THE ROLE OF TRAINING IN RECREATIONAL AVALANCHE ACCIDENTS IN THE UNITED STATES

Ian McCammon*

National Outdoor Leadership School, Lander, Wyoming

ABSTRACT: Avalanche educati?n has beco~e widely availab~e ~n the United States, and yet. trainedrecreationists continue to comprise over a third of avalanche victims. Does avalanche education reallymake a difference? This study investigated the relationship between avalanche education and victimbehavior in 344 recreational U.S. accidents, and found that victims with more avalanche training did infact take fewer overall risks. However, all of the risk reduction in trained recreationists can be attributed tobetter mitigation measures taken by these victims. None of the risk reduction appeared to be the result oftrained groups exposing themselves to less hazard. In fact, victims with basic formal training exposedthemselves to more hazard than any other group, including those with no awareness of avalanches. Inlight of recent findings in ~eci~i~n scie~ce, these results sug.gest that behaviorist and naturalistic teachingstrategies would be effective In Improving avalanche education.

KEYWORDS: Avalanche accidents, avalanche education, human factors, decision making, heuristics, riskhomeostasis, risk reduction, behavioral education, naturalistic decision making.

1. INTRODUCTION

On January 12, 1993, three skiers left the well­marked boundary of Vail Ski Area headed for thebackcountry. The group had been warned of thedangerous avalanche conditions by the Vail SkiPatrol, but these skiers had just completed a two­day avalanche course and were confident thatthey could find safe skiing. Fresh slides werevisible in the area, and a follow-up investigationindicated that the skiers probably experiencedcollapsing of the snowpack as they hiked. Despiteobvious indications of dangerously unstable snow,the group chose to ski a steep, wind-loaded gully.The avalanche they triggered caught two of theskiers, burying and killing one of them.

Accidents like this one raise uneasy questionsabout the influence of avalanche education amongrecreationists. Does it really make them safer, ordoes it create overconfidence that lures them intomore dangerous terrain?

This stUdy investigated these questions in twoways: (1) by examining the relationships betweenavalanche victims' level of training and theirbeh~~ior, and (2) by reviewing current research indec~slon science that applies to avalancheaCCident prevention among recreationists.

2. DESCRIPTION OF THE STUDY

To assess the effects of avalanche educationon recreationists' behavior I reviewed 546 ava-lanch '. ..'

e inCidents involVing 1,050 recreationists. I

~~ut~oraddress: 3363 S. Plaza Way, Salt Lake City,e '. ~A 84109; tel: 801-483-2310;

mall: Imccammon@aol.com.

37

looked at the hazards present at the time of theaccident, the mitigation measures taken by theaccident party, and the highest level of avalanchetraining present in each party.

To quantify the risks that led to each accident, Iused a definition of risk from natural hazardanalysis (Tobin and Montz, 1997):

risk = ( probability ) x (vulnerability). (1)ofoccurrence

For each event, I assumed that the probabilityof occurrence was related to the number of indica­tors that a hazard existed at the time, expressedas a simple "hazard score:' For example, anaccident that occurred on an obvious avalanchepath (a hazard indicator) that had been recentlywind-loaded (a second indicator) during a time ofhigh forecasted hazard (a third indicator) had ahazard score of 3. The higher the hazard score,the higher the probability of an accident occurring.Definitions of hazard parameters appear in tableA1 in the appendix.

Because I computed hazard scores from writtenaccident accounts, the hazard score for any givenincident may have been subject to various report­ing biases. Such biases would arise fromvariations in rescuers' or victims' assessments andobservations of the accident site. However, sincesix of the seven hazard indicators were verifiableby third-party accident reports and weather oravalanche forecasts, any reporting biases shouldbe approximately uniform over all categories ofvictim training. The most potentially bias-pronehazard parameter, the presence of collapsing,

(3)

(2)

40%

20%

45%

As with any analysis based on accident data, itis important to recognize that the results of thisstudy apply only to a self-selected sample: thoseinvolved in avalanche accidents. Extending theseresults to other populations, such as all winterrecreationists, may not be entirely valid.

3. RESULTS

Of the 546 recreational avalanche incidents thatI reviewed, 202 involved victims with unknowntraining. In the remaining 344 cases, 30% of thegroups had no training or awareness, 24% of thegroups had a least one person with an awarenessof the hazard, 31% had at least one person withbasic formal training, and 14% had at least oneperson with advanced formal training. 90% of theavalanches were triggered by the victim or the Vic­tim's party, 6% were natural, and 4% had un­known triggers.

The frequencies of the seven hazard parame­ters for recreationists with no training appear infigure 1. Because these groups lacked evenrUdimentary hazard recognition skills, it is nosurprise that most of their mistakes were made onhigh hazard days in obvious, recently wind loadedavalanche paths. A significant portion (29%) wereburied or killed in terrain traps, and one in fivegroups had noticed recent avalanches, but wereprobably unable to recognize their meaning.

Figure 1. Reported frequency of hazard indicatorsfor accidents involving recreationists withno avalanche training.

The frequencies of the seven hazard parame­ters for groups with avalanche awareness andformal training are shown in figure 2. Thesevictims appeared more likely to heed forecastedconditions but surprisingly, they were not any lesslikely to avoid wind loaded avalanche paths. Thismay have been due to their higher level of skill attheir sport and their tendency to seek out steeperand more hazardous slopes.

Hazard scores for each category of trainingappear in figure 3. Because each distribution ofscores deviated significantly from normality (as

60% 56%

(I) 50%C

":2 40%

~'030%

"0>J!l20%

8q; 10%c.

38

_ M(Hs)c

M(RQ)c - 1 + M(MS)c

Note that RQ and M{RQ) are relative descriptivequantities used to compare the behavior of ava­lanche victims based on their training. They arenot actual probabilities so they are not an absolutemeasure of risk. Definitions of the trainingparameters appear in table A3 in the appendix.

The hazard and mitigation scores in this studycame from avalanche accidents where the level oftraining of the victim(s) was known or could bereasonably inferred. Also, I considered onlyrecreational accidents to minimize biases intro­duced by employment settings, highway incidents,in-area accidents or guided outings. Accidentaccounts came from the Westwide AvalancheNetwork, the Cyberspace Snow and AvalancheCenter, records of the Colorado AvalancheInformation Center, The Snowy Torrents (Loganand Atkins, 1996; Williams and Armstrong, 1984),and archived articles from various newspapersand journals. Results are from the winters 1972-73to 1999-2000.

where hi,k are the hazard parameters, HSk is thehazard score, mi,k are the mitigation parameters,and MSk is the mitigation score, adjusted to avoidundefined RQ values for groups who took no miti­gation measures.

To find the relative risk for each category oftraining (c), I combined the median hazard scoresM{HS) and median mitigation scores M(MS) of allqualifying accidents involving victims with thatlevel of training:

does show an increased incidence with training(suggesting a possible reporting bias), but its over­all effect on the relative risk values was minor.

For each event, I assumed that the vulnerabilityof the group was inversely related to the numberof mitigation measures that the group took prior tothe accident, expressed as a "mitigation score:'For example, a group travelling with two or morepeople (one mitigation measure) wearing beacons(another measure) and exposing only one personat time (a third measure) had a mitigation score of3. Definitions of mitigation parameters appear intable A2 in the appendix.

Combining hazard and mitigation parameterswith equation (1) gives a relative measure (or riskquotient RQ) of the risk taken by each group at thetime of the (k th

) accident:

RQk= L hi,k = HSk

1 + L mi,k 1 + MSk '

0.9580.0090.958

PM-W(O)

2.12.02.62.1

median

minimized contact none orexposure unknown

1.01.00.91.0

Odev

0.3590.2690.1250.542

beacons shovels notalone

0%'--"-----"----20/, 3%

Training

training, who exposed their group to more hazardsthan any other training category.

The frequencies of the six mitigation parame­ters for victims with no training appear in figure 4.Understandably, these recreationists failed to takeany significant precautions other than not travel­ling alone, since they probably did not recognizethe hazard. Other training categories (figure 5) .show a uniform increase in almost all of themitigation measures. Note that the improvement isnot simply due to carrying more rescue gear;victims with more training were actually engagedin a higher incidence of behavioral mitigation(having a plan, minimizing exposure, maintainingcontact) than victims with less training.Remarkably, the tendency to expose more thanone person at time to the hazard remainedsignificant at all levels, as has been noted inprevious studies (Smutek, 1980).

l00%,---------------n-=-:'-:-04.,-,65%

Table 1. Hazard score distributions in figure 3. Pn isthe probability that the distribution isnormal, Odev is the quartile deviation andPM-W(O) is the Mann-Whitney probabilitythat the difference relative to the "notraining" median is due to chance. Samplesizes are the same as in figures 1 and 2.

Figure 4. Reported frequency of mitigation mea­sures for accidents involving recreation­ists with no avalanche training.

The correlation of improved mitigationmeasures with avalanche training appears moreclearly in figure 6. Again, due to non-normality ofthe mitigation score distributions, I used a non­parametric method to asses differences betweenmedian scores. Table 2 shows a very strongcorrelation between avalanche training andincreased mitigation, particularly amongrecreationists with formal training.

noneawarenessbasicadvanced

~80%.."0

.§ 60% .

'0CD

~40%

LO%a.

0% 0%

2.6

2.0

Awareness

Hazard sC?r~s for avalanche victims bylevel ?f training. Boxes indicate the inter­quartile range and whiskers indicate maxi­mum and minimum values.

Figure 2.

Figure 3.

+8%

2O'\-------------+-12-%-~

..CI>"D

1i..'5~ 0% t-,;,;::';'--,i;r-­C

!la-Ii(b) basic training, n = 106

.20%L------------'-'------"'----'

20%~------------------,

·20%~~======::::============~

Reported frequency of hazard indicatorsfor accidents involving recreationists with(a) awareness, (b) basic training, and (c)advanced training. All values are relativeto those in figure 1.

indicated by the O'Agostino-Pearson test), I chosea non-parametric test (the Mann-Whitney tiedrank, normal approximation) to assess differencesin the distributions (Zarr, 1999). For each hazardscore median, I computed the probability that itsvariation from the median score of the untrainedgroup was due to chance (table 1). Surprisingly,hazard scores show no significant reduction withincreased training, and for victims with basicformal training, hazard scores actually increased.Apparently, avalanche training had little influenceon where these people chose to ski, snowmobile,etc., except in the case of those with basic formal

.20% L- ---'(.:..:.c)_'_adv'-'-'-anced~_'_lra;;,;;in_'_ing~,_'_n_=_:.46~

0%

6.5

5.5

e4.5

~3.5l!•=2.5.c:: 2.'

'.5

0.5

None

39

0.5

Basic

0.8

0.7

Awareness

ideal homeostasis

\\

\\

\\

\\

\\_---.--~-.---.deal mitigation

1.0

None

1.2 y-------------------,

Table 2. Mitigation score distributions for figure 6.Variables and sample sizes are the sameas in table 1.

sole cause of the decrease. Factors not examinedby this study include: (1) the relative risk attitudesbetween recreationists who seek out differentlevels of training, and (2) field experience amongthe different groups. Clearly, further study isneeded in this area.

0.0 +------------------

4. DECISION MAKING AND AVALANCHEEDUCATION

Because the majority of avalanche accidentsare caused by the victims (90% in this study),avalanche educators have long recognized theroles that education and decision making play inpreventing accidents. This section examinesrecent findings in decision research and theirimplications for avalanche education. To teachdecision making effectively, it is helpful to knowhow decisions (good and bad) are made in thecomplex and uncertain environment that the ree­reationist encounters in the winter backcountry.

4.1. Victims with Avalanche Awareness

By definition, these individuals could probablyrecognize most avalanche paths and obvioussigns of instability, but they had little experiencemaking decisions in avalanche terrain.

Figure 7. Risk quotients for avalanche victims bylevel of training. If education had no effect.the data would follow an ideal homeostasismodel. If education was maximally effectiveat teaching precautions. the data would fol­Iowan ideal mitigation model.

Training P n Qdev median PM-W(O)

none 0.001 0.3 1.1awareness 0.083 1.3 1.7basic 0.038 1.3 2.2advanced 0.219 1.9 3.4

0.8

1.0

0.2

C.!!!g 0.6<T.><

.~ 0.4

3.4

AdvancedBasic

24%

Awareness

23%

None

1.1

0%+-""""'--

0%

5.5

0.5

Q) 4.5

8~ 3.5o~2 2.5

E1.5

40

60%~===============~

Figure 5. Reported frequency of mitigation mea­sures for accidents involving recreation­ists with (a) awareness, (b) basic training,and (c) advanced training. All values arerelative to those in figure 4.

0%

Figure 6. Mitigation scores for avalanche victims bylevel of training. Box-whisker parametersare the same as in figure 3.

By combining the median hazard scores andthe median mitigation scores for each group inequation 3, we can calculate the median relativerisk for each category of victim training. As shownin Figure 7, the overall risks taken by recreationalvictims does in fact decrease with training,suggesting that avalanche education correlateswith a decrease in the accident rate amongrecreationists. It is interesting to note that thedecrease is not linear; victims with basic trainingappear to have taken more risks than all othergroups with training or awareness, but they stilltook fewer risks than victims with no training.

It's important to note that the decrease in rela­tive risk among trained recreationists does notnecessarily mean that avalanche education is the

6.5 y-------------------,

not minimized none orbeacons shovels alone plan exposlXe contact unknoWn

6O%~=e:::..-=:";';":"'----='------'------:=:':""----,

(c) advanced training, n =48

Tversky and Kahneman (1974) have demon­strated that people in difficult and unfamiliar situa­tions base their responses on simple rules, or"heuristics:' In certain well-defined circumstances(such as estimating probabilities or drawing infer­ences from hypothetical data), heuristics can leadto systematic biases (Plou~, 1993; Siovic,. F.ischoffand Lichtenstein, 1982). Since most heuristicsresearch has focused on understanding thesebiases (Cohen, 1993), it has encouraged a view ofhuman beings as marginally effective decisionmakers (Lopes, 1991; Kleinmuntz, 1985). But in iII­defined, real-world situations, he_uristic decisionstrategies generally perform very well. For therecreationist, a heuristic might as simple as: "avoidslopes over 30° on high hazard days:'

Where do heuristics come from? In unfamiliarsituations, people readily adopt simple guidelinesand recipes for action, and will typically notabandon them until they clearly fail. In theabsence of clear rules, people are adept atsearching a situation and their experience forpatterns that suggest approximate rules, whichthey modify as required by the circumstances(Baron, 1988).

Given people's preference for heuristic reason­ing in unfamiliar situations, it's no surprise thatsuccessful efforts aimed at increasing avalancheawareness favor simple messages (see, forexample, Fredston, Fesler and Tremper, 1994; orTremper, 1990). Munter (1997) has even pro­posed a numerical heuristic set for decisionmaking in avalanche terrain.

If inexperienced recreationists prefer to useheuristics in avalanche terrain, are there ways toteach heuristics more effectively? Studies of howpeople learn show that effective instruction ofmotor and cognitive skills tends to follow a behav­ioral model (Davis and Davis, 1998). For theavalanche educator, this means:

• clearly communicating the specific skills andexpectations of the course (e.g. recognize andavoid avalanche paths),

• subdividing skills and expectations intomana~eable tasks (e.g. measuring slopeangle IS one sub task in recognizing ava­lanche paths),

• role modeling the skill competently andconsistently, and

• ~roviding lots of opportunities for practice, withtimely and effective feedback.

tanBehavioral ~ppr.oaches underscore the impor­educe ~f field time .In introductory avalanche

cation. According to the behavioral model,

41

theoretical instruction beyond basic concepts willhave little impact on the beginner's ability toexecute heuristic-based skills (such as recognizingavalanche slopes). Perhaps most valuable ispractice involving real examples of the problem,preferably in the environment where the skills willbe applied. Current guidelines for avalancheeducation emphasize the importance of a fieldcomponent (AAAP, 1999). Further arguments foran emphasis on field-based activities can be foundin brain-based educational theory, which maintainsthat, under stress, people do what they havephysically practiced rather than what they've beentold (Jensen, 1998).

Avalanche educators can significantly reducerisky behavior among recreationists by simplybuilding better mitigation skills among theirstudents. The lower curve of figure 7 indicateshow the accident data would appear if victims hadtaken all six mitigation measures (MS=6) whilekeeping their hazard exposure the same. Clearly,a small improvement in mitigation habits yields alarge gain in overall risk reduction.

4.2. Victims with Advanced Avalanche Training

At the other end of the training spectrum arerecreationists with extensive training and fieldexperience. Studies of experts in complex real­world situations suggest that these people do not,as a whole, use heuristic strategies (Dreyfus andDreyfus, 1986). Instead, experts seem to recog­nize a situation as typical of a class of situations,mentally test a response, then act (Klein, 1998).The process of recognizing key features of asituation and recalling the appropriate responsehappens qUickly and unconsciously, commonlybeing experienced by the expert as "intuition:' Thisrecognition-primed decision (RPD) model impliestwo important messages for non-experts: (1) itsaccuracy depends on the size of the experiencebase, and (2) the skill to recognize a situation astypical cannot be taught; it can only be learned.

For avalanche educators wanting to buildexpertise in their students, at least three teachingmethods will be effective (Means et aI., 1998):

• focused field exercises and well-designedscenarios covering a wide variety of situations(with quality feedback),

• diligent documentation by the students of theirobservations and decisions, and

• applying new theoretical concepts to showdifferent ways to recognize familiar patterns.

A useful tool for providing feedback toadvanced students is the pre-mortem exercise(Klein, 1998). Once a student has outlined the

specifics of a plan (route, rescue, or other result ofa decision), ask them to imagine their plan beingexecuted perfectly, but failing. Having themexamine possible sources of failure in a futurecontext breaks their attachment to the plan's suc­cess in the present, allowing them to creativelyexplore new ways of perceiving situations theythought were familiar.

4.3. Victims with Basic Avalanche Training

In the middle of the training spectrum arerecreationists who have taken one or two formalclasses but who have limited experience inapplying their avalanche knowledge. Theserecreationists are at something of a decisionmaking crossroads: they may feel that heuristicsare too restrictive but they lack the experience toemploy expert decision making strategies. If theyattempt to employ one anyway, their experiencebase may contain little more than "I high markedthis slope last week-end and nothing happened:'Such statements of "rationalized expedience" arecommon in avalanche accident accounts, evenamong trained victims (Fesler, 1980).

A perceptive instructor can mitigate thenegative effects of an inappropriate expertisestrategy by being alert to its use. Simply asking astudent "What experience did you base thatdecision on?" can be an effective way of empha­sizing the importance of haVing a broad experi­ence base for critical decisions.

Inappropriate use of the RPD strategy is not theonly obstacle faced by this class of recreationists.Recent developments in decision science suggestat least four others.

failure of stage models

Stage models lead a decision maker throughlogical steps to arrive at the best course of action.A simple example is: (1) define objectives, (2)collect relevant data, (3) evaluate alternatives, and(4) pick the best alternative. Stage models areattractive because they appear systematic andportable, and have proven themselves to be verypowerful tools for solving problems when objec­tives are known (Lewis, 1997).

Unfortunately, experiments with people facingill-defined problems (such as those found inavalanche terrain) suggest that stage models canbe ineffective and even misleading. Klein (1998)and Beach and Lipschitz (1993) and others havefound that in ill-defined problems, subjects willavoid using stage models even when they havehad extensive training in stage-based decisionmethods. Furthermore, the heuristic or intuitive

42

strategies they end up using often yield betterresults (Means at aI., 1993).

A qualified exception to the ineffectiveness ofstage models occurs in occupational situations oron guided trips where objectives are simple andclear within the group, or in cases where judge­ments must be justified to others. In thesesituations, stage-based decisions can be usefUl,but they are time-consuming and remain highlyvulnerable to biases introduced by unstatedpersonal objectives (Simon, 1990).

The serious limitations of stage models suggthat they be used sparingly, if at all, in mostavalanche education aimed at recreationists.While stage models are temptingly easy to teachand highly appropriate for well-defined problems(Nickerson, 1994), there is little evidence torecommend them for use by recreationists inavalanche terrain.

recalibration

Recalibration occurs when an individual seekout experiential feedback to re-adjust their expetations (Pious, 1993). Recreationists who havebeen conditioned by avalanche classes or themedia to see avalanches as the "white death" tsweeps away the ignorant and imprudent arenaturally drawn to recalibration activities whenthey see their friends take chances on dangerouslopes and nothing happens. By taking risks inavalanche terrain, these people are simplyattempting to recalibrate their estimate of theavalanche risk to a more realistic standard.Accidents are a natural consequence of thisstrategy. The responsibility of avalanche educa­tors here is clear: avoid "scare tactics" and prerealistic estimates of accident probabilities.

ballistic reasoning

Dorner (1996) has demonstrated that peopletend to protect their perception of their owncompetence, and will actively avoid evidence tothe contrary, partiCUlarly in complex situations.This results in "ballistic behavior" where peopleappear to ignore obvious clues that they aremaking a mistake. In the accident described inIntroduction, the victims were warned of thehazard, they saw recent avalanches and expeenced collapsing, and yet they chose to ski a w·loaded avalanche path ending in a terrain trap.Although it is tempting to view this behavior as"irrational;' ballistic reasoning has an importantfunction within the individual: it reduces confus'and bUilds confidence, allOWing the person tomove on to more challenging problems. Mostpeople reserve ballistic behavior for non-critiC8

situations where the benefits are great and therisks minimal. But when a situation is incorrectlyperceived as low-risk, ballistic behavior is clearly

self-destructive.One solution to ballistic behavior is "external

, attribution;" basically, examining how circum­stances or previous events lead to understandableerrors. When students understand how they madetheir errors, they are less likely to make the samemistake again. An obvious message for theavalanche educator is to stress the limitations ofheuristic reasoning at the outset, and be compas­sionate yet realistic about student mistakes andtheir consequences.

risk homeostasis

This theory maintains that education aimed atreducing accidents will be ineffective becauseindividuals maintain an approximately continuouslevel of risk (Wilde, 1994). As people learn how tomitigate a hazard, they compensate by takingmore chances while keeping their overall level ofrisk (their "target risk") the same. Research resultssupporting this theory can be found in driver safetytraining, drug education, AIDS awareness, andnatural hazards education. In the risk homeostasismodel, recreationists who have completed anintroductory avalanche course may perceive theirnew knowledge as inherently decreasing theirchances of being involved in an avalanche, andthus choose riskier slopes in an effort to maintaintheir target level of risk.

As shown in figure 7, he overall influence ofeducation on relative risk among avalanchevictims does not follow a purely homeostaticmodel. However, risk homeostasis probably playssome role in hazard exposure, particularly amongrecreationists with basic avalanche training.Methods for overcoming the effects of riskhomeostasis are not clear; some educators havesuggested that simply pointing out how people'starget level of risk is set by social circumstances oradvertising will be sufficient to reduce their risklevel.

4.4. The limits of education

~an quality avalanche education, aimed at amotivated aUdience, completely eliminateavalanche accidents? Perrow (1984) hassuggested that in highly complex systems smallevents can combine in unforeseeable ways tocreate a baseline accident rate beyond which we~nnot reduce our risk and still extract benefits~th~ experience. In this study about 4% of theH Ifaccidents had a known haza;d score of zero.

a of these resulted in fatalities. At the current

43

fatality rate a?1ong recreationists, this correspondsto about 0.5 lives per year lost in the United Statesas an irreducible risk of recreation in avalancheterrain.

Because winter recreationists will always seekout steep and dangerous slopes, it's unlikely thatfatality rates will ever approach the irredicible limitregardless of improvements in avalanche educa- 'tion. But in 98% of fatalities, education has thepotential to make a significant difference.

5. SUMMARY

In the 344 recreational avalanche accidentsreviewed in this study, avalanche trainingcorrelated with:

• an overall decrease in the relative risk takenby victims at the time of the accident, and

• an increase in mitigation measures amongvictims.

Avalanche training did not appear to decreasethe hazards that groups exposed themselves to,and in the case of victims with basic training,hazard exposure actually increased.

Recent findings in decision science suggest thatvictims use two strategies for decision making inavalanche terrain: heuristic (rule-based) andexpertise. Heuristic skills can be developed byclassical behavioral education methods and astrong emphasis on practical exercises. Expertisecan be developed by demonstrating conceptualrelationships with detailed scenarios and exercisescombined with various feedback methods.

Ultimately, the real measure of avalancheeducation is the reduction of the accident rate. Bycarefully bUilding on decision skills that studentsalready have, educators can help recreationistsreduce their risks without limiting their experienceof the winter backcountry.

6. ACKNOWLEDGEMENTS

I'd like to thank the students and staff of theNational Outdoor Leadership School, who haveproVided valuable insights into decision makingand education. In particular, I would like to thankDon Sharaf, Allen O'Bannon, Lynne Wolfe andJohn Gookin for discussions and reviews.

Thanks also to Dale Atkins and Knox Williamsof the Colorado Avalanche Information Center inBoulder for allowing me access to avalancheaccident records, and for valuable discussions.

7. DISCLAIMER

The views expressed are those of the authorand do not necessarily reflect official positions ofthe National Outdoor Leadership School.

8. REFERENCES

MAP. 1999. Recommendations and philosophy of theMAP avalanche course guidelines. AvalancheReview, Vol. 18 no. 2, 4-5.

Bacharach, M. and Hurley, S. 1991. Foundations ofDecision Theory: Issues and Advances. BasilBlackwell, Cambridge, MA, 1-38.

Baron, J. 1988. Thinking and Deciding. CambridgeUniversity Press, Cambridge, UK, 16-28.

Beach, L. and Lipshitz, R. 1993. Why classical decisiontheory is an inappropriate standard for evaluatingand aiding most human decision making, in DecisionMaking in Action: Models and Methods, G. Klein etal. eds. Ablex Publishing, Norwood, NJ, 21-35.

Cohen, M. 1993. Three paradigms for viewing decisionbiases, in Decision Making in Action: Models andMethods, G. Klein et al eds., Ablex Publishing,Norwood, NJ., 36-50.

Davis, J. and Davis, A. 1998. Effective TrainingStrategies. Berrett-Koehler Publishers, SanFrancisco, CA, 103-135.

Dorner, D. 1996. The Logic of Failure: RecogniZing andAvoiding Error in Complex Situations. Perseus,Cambridge, MA.

Dreyfus, H. and Dreyfus, S. 1986. Mind Over Machine:The Power of Human Intuitive Expertise in the Era ofthe Computer. Free Press, New York.

Fesler, D. 1980. Decision-making as a function ofavalanche accident prevention, Assoc. Committeeon Geotechnical Research, National ResearchCouncil, Canada, Technical Memorandum 133,Ottawa, Canada, 128-136.

Fredston, J. and Fesler, D. 1994. Snow Sense, A Guideto Evaluating Snow Avalanche Hazard, AlaskaMountain Safety Center, Anchorage, AK.

Fredston, J., Fesler, D. and Tremper, B. 1994. Thehuman factor-lessons for avalanche education,Proc. 19941ntemational Snow Science Workshop,Snowbird, UT, 473-487.

Jensen, E. 1998. Teaching with the Brain in Mind.Assoc. for Supervision and Curriculum Development.Alexandria, VA.

Klein, G. 1998. Sources of Power: How People MakeDecisions. MIT Press, Cambridge, MA.

Kleinmuntz, D. 1985. Cognitive heuristics and feedbackin a dynamic decision environment, ManagementScience, 32(6), 680-702.

Lewis, H. 1997. Why Flip a Coin?: The Art and Scienceof Good Decisions. John Wiley and Sons, New York

Logan, N. and Atkins, D. 1996. The Snowy Torrents:Avalanche Accidents in the United States, 1980-86.Colorado Geological Survey, Special Publication 39,Denver, CO.

Lopes, L. 1991. The rhetoric of irrationality. Theory &Psychology, 1, 65-82.

44

Means, B. at al. 1993. Training decision makers forreal world, in Decision Making in Action: MOdelsMethods, G. Klein et al. eds., Ablex Publishing,Norwood, NJ., 307-326.

Munter, W. 1997. 3 x 3 Lawinen: entscheiden inkritischen Situationen, Agentur Pohl undSchellhammer, Garmisch-Partenkirchen, Germ

Nickerson, R. 1994. The teaching of thinking andproblem solving, in Thinking and Problem SolviSternberg ed., Academic Press, San Diego, CA.409-449.

Perrow, C. 1984. Normal Accidents: Living with High.Risk Technologies. Basic Books, New York.

Pious, S. 1993. The Psychology of Judgement andDecision Making. McGraw-Hili, New York, 222-

Simon, H. Alternative visions of rationality, in Rationin Action, P. Moser, ed., Cambridge UniversityPress, Cambridge, UK, 189-204.

Siovic, P., Fischoff, B. and Lichtenstein, S. 1982. Faversus fears: understanding perceived risk, inJudgement Under Uncertainty, Heuristics andBiases, D. Kahnenman, P. Siovic and A. Tverskyeds. Cambridge University Press, Cambridge. U463-489.

Smutek, R. 1980. Experience and perception ofavalanche hazard, Assoc. Committee onGeotechnical Research, National Research CouCanada, Technical Memorandum 133, Ottawa,Canada, 145-50.

Tobin, G. and Montz, B. Natural Hazards: Explanat'and Integration. Guilford Press, NY., 282.

Tremper, B. 1990. Alternatives to boredom andbewilderment-report from the avalanche class~

Avalanche Review 8(4),3-6.

Tversky, A. and Kahneman, D. 1974. Judgement ununcertainty: heuristics and biases, Science, 185.1124-1131.

Wilde, G. 1994. Target Risk. PDE Publications, ToOntario, 83-107.

Williams, K. and Armstrong, B. 1984. The SnowyTorrents: Avalanche Accidents in the United Stat1972-79. Teton Bookshop Publishing, Jackson.

Zarr, J. 1999. Biostatistical Analysis. Prentice-Hall.Upper Saddle River, NJ. pp. 87-89,146-153.

9. APPENDIX

Table A1. Hazard parameters

high forecast

terrain trap

obvious path

recentavalanches

collapsing

obvious windloading

thaw instability

high or extreme forecast posted forthe region

terrain feature that increasedseverity of the slide's effects

distinct start zone, track or runout,or known path

within last 48 hrs and seen byvictim(s)

cracking, or hollow sounds

obvious wind pillow or fresh cornice

above-freezing air temperaturesor rain

Table A2. Mitigation parameters

beacons

shovels

not alone

plan

worn by party

and probes carried by party

group size > 1

group communication regarding routeand use of islands of safety

minimized minimum number of people exposedexposure

contact visual or verbal contact with the personbeing exposed'

Table A3. Education parameters

none

aware

basic

advanced

no training or awareness

rUdimentary awareness of hazard

1-2 day avalanche course minimum

mUltiple trainings over several years, plusseveral years or more of backcountryexperience.

45