sie 99
Self-Explonations:
How Students Study and Use Examples
in Learning to Solve Problems
MICHELENE TH. Ci
MIRIAM BASSOK
Learn Reser and Deslopment Comer
‘hsburgh PA
MATTHEW W. Lewis
The Rand Corporation
[PETER REIMANN
Unie of Freers
RoBeRT GLASER
Larnng Reser and Developmen Corer
sdemeyocre Amaclansucaantvdinemagtasee se
topes ten
{earnings constructive process in which a student convers words and ex
amples generated by a teacher or presented na et, towable sil, sch2 solving problems. This proces of conversion exsentily form of son
Struine se nstuction Simon, 179). Although the esearch on te ay
of god teching (suchas that which attempts fo deni the characeatit
fof good Socratic lar, Collins & Stevens, 1982}, ewes esearch on the
‘uty of» good text eich a that which manialatesthe quai of labore
tins, Reder& Anderson, 180) ray be informative imately, arin ress
fmt learning kl thai the students therelves bing to beara they ean
“The goal of thisresearch isto understand thestodent' contibaton
ing. partulr, we examine how students ian via et explanations.
“Some ofthe best problem-solving research has concentrated on the con-
version of already encoded knowledge into smooth, fas, skilful problem
solving Thisconvesion process dominate, or example, AndESon's theory
fof shl acqiston (Anderson, 187). thal theory, the procetsf conver
Som is achitved by using general weak methods whieh an conver delraive
Inowledge ito domsinsestc procedore vn the mechanism of co
‘Gon Th, i Anderson's theory, raesomed thatthe effort ul rose ies
In the conversion ofthe declarative knowledge into the produ non,
fig whereas the encoding of the decaraiv knolede ake toe 8
Strlguforward string
in una form our xperine i any domain, eng incon
Sas motteo avec baa ocd ne send
‘tov tnowee nour ye.)
We concur withthe common assumption that earning a sil an be ge
erally viewed ar encotng of instrcton (nthe form of delctive kno
fu) followed by procedualation of some Kind. However, out research
{oust oa the encoding of iasevton, because or leading conjecture
‘hat how well inliduas earn to slve problems ede largely to he “coe
‘letenes with which they have encoded the stration, rather thaa the
[Xficeny wth whieh they can conver the encoded Istruclons ino a sil
‘Anderon (1997) also alluded to the iniiance ofthe encoded represen.
‘on when he suppstd tha,
weak probtensoing maths fie analy ca beach more ete
"hey opt on atch epsetton of he Kredi (5. 208)
Since ou work focuses onthe elation between the completeness ofthe
representation ofthe knowledge that students aque a hey read from a
{ex and problem solving, we expect tht thi feltonship canbe revealed
[yexaminng enna ferences We conjecture ha differences students
list salve problems may arse from dfeences in the was they Ur-
‘esand and learn fom et ncudig the worked-out solution example)
‘Ths, nstead of investigating onl the problem-solving dies toes
encounter, we examine, ination, the diicuis stadents night have i
understanding the worket-out soltion examples in the text, prior tslvng
problems
‘Alinough we view the examination of learner eiffereces a metodo-
logical may by which we can study the relation betwen the Knowledge en
faved and the kil of solving problems, we are aso ilrestd in lar
itferences from the point of view of tying to understand the aequstion of
teenie (Chi, late, & Farr, 198). Whe i takes fone pvods of study
tnd practic before ane ean Become an expt In any dom shart
tensive pratce and study ia necessary but not suiient conn for
tecoming an expert. Thus examining how students ifr the way they
lear new materials may ao shed let om the pacts of shill acon
that determine whether an individual wl owl ot achiev expert,
"We chose to Tocus on students lerrng ofthe work-out examples,
based on theoretical, empiri, a well as isretiona reasons, Theo
cally, theres some conteovers in there concerning how generac
ions are induced from examples, See Dieterich & Michal, 1983, fora
review othe At ieraure, and Murphy & Medi, 1985, for review ofthe
Pycholoial hierture) Thee are two views, Theories based on im
Tany-sased approach aim tha generations are developed by inducing
21 priniple from mulple examples. Such + principe would embody the
(Seti Features shared by all the exampes.On the oer han, theres
based onan expanationased approach (Lewis, 985; Mitchel), Kele,&
Kedar-Cabell, 1986 clan tat generalization can be aban Irn singe
fora few examples. This can be done by consrsting an explanation (or
‘roof thatthe enample Ian instantiation of» previoly given ovine.
‘Although many theories, including Anderson's ACT, can only generalize
from mull examples, there i empire evidence 10 show that student
‘am often generalize fom asingle example (ho & Anderson, 1983; Kirs
‘Boral, 1986. However, what tc i
|. How stdens construct explanations fom thir domnin knowledge in
lozer to prove that the example ean instanceof principe,
2, What isthe nature of ese explanation;
3. The lationship between the depres of undertanding they have of the
domain knowiedge and the explanations constricted;
44 The nature of ther domain understanding ater intantating the ex
ample,
‘Thats, exlaaton-ased theres asume that in understanding an example
and geserasing rom a det mst aeady have complete ndersanding
‘ofthe domain theory. Whether or no hii trac ean ope uestion, Out
‘Stady should provide evidence on the eeditiity ofthe theoretic! sump.
tion underlying am explaaton-bsed theory that complete understanding
‘ofthe domain theory s necessary in ede Lo contract explanations, Thus,‘vo theoreti! issues are inolied her. First san generation of the
Brinciple ccsur wih a few vesis mulpe examples? Second, is complete
‘derstanding ofthe domain theory neessary inorder to constuc expe
‘ations? Our empirical work should shed some light on how generalizations
an emerge fom lesaing few exemple,
“here aera inthe empirical and instructions erate as well
out leaning rom examples, Epil evden sexing to accom
late showing the importance of examples in learing. Rede, Charney, ahd
“Morgan (1885) for example, found thatthe mot effective manne for
structing students how to sea personal computer are the which contain
‘amples. LeFevre and Diton (1986 found that stants stalls peter to
te the example information and nore the write instruction whe ear
ing a procedural task. Piroll and Anderson (985) also fund that 18 of
thei 19 novices relied on analogies fo example in the ery sags of ern
Ingo propram ecutsion Beside there nboratony ning’, Vn (1988)
ko provided indirect esdence that examples ae emportn in reer cl
"oom faring, He fund tat 85 ofthe sptemate ers in aint,
‘alleced from several thousand students, could be explained a dering
from some typeof example driven lenny proces.
‘On the other hand although both stents and isructonal materials
rely heally on worked-ont examples aa instrument fr learning, the
lahorsory research which det examine the foe of exampeslaions
‘on problem solving found that students who have dled erampler en
{Snot sive problems that rele avery sight deviation from te example
Solution Eyton & Haman, 192; Reed Dempster & Enger 1985 Sele
‘Cooper, 1988). The dscrepancy between those sues which show tat
Sudents perform beter fom text materials that cota example, and
‘owe studies which show tat students often fail to generalize fom exam
‘les, may be cused bythe dearest which stents understand the samples
Provided. Generally inthe empirical sues cited, no auestent is made
bout how wel students understood the examples, Ar Pell nd Anerson
(0945) noted, although most ofthe stdents wrote ew programs by aalony
{oesample programs the success depended on how well te studets under
sHood why the examples worke.
‘We propre ha stdets ear an understand an example vi them
‘os they sive while syng i. We hypobese tat such ei-explanaons
ae important and necessary lagey because examples typical contain &
Sequence of userpcsted actions, Son (1979, p 2), for example, noted
‘Geer pking textbooks ne mich oven elie he a
‘truths oF of are ttn syn sang spat whe hse es
ty be ual insta proble. The action the productos wee a
‘seroma nthe domain ofthe exo sei oo sys,
they are no reconnected wi the candi that sald evoke en,
Wie can provide a concrete example ofthe inadeqsacy of worksd-ot
examples by taking one example fom the fifth chapter of pis txt
(GHatny a Resi, 1981), ar shown in Figure 1. We san sohe lack of spe
Eetcation of the exit contions under which the stows shouldbe ene
‘thy one should "consider the aot athe junction of the tre strings fo be
{he body." This is ere! pee of Information because pie hat t
{is foeton (as opposed to the Blok), he sum of he forces zero. Sach
lack of speciation of the exp conditons for action sre throughout
‘heexame n Statement 6 how does the sider Know hat FF a Fe
stealth forces acting on the body, and tha there ate wo other? Statement
‘isessenly a restatement of Newtons Fit nw, ut equ chaiing
several Inerencestgeter, and trasating them into an eauaton e+
‘because the body isa res ihre are a extemal forces, therefor them of
he ores on the ody mst gal 2to)- Statement sol anexpane
shy are the aes choren a och? Ie lea hat the solution sep tin an
{rampleare pot exp about the condons ner whieh the acton spp
Tn order to learn with understanding, stdent neds o overcome thei
compltenerof an example by deeing conclusions and taking inferences
ftom the presented information (Wictlgen, 197). To do s, 4 sadent
seeds Lo provide explanations ether overtly or covery for why a partlae
‘tion taken. Only then wil the radon be able spoy the aed prox
‘sluretonomisomoephic problems tat do nota exactly the contons
{tthe espe sluton, Thus, we sees tha good student "understand
“anexamplesoltio and wilsoceed in generlrng bose he ose make
‘comtiou effort oascetin the conditions ofaplcton ofthe slo
Steps beyond whats expt stated. Todo so the stent mst "expla
how the example iasaniates the principe wich emp.
is ifct 1o define what “understanding” spent in the content of
learning from examples. Operationally, ou study was designed to shes
Ainderstanding with thre eave
1. Solon to omombi problems
2, Solutions for ar raster problems;
3 planation generated whe stodyng exampks,
“The weakest method to detec understanding (efit meat) isto observe
how successfully students solve very similar problems, since very sma
problems ean often by solved by a simple syatace mapping af the example
Procedure to the to.besolved problem. The second metod is sei se
‘ents can successful use the rncpe involved inthe example in adiTerent
land more complex problem (lar tans), Because such problems prevent
‘0 ie cape ge at by fans4: Conor ha kot he anton hate eg be ha os
4: Tbs ramon ent under te acon hehe er own Fy,
4 Roppre woe gen the mogns fone a tase aco.
SNES fe eps fe sar?
etait
EcGrsckerasee
a aa
EES tue amt to
OEE augeenynmemnerteytrnn
‘Fm andes a 70H and ees New York
1k hoy Sore a by Sin ey Sonat poms
‘stents from being able to sve then via ayntactic mapping. However,
ova certain extent, stdying far transfer reveals only that understanding
‘sits ad alloms one osc what conditions faite
The method which permits the most ect assesment of understanding
‘ofan example ft eine the explict explanations that nodes provide
‘hile studying Epinning ia mechanism of study tat allows students to
Inf and expcae the conditons and consequences of exch proce tp
inthe examples wells apply the pnlples and defation of concep
Jusiy them. We posaate that explanations an reveal stoden” ner.
Standing y showing whether ont they Know
1. The condidons of appiation ofthe actions
2. The conseaences of actions
53, ‘The reationtip of stone to goals
4 The relaloasip of goals and ations to natural laws and othr pra
pes
“This atte focuses primarily on the anys ofthe explanations
nthe atl provedre ed, wetted student to over expan 12
esses what they understand, after reading every line of 4 worked
tample Often they ade wo comments at all However, when they id
Beerate a comment, his technique allowed us io yoke the explaation 10
Thestatement linea the example so that we could interpre he protest
nore ely
METHOD
Sebjecs
“Ten students (S males an $ females) were sled from responses to ace
‘ot advertisement. Eight of them were working cowards bachelor’ deers,
‘nit varying major. Two ofthe en sdens Rad atonal post-race
{tuning tn pyehology, None ofthe students ad taken» olege phys
akough al of them (tet one) had taken highschool ply,
wih ferential performance (ported grade) in that course. We intention.
Aly ehowe tedets witha range fabs in erm Of sade point average
fd SAT scotes, so that he could examine learner diferencs. Stents
‘ere pai for thee paripation.
Inthis arte we focus on how stdens study thee worked-ou examples of
‘robes dealing with the application of Newton's laws of motion and on
‘hw ths initia! leering relate fo thet ssoegbent problem olin, Hom
‘er, slnce the televant chapter demands subamive background jer
‘mater, he study of examples andthe problem-solving ats ate eibelded
‘within fonitainal study in which students nou laboratory staded Newtonian meshanss, Stodents etoed the new material ina way they would
normally do wen studying on their om. Inter of the amount of tne
‘devoted Yo sodyng, the ae of self-pacng, and thew of sade bie,
suchas highighting significant parts he ext orerending. The aborsory
Terning fered from the may students would typical len on the ow
in that ll the learning took pce inthe laboratory (without ar estar)
and that student gave talcaloed protocols while studying examples sd
Solving problems. The student pent between #29 hours fo complet the
Study, spread over several weeks. Figure 2 preceis a iagram summarizing
the experimental procedure forthe we sty. Bally the study om
Sted of two major phaser knowledge aquisition spb olin.
Daring the fst pat ofthe knowledgeacqison pha, abject stad
the necessary Baskeound subject mater, covering he topics of mest
‘ment, vectors and motion none dimension. These matras recovered in
the at thee chapters af Haliday and Resick (198). The fourth caper
fon "Motion na Plane” was omited been dd ot have adits bearing
‘nearing Chapter 5, she target caper on "Palle Dynamics” Deion
‘sr this Kind, a6 well as designing of guesions and problems, were made
‘ter conrltations ith phot,
For each of the tree backround chapter, students rend through and
stud the chapter. They weterequesed to record on 2 separate sheet of
pe any questions whi arose ving thet stay. Each ehaper lio con-
falned ante to stop and et the experimenter in arder to ea verbal eo.
tocolof one worked-out example. Thisexecse was meant mail 5 practice
In ghing protocol, When the students lived they were ea tobe tested
‘nthe atrial i thechaper, they nailed the experimenter, ahd were thet
reaued to produce corec answers toast of derive, guaiatve and
‘quantatve questions, ven in tha ode. Declarative qustios wee de
‘Senet avn he recat of cial facts from each chapter, for example,
“Whar se ditference between asclar and vector?” Someot these qs
tions came rom the terbook and some were despned by te esperimeters
‘The quaave questions were desiged tosses reasoning and inferences
shout the concepts in each ofthe chapes without reference to quails,
for erage, “Can an object have am casteard velocity while epeiencng
westward aceeraion?” The majority ofthese problems were taken rom
The teubook, and several were generated by the experimenters The aan
tative questone nesed procedural ils forquanative protien sling,
for example, "Two bodies bein fe fall om fsa he same height, 1.0
seayart How long afer the rst body bins all il the two ods be
1D meters apart" Al ofthe quantitative problems came from tose Pre
soled in the ent atthe end ofthe chapters
=
se
a
ieee 2A sox dsng
nie a heAfter solving each problem st, students rere the problems othe x=
periente for rading. If no erors mere made, they proceeded to the net
Set of questions If an eror war nade, the vertone were ered fr cor
reaion The incorrect anwers were identifi and students wee referred to
‘hore secon of the text which addsered ese questions, For tneoree
‘antaive quertions the correc final answer was lo provided, analogous
{0 the common practice of looking up the answer i the back of he book.
‘Students then nlempted to cocet he salutons and rested hf
trading hee were xl ror on the vcond ty, stdents mee proied
‘ithe workedot slstion forthe incorset answers Students then saded
{he worked-out Solution, nd fei had een emoved, were asked one
‘heir answers and/or to reproduce the coret solution fram memory. IT
ser these two tes the answer wae il nearest the experimenter xpaed
‘the worked-ot soit and the student ad to eprodice Ths the fs.
twee chapers, consisting ofthe background subject mater, were sud
‘un all the stents could eorety solve ast of deat, qualitative
‘a quantitative prosems evant fo each chap,
‘Dring the second pat of the hnowldge soisition phase, students
stud the trae chapter on “Panicle Dynamis” (Halday, & Remick,
{Chapter 3, 1981), Studying the age chapter proceeded exactly as for he
other chapters, encept tat siden wee requrd to ansner coves he
felevant deiartive tnd qualitative questions before studying the thee
‘worked ou examples, Ths was done ore tha the dete had aegdred
{he relevant declarative knowledge from the text. (The quantitative prob:
Teme for Chapter $ were no solved uot the second phae ofthe sly.)
“These of examples was the major focs ofthe knowledge acqulstion
phase sudentssadied tne worked-ut example, een direst fom he
{ext (See Figure! foran example of oe ofthe worked-ot soltions Exch
‘sample solution represented a "19pe"of problem. There were hee pe:
String, intned plane and pole probes. (Se Figure 3.) Stade talked
‘ut lod while studying the examples, and ther protocols were ape
lem. The fst set consisted af 12 "somorphic problems.” wih 4 prolese
corresponding to each "ype f exams ee the ead ares in the prob.
lemslving phase of Figure 2). These problems nee designed tov In
thee dere of smarty tothe worked-ou examples studied inthe previous
Das. Fgute 4 shows one set of fomorphs covtespondig tothe eng
“The second second of sven “chapter” problem, which were prod-
Jems taken det fom the end ofthe age chapter. ners orcteria
for somoephibm to example problems, thes problems can be considered
‘Mar tanster problems Stidets solved bo ets of peblens while giving
te rom fandom Pc p77) Way a Rw, 8, Ho
So atte Sw Dv by i ey Soe pnt sa emoe etme pe 9
“say edo wpm W008 1 2)
ue do Buu ange sa sido spss est Hap poy 6ume 6 HONG
‘Aunts pe6ueug
0 ow woo
am te 6 op wooo
‘toy Gabuey 6 pad won eu.
peop oa seu 09
eu outa 988 nn By 808tal-loud protocol, and no feedback war provided. Only he probe
solving protocol of heirs sto hee isomorgh problems have teen a
lyzed fortis atic
RESULTS
As nica, his article will report onthe sil of he ale ofthe pro
tocol taken while students study the example soltons peed fo the
tart chapter, as well as thease of how they te the examples in the
pyoblemscving protocol ofthe womorphic problems the wo shaded Howes
In Figte 2. Indvidal ferences wil be eported by eonatng the per
Formanc of "Good and "Poor students Fhse two groups were delned
post ho, using thl problem soting succes onthe 12 somone and
{he 7 chapter problems In scoring the problems atime spe were not
onsite as rors and partial softions were ceded proportionally. Al
Droblems recived te seme weigh 20th! the maximum sere fr the io
[orphic problems mas IZ and he maximum coe forthe chapter problems
tra 7. (All coring, Menifcation and clasifeaton of protocols were pe
Forma by wo ade, wih mean interrater flay acon al te onlys
ranging ffom 86te-bi%.) One aden di 0 poly (he eter generated
{ny protocols nor was he able to solve any problem) tha he ws minated
from the analyses. This left ise student We analyzed the rel othe
{oprcoring fou an btton-scoring four suadent, and omit the data of
the "Midale” student There were hat four stadents im each troup. The
‘mean sucess of the Go students was 82% (26% for the omorphic pro
lems and 68% forthe chapter problems). The mean mcs ofthe Pars
ens was 46% (62% for the semorpic and 30% Tor the ehaper problem).
[LEARNING FROM EXAMPLES: STUDYING PROTOCOLS.
‘Amount of Protocols Generaed: Units of Analyses
(Oar est analysis began with avery gross count ofthe amount of rocco!
lines students provided wile stdyng the example solo. A ie of pro
toca any statement tat nots ft seadig ofan svar ines no
fy conversation cared on mth he experimenter tha does at refer tthe
fubject mater of the example (Le, "Do you ave a alto hat can
er”). Tes ako nota eppone to the experimenter’ segue to speak
NSE a hepa tate emai oni on
1a Yaga gt lh
Armin ol fl ener Wie hing nora Sig
mere ohm tm
ae Te mb eset vege pr sen pr ean as
louder Table 1 ia anseript of one student's protocol studying the exan
Ble solution as shown fn Figure I) 9 which we esiged a ie number to
‘a pause boundaries, Line numbers were asigned to statements correspond:
Ing more orks to pases. Ths exams snow tal of 1D ies Nae
that heente we are not necetaril sng sentence boundaries a line Bo
Insert between diferent acts, suchas reading. "
Table 2 (op pane) shows the average numberof ins generated per ex
ampleby the Good and Poor student 8 wel asthe average amount o ime
took them o study an example. These differences re contrasted with the
sverage number of ptocal ies gneate and the amount of time aken 19selve an somorphie problem (bottom pana). The Good students generated
‘couidevably greater number of protocol nes than the Poor stages (142
Ties v.21 Kins, (6)=1.97, <8). These protocol generating Stee
necesaily fed the Good students to spend more ime on each example 8
‘ell (UD sve 7.4 min, 16) =216, p=).
‘Werule oat the pont that Goad sindets re ply reac
or figent, thus producing more protect statements, sinc, in conta the
‘umber ot tine they geneated wile ses prob x aproximall the
ame atthe Poor stadt (v.12 lines per problem, ee Tale 2, bottom
ane) This sugges that he stents pnerted as many ine of poets
1 they deemed necesary i order to lear the example, Likes the fat
‘hat the Good stents took mae ine studying each example sot than
"he Poor stants (3 min v8.7.4 mi) alo elects te choke Co spend
‘more time on each rare rather tana tendency odwell uanecenarly on
{he examples, snc the amount of ine they pen song each omephie
‘problem about he same asthe ie pent bythe Poor sade (138 gin
S183 mn).
“The assumption we make that protocol sateen are generated ome
Snowiedge or inference i being processed or enstuced in memory, mo
siterent than assumptions made with veterence fo ober yer of dependent
measures For eampl, the estomary assumption undering verbal pro
{ovol analyses is that the prclem slr isang what he tang about
‘or dumping the content ei or ber working memory (Newell & Simon,
1972). Thus longer protocols simpy eter toa renter dene of rrocesin.
‘Sinai nreserch win eye movement ooo the assumption tht
the stents processing the locaton at which he o she i ating (Just &
Carpenter, 197) Thy lange isations imply that the stadet sending
ror time procesing that location. Therefore, we view the rete smount
{of protocols produced bythe Good students ara natural conequence of
‘ating to understand the solution example beter, ater than the pos
Diy that they are more arcuate and nt
tight be calmed tht the ean that Good student earned more is
jist that they spent more time studying. Homever, ths snot very deep ex
Plsvation, but merely reatement ofthe corrlon. The important ques
tin s what he Good students whe hey were sting. Ar wil seen
‘ory thee are such lage ualtative difeencs in what aden id
‘nile sadyng examples thatthe simple, shallow, exlantionthat stud
Ing twice as tong mates on eum twie at mech quite impnusbe be
casei cannot expla he ther ferences a student" learning behaviors
‘Although we claim that Inger protects do not necessarily imply vr
bosty using lines as uit of anal doesnot seem adequate for capone
‘tha going on. That, a analysis al the evel finest fie
{o characterize te nature of thee comment, Iad many ines sem 0
tele toa ingle de, and thus we futher collaeed te lines ito unit ex
Dressing singe idea, For example, the 13 protocol ns in Table 1 have
bee ecoded into 7 "idea (Statement I-VI. The boundaries were placed
‘on the basis of aitfereat des, which are often separate by poses or
ferent sort of actives inclading reading o experimenters interjstons,
‘Using thie ort of parsing, we Sen Table 2 (idle panel that God st
ens" protocols have been rsced from 142 ines t31.9 ens and Poor
Studens" protocols remained roughly the sme (ht iy 173 eas v6. 21
lines Tha, ina ease me are penalizing the Good sens for ben atic
late. Wha the dtantaly shows thatthe Good stent often dwell quite
i oma single exer Benne thy rele tht they do ot undestd
ior they want to provides more complete explanation (st net ection)
1 the ideas are jst hare to expat succinctly, Even with th uit of
the Good students generate asigncaly greater numberof ese
than the Poor students (59 vs. 18, 16)= 1.98, p08), Henceforth all
talyses wil be Basel onthe numberof lea states
Kinds of teas Generate
Am idea statement canbe clasifed into thee types. (See Figure $ fora
‘brekdown of al the lasiteatons the dacueed) We reserve the fem
xplonaion to refer ony to those Weas which say somthing vbstantve
shout the physics dsesied inthe example statement The flowing cot
‘ment would be consiered an explanation
Ummm, hs woud make sea, suse he conned by sing
Ihe sig’ goin 10 be rece the cart ing Yo be mond, th
‘sefoce of he inn in be deemed
Other examples can he sen in Table 1
Aiden ie coniered to bea monitoring statements refers to wats of
ompeeasion For example, remark such se
eam se now how hey i
va ang ou wth Poi
Would be considered monitoring statements,
"A third eategory tae sevecl ore yes fies, mostly paraphras-
Ing, mathematieal elaborations, and metatatepcatements Paraphrasing
‘would be comments that ether restate what the example ie std or put
Jnto worés what s shown pltorial For example stent who Fad the
ame ine
i eae lhe fre ating 0 he ods.reprion ) Femey”_ Peper) Faia”
Panis temo a rm 2
(ot 2 ms ® u
fad then remarked that,
Thee ono mor forces
‘ould be paraphrasing. Paraphrasing into words what wat depicted inthe