You are on page 1of 13

Interrelationships among Prior Knowledge, Inference, and Language Proficiency in Foreign

Language Reading
Author(s): Joann Hammadou
Source: The Modern Language Journal, Vol. 75, No. 1 (Spring, 1991), pp. 27-38
Published by: Wiley on behalf of the National Federation of Modern Language Teachers
Associations
Stable URL: http://www.jstor.org/stable/329832
Accessed: 19-06-2017 00:46 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

National Federation of Modern Language Teachers Associations, Wiley are collaborating with
JSTOR to digitize, preserve and extend access to The Modern Language Journal

This content downloaded from 128.2.10.23 on Mon, 19 Jun 2017 00:46:16 UTC
All use subject to http://about.jstor.org/terms
Interrelationships among Prior Knowledge,
Inference, and Language Proficiency
in Foreign Language Reading
JOANN HAMMADOU
Department of Languages
University of Rhode Island
Kingston, RI 02881

COMPREHENDING A SECOND LANGUAGE IS A they, the readers, create. Pearson shows that
complex process that occurs with few unlikeexternal
what generative-transformational theory
signs along the way. Out of recent research might have predicted,
has children recall complex
come a growing understanding that tests compre-
better than simplified ones. The cohesive
hension is not just understanding words, nature of asen-
complex text facilitates the model
tences, or even texts, but involves building
building a
process within the learner. Goodman
model within the mind of the comprehender. describes reading as a "psycholinguistic guess-
In the past, comprehension research inghas
game"been
(18) and is credited with the early
widely influenced by the generative-transfor- concept that reading actively involves readers
mational theory of language, which(26). described
comprehension as building meaning from the
Several first language (L1) studies document
smallest, simplest sentence-level features. the influenceThis
of prior knowledge on comprehen-
view predominated for some time despite sion. Anderson
some and colleagues (4) gave
evidence to the contrary. For example, ambiguous
in 1932 passages to music and physical
Bartlett objected to research that used education
simpli-majors. With one that could have
fied stimuli such as nonsense words to test the described a woodwind quartet or a card game
unsimplified response of comprehension. Heand a passage that could have been a wrestler's
tested subjects' recall of a unified text over a hold or a prison escape, the special background
period of ten years to show that memory andof the readers guided their understanding of the
comprehension were a model building process.piece. Bransford and Johnson gave readers two
In his day, however, his work was largelyambiguous reading passages and demonstrated
ignored. the importance of context (a title page or an
Currently, more interactive models of com- illustration) on ability to comprehend the
prehension have been developing. They passage.
attempt to describe features of language (lexi- In second language (L2) research, back-
cal items used, syntax, topic, etc.) and features ground knowledge has also been proven to play
of the comprehender (e.g., goals, prior knowl- a significant role in comprehension. Johnson
edge, intelligence, proficiencies) and ways in gave ESL readers a passage on Halloween and
which these features interact in the process that demonstrated that cultural familiarity had a
is called comprehension. Spiro describes lan- greater impact on comprehension than the pre-
guage as "a skeleton, a blueprint for the con- teaching of vocabulary. Bernhardt (7) demon-
struction of meaning" (28: p. 245). His research strates, with non-native readers of German,
(27) shows that readers distort the surprise end- that lack of cultural knowledge impedes com-
ings of stories when asked to recall them to prehension of text.
make the ending fit the growing model that General topic knowledge, not just culturally
bound prior knowledge, is significant in L2
comprehension. Hudson's research demon-
The Modern Language Journal, 75, i (1991)
0026-7902/91/0001/027 $1.50/0
strates that cueing readers about an upcoming
S1991 The Modern Language Journal topic with a picture aids comprehension more
than teaching vocabulary and is particularly

This content downloaded from 128.2.10.23 on Mon, 19 Jun 2017 00:46:16 UTC
All use subject to http://about.jstor.org/terms
28 The Modern Language Journal 75 (1991)
Experts now agree
useful for low-proficiency students. Lee thatapplied
inference is a cogni-
tive process used
Bransford and Johnson's passages to to construct meaning. Infer-
non-native
ence is, inthat
Spanish readers and discovered other words,
their a thinking
com- process that
prehension was affected involves
also by reasoningcontext
a step beyond(the
the text, using
availability of a title or angeneralization
illustration). and explanation. For example,
The theme of prior knowledgein the text: She went to
and the hairdresser's
its role and her hair
is blond now,
in comprehension is repeated in many discus- the reader usually reasons a step
sions of language proficiency. Most words
beyond the individual current
on the page to sur-
theories of foreign language miseproficiency
what happened in the hairdresser's salon.
incor-
porate prior knowledge withinTo do so, the the reader
larger usesrealm
generalizations of
typical events
of"context" as either influencing in the above setting
proficiency or and explana-
tory reasoning
as a part of the very construct of how those typical events
of proficiency.
would apply
The ACTFL Proficiency Guidelines to the words/text
describe profi- at hand.' The
importance
ciency according to five scales of inference to
(speaking, comprehension of
listen-
ing, reading, writing, andeven the simplest
culture), of texts
each has long been
with
three dimensions (linguistic functions,
demonstrated content,
in L1 studies of comprehension
and degree of accuracy).(e.g., 3; 30). Some
Canale andviewsSwain
of inference are
offer four different scalesmostly text-based and describe
(grammatical compe-readers' infer-
tence, sociolinguistic competence, discourse
ences as bridges among propositions of the text
competence, and strategic(e.g.,competence)
17; 22). Other views ofwithinference are
mostly model-based
three dimensions (basic language, and describe readers as
communi-
cative language, and autonomous language).
quickly building whole models of understand-
The Council of Europe (15)
ing ofdefines language
texts that are subsequently refined based
proficiency as varying according
on the constraints to learner
imposed by the text. Infer-
needs; Cummins identifiesence
two within this theory of comprehension
dimensions to is
the use of language (context
viewed asdependency
one of the key cognitiveand
processes by
cognitive involvement). A which readers formulate
common and refine their own
denomina-
model of the text
tor in all the preceding definitions of (13).
language
Much
or language proficiency is theof the current studyof
influence of inference
"context." focuses on the identification of inferencing
In the ACTFL Guidelines, context of language strategies used by different types of L1 readers.
encompasses the task that the language user Little work has been done on the construct of
must perform (function) and the topic or con- inference in L2 or the interaction among infer-
tent of the language. In Canale and Swain's ence, background knowledge, and language
model, context influences the sociolinguistic, proficiency for L2 readers. Afflerbach notes (p.
discourse, and strategic competencies of the 40) that only readers with prior knowledge of
language user. One's performance may depend topic were able to infer/construct main idea
on or be based on the ability to understand lan- sentences when none was provided for difficult
guage in different sociolinguistic contexts and L1 texts. He concludes (p. 43) that prior
different modes (casual conversation, argument, knowledge may be an important factor in accu-
recipe) and to compensate for breakdowns in rate inferring. In contrast, Phillips finds, in
communication (11: p. 339). Cummins identi- studies of inference strategy use, that strategy
fies a strong relationship between language use differs between readers with high or low
proficiency and context because at one extreme background knowledge only when readers are
"context-embedded" language is supported by proficient; no differences occur if the readers
a wide range of types of cues from the situa- are not proficient in that skill. She concludes
tional to the paralinguistic that give redundant (p. 17) that prior knowledge is insignificant in
information; and, at the other, "context- the absence of reading proficiency.
reduced" communication occurs with language
that relies on cues purely linguistic in nature.
RESEARCH QUESTIONS
Although each model gives the notion of con-
text a slightly different role and degree of Two parallel studies, one in French and one
importance, all suggest some relationship in Italian, addressed the following research
between context (and its component, priorquestions: 1) Do non-native readers compre-
knowledge) and language proficiency. hend familiar topics significantly better than

This content downloaded from 128.2.10.23 on Mon, 19 Jun 2017 00:46:16 UTC
All use subject to http://about.jstor.org/terms
JoAnn Hammadou 29
they do unfamiliar protocols, they wereones?
handed a list in English
2) D
between familiar and unfamiliar recall lessenof the topics they had just read and instructed
to rank order these according to their prior
for more proficient readers than for less profi-
knowledge of the topics.2
cient ones? 3) What are the qualitative differ-
ences in L2 readers' inferencing according toResearch studies have operationalized prior
topic familiarity and language proficiency? or background knowledge in a number of dif-
Subjects. Participants in the study were stu-
ferent ways. These range from simply classify-
dents enrolled in levels 101 and 104 in both ing a subject as unknowledgeable in a topic a
French and Italian at the University of Rhode
priori (e.g., the history of Halloween for newly
Island. The study sample consisted of eighty-
arrived ESL students) to extensive testing of the
subject's knowledge base. In this study, the
nine students of French (forty-one in the begin-
approach was to maximize probable prior
ning level, forty-eight in the more advanced
one) and seventy-seven students of Italian knowledge differences by selecting topics from
(forty-three beginners, thirty-four morethe ad-top, middle, and end from a list of thirteen
vanced). All subjects were native speakers of presented to another sample of Begin-
topics
English. The 104 level (fourth semesterning of and Advanced subjects. Therefore, sub-
study) was used as the more advanced profi-
jects were asked to assign a number from one
to three to each of the three topics. Their
ciency level rather than a higher course level
because data from this study were also being
familiarity was judged to result from their prior
used as part of program evaluation of thisknowledge.
ter-
Scoring. A list of propositions was created
minal stage in the general university language
from each of the original passages based on
requirement. The category names "Beginning"
pausal or breath units. These units are based
and "Advanced" are used for comparison within
this essay and the latter does not refer toonthat
pauses chosen by native speakers during oral
reading. The ending of each pausal unit creates
term as used in the ACTFL Proficiency Guidelines.
a syntactically related unit.3 For example,
Materials. Reading texts were all authentic,
unedited, expository passages selected from Le
Wednesday, February 8/the American Secretary of
Monde and Paris-Match for French and from State/James Baker/called on the Sudan government/and
Lettere dall'Italia in Italian. The selected passages
the southern rebels/to initiate a cease-fire/immediately/
were from 178 to 285 words in length. toBased facilitate the conveying of help/to hundreds of thou-
on a survey of similar university students, sands of people/touched by the famine/in the south of
topics were selected that appeared likely to
the Sudan.
maximize the possible range of familiarity dif- Each recall protocol was scored based on the
ferences. The titles selected for all French sub- percentage of the possible propositions con-
jects were: Sida: la leqon de San Francisco [AIDS:tained in the protocol. Due to the inherent
imprecision of recall protocols written in Eng-
the lesson of San Francisco], L'Avion secret demasque'
lish about texts in French or Italian, credit was
. . . du ciel! [The secret airplane unveiled . . . from
the sky!], and Soudan: guerre civile etfamine [Sudan:given based on sufficient gist of the meaning
Civil war and famine]. The titles read by all of the proposition rather than verbatim recall.4
For example, full credit was awarded for "put
Italian subjects were: Lo sport tra spettacolo e cul-
tura [Sports between theater and culture], Ricardo an end to the fighting" for "to initiate a cease-
fire." Furthermore, "Secretary of State" without
Bauer: antifascista e educatore [Richard Bauer: anti-
mention of "American" received full credit, but
fascist and educator ], and Il cinema dietro le quinte
[Backstage at the cinema]. These articles were thenno credit was given for "Secretary of America."
retyped to a uniform paragraph format without The researcher and three trained assistants
visuals. (See Appendix I for a sample of each.) scored the 166 recall protocols. A sample of the
Methodology. During a regularly scheduled recall protocols was scored by two raters with
class session, all students read three differentan inter-rater reliability of .86. Differences in
texts that were identified as French or Italian the awarding of credit were resolved by discus-
newspaper articles. Ample time was provided sion.
for subjects to read and reread each passage.In the second stage of each study, all written
recall protocols were examined for propositions
Subjects were instructed to recount in English
what they had just read after each passage that did not match propositions from the origi-
before continuing to the next article. When nal text. A list of inferences, or propositions
outside the narrow bounds of the original texts,
subjects had finished writing all three recall

This content downloaded from 128.2.10.23 on Mon, 19 Jun 2017 00:46:16 UTC
All use subject to http://about.jstor.org/terms
30 The Modern Language Journal 75 (1991)
was created. These propositions
TABLE I
were searched
Mean Familiarity
for logical readers' inferences Ratings and sup-
(conclusions Recall Scores
ported by the text) or illogical inferences (con-
Topic Proficiency Familiarity Recall
clusions unsupported by the text). The inter-
Italian Sports 1 1.95 1.88
rater reliability for coding inferences was .93.
4 2.24 7.31
Cinema 1 1.88 4.93
DATA ANALYSIS AND RESULTS
4 2.12 13.48

Biography 1 2.16 16.11


Six analyses of covariance were conducted
4 1.65 23.47
for the six passages read.5 The factors in each
French AIDS 1 2.71 14.44
analysis were proficiency (beginning or ad- 4 2.79 23.54
vanced) and familiarity with topic (most, mid,
Stealth 1 1.98 23.37
least), and the covariate was grade in course
bomber 4 1.74 26.51
recorded on the traditional four-point univer-
Sudan 1 1.32 17.58
sity scale (0 to 4.0). The grade in course vari- 4 1.60 25.09
able was included to enhance design efficiency
by reducing the variance due to the already
recognized factor of success in This was not the case, h
class. The
dependent variable was the percentage of
language was a strongly
propositions recall. recall of all passages in
values of
The French passages in order of recall .0001 were recorded for four of the
were:
1) Stealth bomber; 2) Sudan; 3) six recall The
AIDS. sets. The more proficient readers
clearly
ranking by familiarity rating of each comprehended more of every text than
passage
theStealth
topic in French was: 1) AIDS; 2) novices. Differences between the two
groups
bomber; 3) Sudan. The best recalled according to familiarity with topic were
Italian
not evident,
passages were: 1) Biography; 2) Cinema; 3) however.
Sports. However, the overall ranking by of proficiency and of familiarity
The effects
were
familiarity rating of each passage the same
topic in for both language groups. In
Italian was: 1) Sports; 2) Cinema; both3)language
Biog- groups language proficiency was
raphy. In the Italian study, subjects awere
strongly
moresignificant factor and familiarity
evenly divided in their selectionrating was not (Tables II and III).
of a most
familiar topic, and the Advanced and TheBegin-
third research question examined stu-
dents'
ning groups did not agree on the order of inferencing
topics behavior. Chi-square tests
as they did in the French study. For were conducted to determine if inferencing was
students
of both languages, the ranking of therelated to language proficiency or background
passages
knowledge
according to recall score and by familiarity of topic. In this area, clear differ-
rat-
ing does not match (see Table I). ences
Thisemerged
dis- beyond simple amounts of recall
crepancy is upheld by the analysis ofbetween novice and more proficient readers.
variance
for each set of readings. FamiliarityThe overallisamount of inferences was signifi-
rating
cantly
a nonsignificant factor for all reading setsgreater
and for the less proficient readers
only approaches significance for the than for the more proficient ones (X2 = 4.063,
subjects
p= .04 in(See
reading the Italian cinema passage. French; X2=4.153, p= .04 in
Appendices II and III for ANCOVA Italian). French readers produced 124 infer-
tables.)
ences in the
In response to the first research question, Beginning group and 111 in the
this
Advanced group. Italian readers produced, re-
sample of student readers did not comprehend
familiar topics significantly better spectively,
than un-121 inferences and 70 inferences.
familiar ones. Neither study produced a significant rela-
tionship between background knowledge and
The second research question asked whether
amount of inferences in recall when all infer-
the difference between familiar and unfamiliar
ences were studied together. Among the Italian
recall lessened for more proficient readers than
readers, however, a relationship between lan-
for less proficient readers. Some models of pro-
guage proficiency, self-reported familiarity, and
ficiency suggest that differences between recall
logic of inferencing was evident (Table IV). In
of familiar topics and that of unfamiliar ones
the only other results on inferencing where the
would be significant for beginning readers, but
that recall of unfamiliar topics would not be two
so studies differed, French readers recalled
more logical inferences (sixty-five percent) for
adversely affected for more proficient readers.

This content downloaded from 128.2.10.23 on Mon, 19 Jun 2017 00:46:16 UTC
All use subject to http://about.jstor.org/terms
JoAnn Hammadou 31
TABLE II

Mean Recall Scores by Familiarity and Proficiency for French Texts

Beginning Advanced
Familiarity
AIDS Stealth Sudan AIDS Stealth Sudan

Least 14.5 (16) 18.1 ( 8) 17.7 (20) 20.0 ( 1) 26.7 (23) 23.8 (24)
Mid 17.0 ( 4) 24.4 (26) 17.8 (11) 23.1 (11) 26.4 (16) 26.9 (21)
Most 14.1 (33) 26.4 (16) 11.6 ( 1) 23.8 (36) 28.8 ( 9) 23.1 ( 3)
N in parentheses.

TABLE III

Mean Recall Scores by Familiarity and Proficiency for Italian Texts

Beginning Advanced
Familiarity
Cinema Sports Biography Cinema Sports Biography
Least 4.5 (16) 2.3 (16) 16.3 (11) 13.0 ( 6) 6.2 (9) 22.8 (19)
Mid 4.1 (16) 1.4 (13) 16.3 (14) 12.8 (18) 8.5 ( 8) 29.7 (8)
Most 7.1 (11) 1.9 (14) 15.9 (18) 16.0 (10) 7.4 (17) 18.2 ( 7)
N in parentheses.

TABLE IV
FIGURE I
Observed and Expected Frequencies of
Inferences by Proficiency Levels Inferences by Proficiency

Proficiency 150

Language Beginning Advanced Chi-square Illogical


SLogic
French 124 (108)" 111 (127) 4.063*
40
Italian 121 (106) 70 ( 84) 4.153* Q 100

4,
aNumbers enclosed in parentheses are expected values.
p = .04.
50 -

CI

which there was direct support in the original


texts and Italian readers recalled more illogical
inferences (sixty-three percent) without direct French Italian French Italian
support from the texts. The greater overall Novice Advanced

inferences in recall by Beginning readers and


the distribution of logical vs. illogical inferences
are displayed in Figure I. a number of ESL studies
clearly shown the importan
DISCUSSION
edge to comprehension, the
prior knowledge is measure
Question 1: Familiar vs. Unfamiliar Topics.
important difference in th
Overall, the topic judged most familiar Finding
(sports a shortcut method
or the disease AIDS) was the passageknowledge least well clearly remains
recalled. Although seventy-eight percent study ofthe
all contradictions bet
readers rated AIDS as the most familiar of the ity ratings and actual recall
three French topics, the comprehension of the the question of whether a su
AIDS-related passage was the weakest of the reporting that topic A is mo
three French passages for all groups. The sub- than topic B was indeed ac
for the lack of significance
jects' rating of their own familiarity with a topic
was unable to predict their ability to compre- variable, therefore, may be
hend what they read about that topic. Although reports.

This content downloaded from 128.2.10.23 on Mon, 19 Jun 2017 00:46:16 UTC
All use subject to http://about.jstor.org/terms
32 The Modern Language Journal 75 (1991)
The strength of that familiarity
ciencies. If topic itselfor, more
could be included within
a single
important here, the relevancy ofanalysis,
that anyfamiliarity
interactions of topic with
L2 proficiency
to the text at hand may differ too or with familiarity could be
dramatically
from individual to individual
found. Someto hold
indication up
for the to
need to inves-
scrutiny within group means. Both
tigate in that groups'
direction exists in the variation
forced-choice ratings ofbetween
three topics
proficiency were
groups in the Italian study.
unable to predict the ability ofbarriers
Statistical the group
to such studytocould be lifted
recall a passage about thatby topic.
constructing These
measures find-
of familiarity that
ings mirror Bernhardt's (8) produced
findings continuous
that variables
a free rather than
association task that is ranked according
simple rankings, though the toequating of pas-
sages onideas
quality and quantity of related all factors
to a except
topictopic remains a
that subjects can produce is also not
stumbling block.sensitive
enough to predict comprehension. Assessing
Measurement Difficulties. It is hard to determine
prior knowledge of a given topic remains
why familiarity had no greater impact on the
problematic. novice L2 readers than on the more proficient
Question 2: Familiarity and Language Proficiency. ones. Among the readers of French, the differ-
Because of the difficulty in measuring topic ences between recall of familiar topics vs. that
familiarity, it is also difficult to assess the effects of less familiar ones did lessen for the advanced
of topic familiarity on the two levels of profi- readers on two passages. As hypothesized, the
ciency. Current language theories imply that differences between most and least familiar
more proficient language users rely less on prior recall were slight for Advanced proficiency
knowledge of topic and rely more on knowledge readers of the Sudan and Stealth bomber texts
of elements of language. This concept has been and were more pronounced for the Beginning
cited frequently as part of the definition of pro- group. This result did not extend to the other
ficiency. Proficiency is defined, in part, as a four passages, however. Among the readers of
measure of context dependency in measuring Italian, those who identified biography as their
comprehension of meaning. If that is true, least familiar topic actually outperformed those
beginning readers might be expected to show who identified it as their most familiar one. Al-
significant differences between recall of familiar though those who considered cinema or sports
and unfamiliar texts; more proficient readers their most familiar topic did better than those
might show much smaller (or no) differences, who considered it their least familiar one, the
possibly triggering a statistical interaction differences between the most and least familiar
effect. Surprisingly, no such effects were found. groups did not lessen for the more proficient
although the more proficient readers did com- readers. Again a resolution seems to be locked
prehend more of what they read, the differences in the complexity of the variables at work and
between most-familiar-topic recalls and least- their natural resistance to oversimplification.
familiar-topic recalls did not diminish with this A review of the content of individuals' recall
increased comprehension. Or, conversely, the protocols reveals one major stumbling block to
least proficient readers did show signs of lessen- answering the second research question in par-
ing comprehension, but it was not based on less ticular and to measuring the effect of prior
familiarity. knowledge in general. The readers of French,
One possible explanation is that students in for example, overwhelmingly identified AIDS
their fourth semester of language study were as their most familiar topic. To avoid influenc-
not yet sufficiently beyond the novice stage ing reader comprehension, this rating occurred
themselves to show such linguistic autonomy. after passages were read. And although most
Nevertheless, they were distinctly different agreed AIDS was a familiar topic to them,
readers from the first semester students; their some had not recognized the newspaper article
recall abilities stood clearly above the beginning that they had just read to be about AIDS. Some
group with all texts. If more proficient readers readers seemed to have decided early on in
are less dependent on knowledge of topic, evi- reading Sida: les legons de San Francisco that this
dence of that effect here was masked by other was an article about a place. "The history of
factors. 'Sida,'" the talk of budgets, communities, popu-
One fascinating question that remains is lations, and comparisons to New York, Paris,
whether the definition of reading difficulty may Rome, and Kinshasa were all incorporated into
be qualitatively different for different profi- a description of a place. One reader offered that

This content downloaded from 128.2.10.23 on Mon, 19 Jun 2017 00:46:16 UTC
All use subject to http://about.jstor.org/terms
JoAnn Hammadou 33
"nearly 300,000 better is more people
complex and therefore act
remains
and another more elusive. Excepting Bernhardt,
explained that all of the"
7,000 researchers background knowledge
and studies summarized in
special
and France the introduction
put together involved constructedmo passages
for a clinic in Sida." Yet another recalled that which can maximize the factors being studied
"the United States contributes 860 million rather than unedited, authentic passages. Bern-
dollars to Sida. Here, people go to have hardt's their data shows the effects of gaps in cultural
sicknesses cured. . . . In the past, homosexuals background knowledge and the resulting dis-
went to the Sida with their partners for infec- tortions in comprehension.
tions to be cured." The relative weight of the effect of more vs.
The problem experienced by the above
less prior knowledge on comprehension is a
readers underscores a problem in research delicateon issue - which is not to say that we can-
reading. The measures of familiarity may not wellaccount for varying amounts of background
have been perfectly accurate, but group results knowledge. My assessment of prior knowledge
are confused because of those readers who of the passage topics of the human eye and the
never instantiated the appropriate prior game of pitanque by means of multiple-choice
knowledge in memory. Attempts to understand tests of ten items each proved to be a signifi-
the reading comprehension process are inevi- cant factor in readers' comprehension (19).
tably hampered by the invisibility of the process However, to validate the tests and confirm their
to the outside observer. All further attempts in reliability, each had to be piloted with a large
this study to judge the effect of familiarity are group of both experts and novices for each
mitigated by the awareness that some readers topic. Each test was subjected to a discriminant
had listed as their most familiar topic one that analysis and item analysis. Needless to say,
they had never recognized while reading. such procedures are costly and time consuming
No readers scored a zero on the AIDS pas- for regular use. The simple self-report used in
sage. All understood some parts of it accu- this study does not appear, however, to be a
rately. Occasional confusion over this passage fail-safe shortcut.
simply dramatizes the fact that no one factor Question 3: Inferences and Evidence of Beginning
produces an absolute effect. One reader (who Reader Input. The inferencing data provide
also ranked AIDS as her most familiar topic) important evidence that there are differences
decided that "they were trying to find a cure between different proficiency groups that go
for an illness. It is a long term illness, but it beyond quantity of recall. The beginner readers
is not the same as AIDS." These preceding were not required--or even encouraged--to
recalls suggest the dizzying array of possible "guess." Their recalls could simply have been
interactions between individual prior knowl- briefer than those of advanced readers. How-
edge and text elements. All factors seem to have
ever, rather than being shorter and less complete
the possibility to interact in different ways and
versions of their more proficient counterparts'
with different strengths. recalls, those of the novices show clearly the
The readers of Italian were more evenly traces of the readers' own schema, those con-
divided over what topic was most familiar to cepts that the reader brought to the creation
them than readers of French. The overall of meaning of the text. For example, the follow-
favorite, sports, was most familiar only toingforty
two written recalls from subjects of the same
percent of the subjects. As a result, differences
familiarity group contain approximately the
same amount of accurate recall. The second,
between most familiar and least familiar groups
tended to be less pronounced. At what beginner,
point recall shows a greater amount of
or strength familiarity is able to increasereader
com- initiated propositions that were inferred
prehension significantly beyond that of from lesserthe original text than the first, advanced,
known topics is still not known. Earlier studies
recall. Inferences in the unedited recalls are
italicized:
have shown the effect of the construct by show-
ing that recall of topics known (recently experi-
B2. This article is about the new Stealth bomber con-
enced Halloween events) is significantly better
structed by Northrup. Its design enables it to be
than topics unknown (the history of Halloween) "invisible" to enemy radar. 130 B2s have been ordered.
for non-native readers of English. The effect The budget is 67 million US or 400 million francs. This
of various degrees of familiarity such as knowing
is the equivalent of double the budget of the French
both sports and cinema but knowing sports
army. The B2's structure was built using new materials

This content downloaded from 128.2.10.23 on Mon, 19 Jun 2017 00:46:16 UTC
All use subject to http://about.jstor.org/terms
34 The Modern Language Journal 75 (1991)
novice
including carbon fiber, graphite, andreaders leftThese
resins. such evidence
help of their per-
resist radar detection. Its exterior sonal schema within
is painted their
a black immediate recall
matte
to further it's [sic] invisibility. The bomber is powered by
protocols.
2 engines that enable it to reach 4x the speed of sound.
Inferencing Differences. The Italian subjects
recorded
The ultimate American bomber. Themorefuturistic
illogical inferences,
B2 and the
French
bomber constructed by Northrup subjects recordedinvisible
is supposedly more logical ones.
to enemy radar detection. The This
Pentagon
difference iswants
probably 130
due to B2s
the apparent
built to form an army absolutely unpersuadable by the
greater difficulty of the Italian passages as
Soviets, for a collasall [sic] 67reflected
million dollars
in lower or scores.
overall recall 400 Whereas
million francs. That is double the French army budget
the nature of these individual texts
may have
for 1989. The structure is made of a new composite of
directed subjects toward
carbon, graphite and resin. Its resistance is revolu- more or less logic in
their[sic]
tionary. It is painted camoflage style inferencing, the propensity for
to be undetectable inferencing
by not only radar but color as remained strong
well. Its among all have
2 engines beginning readers
regardless
speed 4 times the normal bomber. Until of text
the and across both
inauguration second lan-
of the bomber, the US Air Forceguage
is keeping
groups. it top secret.
The
Imagine the headlines. The photographs Italian readers'
immortalize theinferences
scenic did show a
view, the clouds!
relationship between language proficiency, self-
Surprisingly the beginners' written
reported familiarityrecalls
of topic, and logic of infer-
were often longer than those ences. of the
Unlike advanced
beginning readers, the advanced
readers, as in the case above. readers The beginners'
who reported greater familiarity with
written recalls contain frequent inferences.
the topic showed a significant decline in illogical
Both readers have left out inferences. a greatThe number of
three familiarity groups of ad-
the original propositions, but vanced readers
only the produced
beginnera nearly even number
reader makes the leap fromof logical inferences (6, "four
understanding 7, 7) while producing
times" to inferring that this fewerunusual
illogical onesbomber
with increasing familiarity
flies four times faster than other
(23, bombers,
15, 12). Advanced for
readers showed a sharp
example. decline, therefore, in illogical inferencing with
Naturally, both readers may have used iden- greater background knowledge. Beginning
tical internal cognitive processes during read- readers, on the other hand, simply inferenced
ing. One cannot determine from a recall in greater overall quantity when self-reporting
protocol that processing was actually different greater familiarity. For example, all readers
for the beginner reader. The study did not except advanced readers with greatest familiar-
record the ongoing thinking processes of either ity with the topic of the antifascist's biography
proficient or beginner readers. All readers may produced many illogical inferences from the
be assumed to have used the inferencing following segment:
process as described in earlier descriptive His father, a rich Bohemian merchant, after having
studies. Nevertheless, more of what remains as settled in Milan at the end of the last century, took as
"recall" comes from the reader's own knowledge his bride one of the Cairoli women from Mortara with

or speculation about the topic for the less pro- whom he had three children. .. .6

ficient readers.
The greater amount of inferencing during Those unfamiliar recalled that "the father
recall done by the beginner readers indicatesfinished his studies in Cairo," "the father was
the influence of prior knowledge on these in Milan at the end of the second century," "his
readers. The self-reports of prior knowledge father was stationed in Cairoli," or "his father
married Cairoli in the second century." No such
earlier did not show a relationship between L2
proficiency and prior knowledge. However, similar details were recalled by advanced profi-
closer examination of the written recalls does ciency readers who reported familiarity with the
reveal a relationship between L2 proficiency topic.
and "recall" of nontextual propositions. In other Readers in the French study showed a simi-
words, the novice readers' written recall proto- lar tendency for advanced readers' inferencing
cols show visible signs of their attempts to to increase in logic (21, 25, 26) and decrease
construct meaning from prior knowledge. The somewhat in illogic (14, 14, 11) based on self-
elements of these attempts that remain visible reported background knowledge, but this ten-
are those propositions that deviate from the dency did not reach statistical significance as
original text. Across texts and across languages, it did in the Italian study. The possibility of a

This content downloaded from 128.2.10.23 on Mon, 19 Jun 2017 00:46:16 UTC
All use subject to http://about.jstor.org/terms
JoAnn Hammadou 35
complex link cient readers are not seen in novicethe
between readers. pr
background Below knowledge, an
a certain level of linguistic knowledge,
proficiency readers may not be able
should not to infer be
in the same
p
nated. way (as evidenced by Phillips' L1 readers) or
The question still remains whether additional to the same degree (as suggested here). How-
such relationships with topic familiarity mightever, the support for this notion remains
become apparent with a more finely tunedtenuous because the French study only ap-
measure of familiarity. Meanwhile, the overallproached the same patterns without reaching
inferencing trends are a very important findingstatistical significance. Furthermore, the find-
of the study. ings both here and in Phillips are based on the
The greater inferencing done by novice relationships between number of inferences or
readers in both French and Italian for all six inferencing strategies to proficiency without
texts is clearly an indication that recall and regard to relationships to the quality of com-
prehension that actually resulted. Further
comprehension are not products of the text
alone. If that were the case, the novices' recalls
investigation of the relationships between in-
simply would have been less complete. Instead,ferencing and prior knowledge as they relate
to language proficiency may aid attempts to
the greater inferencing points to the important
role that the reader's own ideas play in com- better understand the construct of proficiency.
prehension. These inferences also support the No attempt was made to influence these
idea that greater or lesser language proficiency
readers' comprehension with varying types of
is not a simple question of quantity but rather
instruction. Therefore, no mandates for foreign
reflects qualitative differences as well. language teachers result. But language teachers
Inferences remain as external evidence of the can see the evidence of the active and inter-
interaction between reader and text. They active nature of comprehension of a second lan-
document some of the encounter between guage. As a result, language teachers can learn
reader and text. They show some of what the deal about that process from their own
a great
reader brings to the process and are part students
of the by listening carefully to what students
interaction between readers' prior knowledge
say they have understood. Although the stu-
and the text. dents' prior knowledge will clearly have an
impact on what, if not how, they comprehend,
CONCLUSIONS AND IMPLICATIONS
simple measures of prior knowledge will not
easily identify this impact. During testing this
Explanations of language proficiencywillremain
remain a problem. During instruction pre-
elusive, and its component parts defyreading
enumera-discussions or clarifications of topics
still seem
tion. Nevertheless, further careful study may important.
The studies
clarify learner characteristics as they relate to raise several additional ques-
varying L2 proficiencies in comprehension.
tions. It would be interesting to know the meta-
The preceding studies provide evidence that as
cognitive awareness of the readers, particularly
comprehension proficiency improves, those readers whose recall protocols contain
qualita-
tive differences appear rather than just
largesimply
numbers of propositions from outside the
quantitative ones. Although the self-reports of Is there any difference in aware-
original texts.
ness of process between the novice and ad-
prior knowledge were not able to demonstrate
clearly an effect for prior knowledge vanced reader? How confident would the
on com-
prehension, the effect did surface. Itnovice
became reader be of the validity of each proposi-
visible, not in the relationship between self- Would he or she be less confi-
tion in recall?
reporting of prior knowledge and quantity ofthose propositions from outside the
dent about
original
accurate recall, but rather in the vestiges of text
the than of others?
reading process that were unique to the Allreader
of the beginning and advanced readers
that remained in the written recall protocols.
from this study probably come from the cate-
The analysis of readers' inferencing gories "Novice" and "Intermediate" as described
did, indi-
in the ACTFL Guidelines. None of the ad-
rectly, demonstrate that readers' background
knowledge was affecting the comprehension
vanced readers was approaching near-native
process. The Italian study lends some fluency,
support but all were still struggling to learn the
to Phillips' L1 findings that the effects of back-
foreign language themselves. Because real dif-
ground knowledge that are visible with profi-
ferences were identified between these two

This content downloaded from 128.2.10.23 on Mon, 19 Jun 2017 00:46:16 UTC
All use subject to http://about.jstor.org/terms
36 The Modern Language Journal 75 (1991)
groups, it would be useful to extend
prehension. the study
Detailed com- of the nature of the
parison to other, more experienced process should continue in order to understand
or proficient
readers as well. better what foreign language learners actually
Regardless of how elusive the internal com-
do when engaged in the activity we call foreign
language learning.
prehension process is, language educators will
continue to attempt to influence learners' com-

ment methods for scoring written recall protocols. This


ACKNOWLEDGMENT
method was the most time efficient and provided the same
information as more cumbersome methods for the purpose
of this study.
I thank Professor Remo Trivelli, Section head of Italian
4The researcher followed the now established practice of
at the University of Rhode Island, for writing
his invaluable
recall protocols in the subjects' native language to
assistance during all phases of this project.
avoid the confounding of ability to comprehend with the
ability to produce in the target language. See Bernhardt
(6; 7; 8) for further examples of the use of the procedure
NOTES and see Lee for a discussion of the implications of choice
of language for written recalls.
5The analysis of each passage was conducted separately
'Inference is sometimes divided into two subclasses of rather than in combination with the other two passages of
the same
implication and presupposition. See Carrell for examples and language for the following reasons: 1) Comparing
discussions of these categories of inference, which them within the same analysis of variance would require
derive
making claims about their equality on features other than
from sentences containing implicative and factive predi-
cates. This study does not differentiate between thetopic or expanding the qualitative segment of the study
sub-
classes of inference. beyond its original intent or scope. Such uniformity was
not possible with naturally occurring texts, which were
2All current measurement techniques share the common
deemed necessary for the validity of the study. 2) The result-
weakness of being indirect measures of the internal processes
of comprehension and inference. Recall as an operationaling link between simultaneous repeated measures of passage
topic and topic familiarity would violate the assumptions
definition of comprehension has gained widespread accept-
of the model of the analysis of variance. Each subject would
ance in current empirical research for two broad reasons:
1) it avoids the contamination of text comprehensionhave
by had three repeated measures of familiarity and three
repeated measures of passage topic, but each measure could
further discourse in the form of questions; 2) it closely
matches current models of comprehension that emphasize only be examined in relationship with one of the other
the central role of the individual reader (e.g., 3; 4; 6;measures.
8; This linking of two otherwise separate repeated
measures is problematic for the ANOVA model.
14). All current measurement techniques share the common
6"Il padre, ricco commerciante boema, stabilitosi a Milano sul
weakness of being indirect measures of the internal processes
finire del secolo scorso prese in sposa una Cairoli, proveniente da
of comprehension and inference.
Mortara, dalla quale ebbe trefigli. .
3See Bernhardt (8). The author compares several assess-

6. Bernhardt, Elizabeth B. "Testing Foreign Language


BIBLIOGRAPHY Reading Comprehension: The Immediate Recall
Protocol." Unterrichtspraxis 16 (1983): 27-33.
7. - . "Reconstructions of Literary Texts by Learners
1. Afflerbach, Peter P. "Main Idea Construction Strategies."
of German." New Yorker Werkstattgesprdch 1984:
Reading Research Quarterly 25 (1990): 31-46. Literarische Texte im Fremdsprachenuntericht. Ed. Man-
2. American Council on the Teaching of Foreign
fred Lan-
Heid. Miinchen: Kemmler & Hoch, 1985.
guages. ACTFL Provisional Proficiency 8.Guidelines.
- . Merging Research and Assessment Perspectives in For-
Hastings-on-Hudson, NY: ACTFL, 1986. eign Language Reading. Technical Report. Washing-
3. Anderson, Richard C. & P. David Pearson. "A Schema-
ton: National Foreign Language Center, 1988.
9. Bransford,
theoretic View of Basic Processes in Reading Com-John D. & M. K. Johnson. "Contextual Pre-
prehension." Handbook of Reading Research. requisites
Ed. P. for Understanding: Some Investigations
David Pearson. New York: Longman, 1984: of Comprehension and Recall." Journal of Verbal
255-92. Learning and Verbal Behavior 11 (1972): 717-26.
10.L.
4. Anderson, Richard C., Ralph E. Reynonds, Diane Canale, Michael. "On Some Dimensions of Language
Schallert & Ernest T. Goetz. "Frameworks for Com- Proficiency." Issues in Language Testing Research. Ed.
prehending Discourse." American Educational Research John W. Oller, Jr. Rowley, MA: Newbury House,
Journal 14 (1984): 367-82. 1983: 333-42.

5. Bartlett, Frederic C. Remembering. London: Cambridge 11. - & Merrill Swain. "Theoretical Bases of Com-
Univ. Press, 1932. municative Approaches to Second Language Teach-

This content downloaded from 128.2.10.23 on Mon, 19 Jun 2017 00:46:16 UTC
All use subject to http://about.jstor.org/terms
JoAnn Hammadou 37
ing Testing." Appliedon Building Background Knowledge."
Linguistic TESOL Quar-
12. Carrell, Patricia terly 16 (1982):L.503-16. "Inferenci
tions and Implications
22. Kintsch, Walter. The Representation of Factiv
of Meaning in Memory.
Predicates." Language Hillsdale, NJ: Erlbaum, Learning
1974. 3
13. Collins, Allan,23. Lee, James
John F. "Background Seely
Knowledge in L2 Read- Brow
"Inference in Text ing." Modern Language Journal 70 (1986): 350-54.
Understanding.
in Reading 24. Pearson, P. David. "The Effects of Grammatical
Comprehension: PerspectCom-
Psychology, plexity on Children's
Linguistics, Comprehension, Recall, and
Artificial Inte
tion. Ed. Rand J. Spiro, Bertram C. Bruce & Conception of Certain Semantic Relations." Reading
William F. Brewer. Hillsdale, NJ: Erlbaum, 1980: Research Quarterly 10 (1974-75): 155-92.
385-407. 25. Phillips, Linda M. Inference Strategies in Reading Compre-
14. Connor, Ulla. "Recall of Text: Differences between Firsthension. Center for the Study of Reading Technical
and Second Language Readers." TESOL Quarterly Report No. 410. Champaign, IL: Univ. of Illinois,
(1984): 239-55. 1990.

15. Coste, Daniel J., et al. Un niveau seuil. Strasbourg:


26. Samuels, S. Jay & Michael L. Kamil. "Models of the
Council of Europe, 1976. Reading Process." Handbook of Reading Research. Ed.
16. Cummins, Jim. "Language Proficiency and Academic P. David Pearson. New York: Longman, 1984:
Achievement." Issues in Language Testing Research. Ed. 185-225.

John W. Oller, Jr. Rowley, MA: Newbury House,27. Spiro, RandJ. "Remembering Information from Text:
1983: 108-30. Theoretical and Empirical Issues Concerning the
17. Frederiksen, Carl H. "Representing Logical and Seman- 'State of Schema' Reconstruction Hypothesis."
tic Structure of Knowledge Acquired from Dis- Schooling and the Acquisition of Knowledge. Ed. Richard
course." Cognitive Psychology (1975): 371-458. C. Anderson, RandJ. Spiro & William E. Monta-
18. Goodman, Kenneth S. "Psycholinguistic Universals in gue. Hillsdale, NJ: Erlbaum, 1977: 137-66.
28. - . "Constructive Processes in Prose Comprehen-
the Reading Process." Psycholinguistics and Reading.
Ed. Frank Smith. New York: Holt, Rinehart, 1973: sion and Recall." Theoretical Issues in Reading Com-
21-27. prehension: Perspectives from Cognitive Psychology, Linguis-
19. Hammadou, JoAnn. "The Effect of Analogy on For- tics, Artificial Intelligence, and Education. Ed. Rand J.
eign Language Reading Comprehension of Spiro, Bertram C. Bruce & William F. Brewer.
Advanced and Novice Readers of French." Diss., Hillsdale, NJ: Erlbaum, 1980: 245-78.
Ohio State Univ., 1988. 29. Steffensen, Margaret S., Chitra Joag-Dev & Richard
20. Hudson, Thom. "The Effects of Induced Schemata on C. Anderson. "A Cross-Cultural Perspective and
the 'Short Circuit' in L2 Reading: Non-Decoding Reading Comprehension." Reading Research Quarterly
Factors in L2 Reading Performance." Language 15 (1979): 10-29.
Learning 32 (1982): 18-23. 30. Van Dijk, Teun A. & Walter Kintsch. Strategies of Dis-
21. Johnson, Patricia. "Effects of Reading Comprehension course Comprehension. New York: Academic, 1983.

APPENDIX I

Two Sample Passages

Sida: les legons de San Francisco Le Sida, a l'image de son incubation, est une maladie
d'evolution tres lente, aussi nul ne doute desormais, compte
Le Sida mobilise pres de 7 000 chercheurs et specialistes,
tenu des resultats en cours, des divers traitements prescrits,
les Etats-Unis lui consacrent en 1989 un budget que lesde 860
praticiens gu&riront la maladie avant de proteger
millions de dollars, soit 36 fois le budget que la
les France lui saines par une vaccination efficace.
populations
L'histoire
octroie pour la meme annee (150 millions de francs). Certes,du Sida n'est pas la meme 'a San Francisco,
sommes-nous cinq fois moins nombreux et dix-neuf foismaintenant jugulee, a frappe aveuglement
o0 'lpidemie,
moins malades (5 174 cas recenses en decembre 1988 dans homosexuelle des 1978, a New York, a Paris
la communaute
notre pays contre 93 723 outre-Atlantique). ou a Rome, oi se repand a travers les toxicomanes et leurs
On estime a pres de 300 000 le nombre departenaires
personnes sexuels, ou encore 'a Kinshasa, oii il progresse
qui sont actuellement atteintes du Sida dans le monde
dans et
la population h6terosexuelle, donc dans la population
qui attendent, au-delh des traitements palliatifs tout
qui court.
leur sont
proposes, une veritable cure de leur maladie. L'histoire du Sida 'a San Francisco meritait une attention
Plusieurs centaines d'equipes de praticiensparticuliere,
dirigent en ce que, pour cruelle qu'elle soit, ses
plusieurs dizaines d'essais cliniques dans lesquels plusieurs sont une contribution essentielle a la
enseignements
milliers de volontaires testent pour eux-memes et leurs
recherche medicale.
semblables l'efficacite et la suirete des m dicaments qui leur
sont administres.

This content downloaded from 128.2.10.23 on Mon, 19 Jun 2017 00:46:16 UTC
All use subject to http://about.jstor.org/terms
38 The Modern Language Journal 75 (1991)

APPENDIX I (continued)

Riccardo Bauer antifascista e educatore

Riccardo Bauer, milanese di nascita e di adozione tenente di artiglieria da montagna, partecip6 alla guerra
guadagnandosi una medaglia di bronzo al val militare e
(Milano 1896-1982) e una delle figure piui rappresentative
dell antifascismo italiano militante. una citazione all'ordine del giorno della IV Armata per le
Il padre, ricco commerciante boemo, stabilitosi a Milano
numerose ferite reportate e per il coraggio dimostrato. Muti-
sul finire del secolo scorso prese in sposa una Cairoli,
lato, ottenne une pensione di guerra per aver perso l'udito
da un orecchio perforato da una scheggia di granata e per
proveniente da Mortara, dalla quale ebbe tre figli: Augusto,
ingegnere affermatosi nel settore dei trasporti, Riccardo, la ritenzione di un proiettile sotto la scapola destra.
laureatosi alla Bocconi in Scienze economiche, e Adele. Ritornato dalla guerra complet6 gli studi che aveva
I due fratelli rinunciarono alla cittadinanza austriaca allo interrotto, coerente con la sua richiesta di chiusura delle
scoppio della Prima guerra mondiale e si arruolarono volon-
universitt durante il conflitto.
tari senza chiedere il cambio del cognome. Riccardo,

APPENDIX II

Analyses of Covariance: French Study

AIDS Sudan Stealth Bomber

Source df MS F df MS F df MS F

Proficiency (A) 1 183.19 26.51* 1 1247.41 12.43*** 1 387.37 4.70**


Familiarity (B) 2 2.16 .03 2 51.82 .52 2 154.10 1.87
GIC 1 340.89 4.93** 1 179.06 1.78 1 139.84 1.70
AxB 2 32.59 .47 2 28.30 .28 2 69.79 .85
Error 82 69.12 82 100.33 82

*p>.0001; **p>.03; ***p>.0007.

APPENDIX III

Analyses of Covariance: Italian Study

Cinema Sports Biography


Source df MS F df MS F df MS F

Proficiency (A) 1 1450.85 64.8* 1 558.22 54.88* 1 1029.05 18.66*


Familiarity (B) 2 67.45 3.01 2 .63 .06 2 97.78 1.77
GIC 1 66.48 2.97 1 129.69 12.75** 1 742.25 13.46**
AxB 2 .20 .01 2 25.61 1.26 2 151.40 2.75
Error 76 76 76

*p>.0001; **p>.0006.

This content downloaded from 128.2.10.23 on Mon, 19 Jun 2017 00:46:16 UTC
All use subject to http://about.jstor.org/terms

You might also like