus, however, is not on facts per se, but on cause. Cause, however, is not a fact, but is the analytical assessment of an historical relationship, the considered assessment of causality and consequences. And while the cause of an event might seem obvious, a significant event is rarely apparent from a given action, but rather from that action's consequences, many of which are known only far in the future. So while counterfactuals always focus on altering pivotal causes, these are identifiable only as a result of their consequences, not in and of themselves. And it is this historical notion of causation, akin to motivation rather than to a direct mechanistic cause and effect, that precipitates many of the problems in historical analysis.
To change a fact to create a counterfactual analysis, one must select out of the innumerable circumstances of the day a meaningful event or condition, not a physical event or condition but rather one that is significant to the people involved. For an event to be meaningful, it must have been taken into consideration by other historical actors and have influenced their subsequent actions. Thus, the focus must be on facts as they appear to the actors in that culture, not necessarily to the analyst. Moreover, since norms and actions often differ, in identifying what is meaningful to actors in another culture, the focus must be primarily on behavior rather than on verbal claims or written statements.
There are, however, real limits to our ability to assess actions, both historically and ethnographically. Just as particle physics has the Heisenberg Uncertainty Principle, which states that a particle's position and its velocity can each be determined but not at the same time -hence the uncertainty -so too does ethnohistory have an uncertainty principle.
The analyst is dependent upon the vagaries of record keeping, document survival, and the good fortune of discovery to reveal historical events. The ethnographer, by contrast, can actively inquire, although some things, such as wealth or ethnic background, may be more difficult to discover during someone's lifetime than posthumously. The importance of an event or circumstance, however, depends not on the thing itself but on its consequences, with the result that a significant occurrence can only be recognized historically. For instance, ethnographers may focus on a murder in their village as the most significant event because they believe it is important, not of itself, but in terms of its longer-range consequences based on the impact such events have had elsewhere. But in that same village, an Abraham Lincoln or an Adolf Hitler might be born and the ethnographers will have missed this more significant event because its importance cannot be recognized immediately. It becomes important far later, as a result of the adult's subsequent actions. So, while ethnography offers the best data, history offers the best questions, and the two can never be completely brought together. And these difficulties are magnified in the field of ethnohistory.

III. ETHNOHISTORY AND COUNTERFACTUALS
While much counterfactual analysis presumes 'perfect knowledge', these idealized counterfactuals do not fit well with the non-Western groups with which ethnohistorians deal because often the data are poor. 2 While the focus on pivotal causes can be addressed with a certain degree of confidence for modern Western history, where the data are generally good, and even excellent for political actions, the same is often untrue outside the Western, especially where there is no indigenous literate tradition.
Some counterfactuals are more plausibly situated in the recorded accounts of the event in question, and it is important to differentiate between those for which the data are fuller and those for which they are not. A counterfactual in the latter case is significantly less compelling because there are many logical possibilities and little support in the historical record to focus on the one proposed: they are, in effect, possibilities rather than bona fide counterfactual probabilities.
In the case of ancient Mexico, virtually nothing can be answered by recourse to an official transcript of debated actions and considered alternatives. The data are few and often questionable, and all causal explanation hangs on how events and their significance are reconstructed. And given the radically divergent cultural traditions of Spaniards and Aztecs, even when the focus is on such a seemingly fixed event as Cortés's landing on the Veracruz coast, it is necessarily interpreted in quite different ways, because each side does so from the perspective of its own cultural background. As a result, an analysis of a pivotal event in the causal chain as seen within one cultural framework will not necessarily focus on the essential cause from the other cultural perspective. This is not fatal to historical analysis, since, after all, the history as written is the historians', not the Spaniards' or the Aztecs', and it is their assessment of pivotal causality that is being evaluated. But it does have a major impact on the data, since each culture will record (or ignore) events in light of their significance for itself, which will most probably produce very different records that are then likely to skew the interpretation of the analyst. 3 A great number of counterfactual examples rely on the unobservable, such as the mental act of considering and selecting alternative courses of action (e.g. Clark, 1997). But these intellectual considerations are seldom reflected in the Western historical record of non-Western societies, which suggests that counterfactual interpretations that are based on the private deliberations of an individual are unconvincing for such societies. Ethnohistory tends to rely more on culturally congruent causes than on individual decisions precisely for this reason. And as long as it offers recorded patterns of behavior, it should yield plausible counterfactual analyses.
While ethnohistory does suffer from data limitations, it often benefits by the use of comparison. Lacking considered alternative courses of action in the specific historical case being studied, the tendency is to fall back on comparable examples. By examining alternative choices made by other groups under similar circumstances, a range of alternatives is presented that is effectively analogous to considered, recorded alternatives. Moreover, these have some advantage over considered, but rejected, alternatives, because the former establish that these were viable alternatives adopted by someone and they offer real consequences rather than merely counterfactually deduced ones. In effect, then, counterfactuals are a form of projected comparative analysis that relies on similar situations in other places and other times, so that the likely outcomes of the choices not made can be known.

IV. COUNTERCAUSALITY
One of the seeming paradoxes in counterfactual analysis is that while it focuses on changing causes, effects are more apparent than causes in historical records. Moreover, it is effects that explanation seeks to account for, with causes invoked only in relation to them as explanatory. Causes are not always obviously significant when they occur, and because they are not always immediately perceived as important, they are less likely to be recorded when they occur (and if they are, it is typically in a sketchier version) than are effects. Moreover, there is little need to account for causes when they occur because they need only be explained in relation to their effects. And the farther cause and effect are separated in time, the less will be recorded and known of the former.
As odd as it may seem at first glance, distinguishing cause from effect is not always easy. Every cause is, presumably, also an effect, yet we label them causes because of their relationship to the effect of interest. Effects are the focal events and are thus defined as effects, and once this has taken place, the search is then on for their causes. In essence, effects are effects because we seek their causes, and we seek these because we want to explain events that we see as having an impact on us. So first and foremost, pivotal events are such because we consensually agree that they are. And we do so because we share basic epistemological notions about history and causation, and the perception of pivotal events is our common cultural currency.
Although effects are what is important in history, our counterfactual analyses focus on cause because we believe we already know the crucial historical link between the cause and its effects. So while counterfactuals nominally change causes, their real focus is on the effect: the former is altered to change the latter in order to produce different consequences. But while we know the effect, or it would not have been the important focus of the inquiry, the same is not true of the cause. The cause is generally discovered only by reasoning back from the effect.
Analytical history is the attempt to link effects to causes, but in feeling our way blindly through the historical record, the most impressive bumps -in records as well as in impact on the people -are the effects, and we then must tentatively work our way back to the less impressive, more poorly recorded causes. And if this is not readily apparent in counterfactual analyses, it is largely because most examples of that genre occur in data-rich historical periods, such as modern Western history, where the difficulties in discerning apparent causes are minimized. 4 Moreover, because history, common understanding, and causal reasoning emphasize a chronological sequence, the normal manner of presentation, from the cause forward, obscures the fact that the route of research is from effect backward to their causes. And it is precisely in ethnohistory, which typically deals with peoples who have a relatively impoverished historical record, that the distortion of causal reasoning is greatest. The data are the least secure and the likelihood that the cause imputed in the analysis of a different culture is in error is the greatest.

V. CAUSAL REVISIONISM
The essence of traditional counterfactual analysis is the identification of the pivotal causal points in an historical explanation and its alteration and the consequent change in effect to illustrate the essentially contingent nature of history. If a cause is altered in a counterfactual, then the established effect should change too, and if it does not, the 'cause' selected was not significant.
A major limitation of counterfactuals is not that it is impossible to reassess an historical situation, alter the causes, and draw new logical consequences from them, but rather that in doing so the anticipated counterfactual consequences will necessarily be logical ones. That is, they are limited to the consequences one can reasonably foresee. Unfortunately, the world frequently produces results that, while logical when seen after the fact, all too often are not logically anticipated in advance. While historical analysis emphasizes ex post facto causal reasoning in which both the probable and the improbable can be seen, counterfactual analysis demands predictive causal reasoning, which is notoriously more difficult and generally does not capture the improbable. And because counterfactual consequences can only be logical, anticipatable ones, counterfactual examples are inherently impoverished in comparison to the richness of history as it is lived, and its exercise eliminates much that actually happens in history.
But there is another way in which causes can be manipulated to produce new and potentially provocative insights in historical analysis. Instead of altering a cause to change its consequences, I propose changing the pivotal cause while holding the effect constant. This causal revisionism is similar to counterfactual analysis in altering the cause to produce a different explanation. But rather than accepting the standard interpretation as true and assuming that a known causal link is being disrupted, causal revisionism challenges that interpretation, contests that link, and substitutes an alternative cause. This is done, however, not to alter the effect and subsequent events, but to force a reinterpretation of these events as they are currently understood. I suggest this approach primarily for two reasons. First, the primary data are often poor and, since effects tend to be more securely documented in the historical record than causes, the former are more apparent than the latter and we can have greater assurance of their accuracy. And, second, this approach eliminates the need for the problematic, predictive causal reasoning. In causal revisionism, the attempt is to produce a more satisfactory revisionist analysis through the use of reasoning drawn from counterfactuals to focus more on 'how did this effect arise' than 'would altering this cause change that effect'. Addressing the latter requires a solid knowledge of the cause-effect relationship in historical settings of a sort that often does not exist and which is not required in the former. One illustration of this approach is the Cholollan (Cholula) massacre of 1519.
During the Conquest of Mexico, Cortés allied with the Tlaxcaltecs and then marched one day south to the city of Cholollan. He was allowed to enter the city, but Marina, an Indian woman who accompanied the expedition, learned of a Chololtec plot to massacre the Spaniards, aided by a hidden Aztec army 50,000 men strong. She alerted Cortés, who acted first. He assembled the Chololtecs in the main courtyard, placed armed Spaniards at every entrance, and then massacred the enclosed and unarmed Indians.
Since Marina's warning prompted Cortés to carry out a preemptive attack, a typical counterfactual analysis might ask 'what would have happened if Marina had not learned of the plot or failed to inform Cortés of it'. All other things remaining constant, altering this cause would logically have led to the death of Cortés and presumably the abandonment of the Conquest of Mexico. But this counterfactual, logically reasoned result need not have happened. Perhaps someone else would have uncovered the plot, or Cortés might not have died so easily, or any of a number of unforeseeable events might have intervened to alter the outcome. So instead of presuming that changing the cause as proposed above would have flawlessly led to the failure of the Spanish Conquest in a traditional counterfactual of questionable utility, I propose a different type of counterfactual analysis, one that leaves the effect intact and therefore does not require the establishment of a tricky predictive line of reasoning. In this counterfactual, the Chololtecs are not presumed to attack first and kill Cortés so that events subsequent to this are entirely altered and demand logical predictions. Rather, I suggest that the cause be reinterpreted and the consequences be left intact so that only ex post facto reasoning is required. Such an exercise would go as follows.
Despite nearly unanimous Spanish support for the above account of the Cholollan massacre, it does not ring true. There probably was no Aztec army, as it is unlikely that Moteuczoma had tens of thousands of soldiers available to send to Cholollan during the agricultural season when most men were occupied. And even had the soldiers been available, only three days elapsed between Cortés's arrival in Cholollan and the alleged reports of the Aztec force -barely enough time for a message to be sent to Tenochtitlan, much less raise, arm, supply, and dispatch an army. Thus, an armed Aztec threat does not seem credible, although Cortés may well have accused Moteuczoma of sending one to keep him on the defensive. Moreover, the Spaniards also claim to have been alerted to Chololtec duplicity by the presence of barricades and stones piled atop houses in the city and concealed pits with sharpened stakes outside. The former is entirely expected since nearby Tlaxcallan was an enemy, and the latter is almost certainly a Spanish projection or misinterpretation. Concealed pits are relatively ineffectual against infantry but they were the standard European counter to cavalry. So while the mounted Spaniards might expect to encounter these, their use is otherwise unattested in the historical sources, and they would have been entirely alien to Mexico as there were no horses.
There was a massacre, as reported, but of the Chololtecs, and this was probably an intentional act by Cortés to destroy their leaders as a warning to other cities. There was no sound logistical reason to go to Cholollan but there was animus between that city and his new ally, Tlaxcallan, which threatened Cortés's plans. Cholollan straddled the main route between Tenochtitlan and Vera Cruz and would have threatened Cortés's rear once he continued his march. His need for reliable links to Vera Cruz had been brought home to him by the rapid exhaustion of his shot and gunpowder in the battle with Tlaxcallan. Vera Cruz was the only available source for these essential armaments and his access could not be impeded. Moreover, having established an alliance with Tlaxcallan, Cortés felt less constrained in his dealings with Aztec allies. 5 Thus, his decision to go to Cholollan can best be understood as political -to secure his rear and his lines of resupply, and chastise his friends' enemies.
The assessment above is not mere fantasy. Even though it varies from the widelyaccepted conventional account, it is what facts and perspectives that are not generally considered indicate happened. While we customarily think of counterfactuals as reconstructions that are contrary to the facts, this is a counterfactual that goes against the data as known, at least when one looks only at the Spanish accounts, and examines matters somewhat more broadly and emphasizes other facts to posit an alternative causal scenario. Thus I argue that the real counterfactual is the Spanish version, and that I am proposing a fuller, more accurate, account. Tetlock and Belkin (1996: 10-11) opened the theoretical door for this type of counterfactual with an example in which a different cause was posited to frame an alternative theory, that of dinosaur extinction as a result of comet impact rather than a climatic shift, a theory which has since been accepted by a majority of scholars in that field. Tetlock and Belkin used the illustration only as an example and did not extend the approach to human affairs, but the Cholollan example described here does. Moreover, this is a true case of minimal rewrite. Rather than altering both causes and consequences, this approach alters only the cause, leaving the event and its effects intact. That does not mean that I am arguing that changing a cause will necessarily retain the same effect. Rather, I argue that the actual cause of the Cholollan sequence of events was misunderstood and the presumed cause has been used to produce a specious causal link to the historically known effect. So this causal revisionism argues for a reinterpretation of the cause that led to the actual, experienced consequences and implicitly suggests that the cause as traditionally conceived is not and perhaps would not, in actuality, have produced the effect we accept as historically true.
Adopting this strategy and not changing the effect emphasizes ex post facto reasoning rather than predictive, and does not lead to an altered causal line that stretches indefinitely into the future as do conventional counterfactuals. Because this approach does not alter the effect, there is no further consequence for the historical record, though there will be for its interpretation. What is the practical difference between a counterfactual example and a theoretical explanation that emphasizes the wrong facts and ignores those that are crucial to another explanatory model? There is little, unless either case then tries to go beyond the known consequences to argue an alternative reality. So the true effect of a counterfactual lies in its deviation from established consequences and not from its alteration of the initial facts because that -the suggestion of an alternative causal mechanism -lies at the base of all revisionist history. This is precisely Cortés's account of the Cholollan massacre -he posited something that did not happen (a planned massacre of which Marina warned) and then fitted the subsequent events to that initial 'cause'. My reinterpretation does not disturb the consequences, but substitutes what I argue are the actual causal events for Cortés's counterfactual ones.
But is this an isolated case of an infrequently encountered phenomenon? I do not think so, though the exercise of critically examining such 'well established' causes may be. To illustrate the point, let me introduce another example from the same Conquest.
After Cortés had returned to the Valley of Mexico for the final assault on Tenochtitlan, he built 13 brigantines to control the lakes and then sent to Tlaxcallan for 20,000 more warriors, as well as requesting soldiers from his allies within the valley. The Tlaxcaltec forces were led to Tetzcoco, where the Spaniards were headquartered, by Xicotencatl the Younger and Chichimecateuctli. But that night, Xicotencatl left Tetzcocoin one version, because he was in love with a woman in Tlaxcallan -and Cortés sent a party of Spaniards in pursuit. He was seized and hanged for treason.
Xicotencatl was, in fact, killed, but the Spanish explanation is weak: other native leaders left combat, often with their troops, and no action was taken against them, nor was the majority of disaffected Spaniards harshly disciplined. The Spanish account satisfied legalistic reasoning by offering a cause that would justify the result, but an examination of the larger context suggests a political purpose behind Xicotencatl's execution and a different cause.
When Cortés first reached Tlaxcallan, power in the four confederated provinces rested with the rulers of only two, and of those two, Tizatlan's ruler, Xicotencatl the Elder, was dominant. When his son and heir apparent, Xicotencatl the Younger, led the Tlaxcaltec army against the Spaniards, the ruler of the competing province of Ocotelolco, Maxixcatl, took the counter position and supported the Spaniards. When the Spaniards were not quickly dispatched, support shifted to Maxixcatl's side and he assumed greater importance in the eventual coalition, while the fortunes of both Xicotencatls waned. The political balance in Tlaxcallan was upset and Maxixcatl's enhanced importance continued for the next year and a half, until he died of smallpox. Two other factors threatened Ocotelolco's prominence: Maxixcatl had sought to create a political link with the Spaniards by giving one of his daughters to Juan Velásquez, but both were killed during the flight from Tenochtitlan. And when Maxixcatl died, he was succeeded by a teenage son who was not yet a formidable ruler. As a consequence, the fortunes of Xicotencatl the Younger now significantly improved, while those of Cortés worsened. Had this happened earlier, it might have proven fatal to the entire Spanish enterprise. But now, a few months into 1521, even though Xicotencatl's domestic position was strengthened, events external to Tlaxcallan made his support, though still important, less essential -notably Cortés's alliance with the Acolhua in the Valley of Mexico. Nevertheless, Xicotencatl the Younger was now a greater threat to Cortés, and his death was likely a calculated Spanish effort to eliminate a hostile ruler in Tlaxcallan and to improve the positions of those with closer ties to Cortés. 6 But did the events as I have briefly reconstructed them actually happen, and what does either of these interpretations do that the conventional ones do not? In short, how does this approach differ from good, solid historical analysis as currently practiced? Historical analysis that is canonically historical, that is, history that relies on the demonstrable linkage of events from those events themselves cannot effectively adopt this approach. The standard interpretations are suggested by the data and are often the perspectives offered by the historical participants themselves, so if an alternative understanding is to be adopted, one must go beyond the traditional historical frame, for alternative answers will generally not be found within it. What guides such reanalysis is social context, a wider cluster of events and conditions than those historically or intuitively evident from the immediate documentary record. Yet merely broadening the scope of inquiry guarantees no answer: the potentially infinite social field must be delimited to what is actually causal. And this is where social science research in the form of comparative analysis merges most naturally and gracefully with the historical enterprise.
Comparative analysis generates probablistic relationships between phenomena where the causal links are not always intuitively obvious and cannot necessarily be deduced from individual historical cases. Yet comparative analysis can convincingly show their causal role. And since a major goal of historical reconstruction is to seek the unknown cause of a known effect, comparative analysis broadens the alternative possibilities beyond that (or those) proposed in the historical accounts, which may be self-serving and are certainly constrained by the notions of cause of their time. This approach, then, is not strictly counterfactual -it is, after all, the often-unrecorded cause that is contested, not the recorded effect. Rather, it is documentarily counterfactual, which is a niche opened by social science approaches in history.
It is difficult to say conclusively whether my reinterpretations of events are demonstrably, factually, superior. But they do bring together more data, give a role for Indian interests and perspectives that is largely absent from the Spanish version, and offer political reasons for actions that are otherwise difficult to understand. Moreover, while they do not alter the subsequent historical consequences, they do put them in an entirely different light. They suggest that events arose from causes other than those conventionally offered and, if such is the case, they provide alternative perspectives on the Conquest in which Indian factionalism and infighting play a dominant role and therefore cast the Conquest in an entirely different light, even with largely the same data, and open up a new avenue of research, while the traditional views have long since become entrenched, uncritically accepted, and have lost their intellectual challenge.

VI. CONCLUSION
In sum, counterfactual analysis is not merely entertaining fiction. Rather, by forcing the analyst conceptually to alter the key events, it offers a sound approach for testing and proofing causal arguments that rest upon the inherently limited historical data typical of non-Western and contact situations. Even though the assessment cannot cover all contingencies because it is limited to what can be logically foreseen, conceptually altering the pivotal causes in one's argument and then reassessing the sequence of events also provides a ready means of correcting and strengthening the analysis. But more than merely testing historical arguments by changing causes that are presumed to alter effects which then produce different consequences, the counterfactual approach provides an effective means of challenging conventional interpretations, often long accepted. And it is especially useful in ethnohistory, where the historical record tends to be thin. Effects are typically recorded, but only rarely are their causes. Thus, the counterfactual approach is useful in establishing what is, in fact rather than in convention, a pivotal event, and once established, the stage is set for causal revisionism. The accepted cause can then be critically examined and, where pertinent, challenged, as these causes are far from certain, resting as they typically do on far shakier documentation than the effects, and the ascribed causes are often the assumptions of genuinely perplexed or self-serving outsiders to the culture being chronicled. By examining these causes and comparing them to similar situations elsewhere, and by assessing them in relation to an anthropological understanding of the indigenous culture, the actions of all parties are placed in a broader social context, and those actions can be reassessed to yield a more satisfactory interpretation of the known consequences.
1 It should be noted that, in this regard, propaganda is an effort to change the perceived circumstance environment of the enemy through deception and to do so in a way that causes him/her to act disadvantageously. 2 But how does the plethora of alternative choices in life square with the theoretical use of counterfactuals? Niall Ferguson (1997: 87) claims that the only acceptable counterfactual is not that which is merely premised on plausible alternative courses of action, but on courses of action that were actually considered and are documented in the historical record. His is not, of course, the only perspective, and Herrmann and Fischerkeller (1996: 163) argue for the use of general motivations, such as a theory of power-driven states, in framing counterfactuals. But, while Ferguson's standard yields the strongest counterfactual analysis, at least on an evidentiary basis, this claim, as stated, implies that the validity of a counterfactual as a theoretical construct depends on the data! However, the validity of a theory rests on its logic, not on its evidence, which is recognizable only in relation to a theory and goes only to confirmation. So while Ferguson's approach yields more adequate evidence of a counterfactual's plausibility, it is not logically superior to cases of more speculative counterfactuals. 3 If our notions of historical causality are culturally determined, how does counterintuitive causation fit in? In part, counterintuitive causality is academic, as opposed to common sense, or popular, causation. But it can shape popular views of causality in history, and a well-established counterintuitive cause can force the type of perceptual shift that I argue for, and it then becomes the popular perception. 4 Because we look at effects and try to link them to prior causes, any arbitrary limit on historical inquiry, whether by pre-/post-Conquest divisions, centuries, or anything else not delimited by a problem focus, is likely to separate the cause and the effect and foster a misinterpretation of what we are seeing, e.g. a cause without its decapitated effect may be taken as the effect. 5 For a fuller discussion of the Cholollan massacre and Marina's role in it, see Hassig, 1998. 6 For a fuller discussion of the death of Xicotencatl, see Hassig, forthcoming.