Language Learning
E1.v
Strength of Motivation
An additional variable of major concern in this study is level of
motivation. This construct, termed Strength of Motivation, concerns the
degree to which an individual desires to learn the second language. A
number of studies have investigated motivational level; research that has
separated level of motivation from type of motivation (e.g., instrumental
or integrative motivation) has generally found higher levels of
motivation associated with grades (but not with scores on standardized
Other approaches to assessing anxiety in the foreign language classroom are reported by
Kleinmann (1977) and McCoy (1976).
6Anxiety, however, was predicted by prior achievement (Gardner, Lalonde. and Pierson
1982).
Language Laarning
'
'The results of a descriptive study of motivational type were included in the analysis. Space
does not permit a discussion here of this aspect of the research (see Ely in press).
'In the initial planningfor this study, a distinction was made between quantity (number, not
length o r complexity, of utterances) and quality (correctness) of Classroom Participation. It
was hypothesized that Language Class Risktaking and Language Class Sociability would be
linearly (and positively) related to quantity of Classroom Participation. but curvilinearly
related to quality (correctness) of Classroom Participation, with moderate amounts of
Language Class Risktaking and Language Class Sociability most promotive of correctness in
class. Subsequently, it was decided to simplify the model by focusing only upon quantity of
Classroom Participation.
The hypothesis that risktaking is curvilinearly related to accuracy may be especially
attractive in cases where the individual has an opportunity to use the language somewhat
freely, as in a naturalistic environment or at relatively advanced stages of classroom language
learning. For a further discussion of curvilinearity in the association between risktaking and
L2 acquisition, see Beebe (1983).
Language Learning
Vol. 36, N o . 1
Proficiency
9 t r m vt h
\//*
Language
Class
Discomfort
Class
Language
Class
Sociability
E1.v
have found students attitudes toward the teacher and the course to be
predictive of motivation. In the present study, Attitude toward the
Language Class was hypothesized t o be a positive predictor of Classroom
Participation.
It was theorized that students who have a strong desire to achieve high
grades (high Concern for Grade) manifest high Classroom Participation.
Concern for Grade was also seen as a positive predictor of proficiency.
Another positive contributor to Classroom Participation, as well as to
proficiency, may be the students Language Learning Aptitude. Gardner
and Lamberts (1972) extensive factor-analytic study of U.S. high school
learners of French yielded numerous significant common factor loadings
of aptitude and proficiency. Lalonde (1982) found aptitude to be a
predictor of French achievement among Canadian university students.
In addition to the main constructs (shown in Figure l), two variables
representing prior learning of Spanish were introduced as control
variables. These were: amount of previous classroom study of Spanish and
presence or absence of a Spanish-speaking home environment. These
variables were included in the prediction of Classroom Participation and
proficiency.
METHOD
Setting of the Study
The subjects were students enrolled in first year Spanish courses at a
university in northern California. The study involved six classes: three
classes of students in their first quarter of Spanish study (Level 1) and three
classes of second quarter students (Level 2). The students, who gave their
written consent to participate in the study, were told only that the study
concerned various aspects of language learning. Similarly, the six
teachers were not advised of the specific objectives of the research; they
were aware, however, that it was the students, rather than the teachers, who
were the focus of the study. The data were gathered during the winter
quarter of the 1982-83 academic year.
Language Learning
10
Language Learning
I.
2.
3.
4.
5.
Id like more class activities where the students use Spanish to get to
know each other better.
I think learning Spanish in a group is more fun than if I had my own
tutor.
I enjoy talking with the teacher and other students in Spanish.
I dont really enjoy interacting with the other students in the Spanish
class. (-)
I think its important to have a strong group spirit in the language
classroom.
Strength of Motivation
A nine-item Strength of Motivation scale was designed for the pilot
questionnaire. After item analysis of the results, two items with low
corrected item-total scale correlations were deleted; the seven items used in
the final questionnaire are listed below. (Items 1, 2, and 3 were patterned
after items on the scales of Motivational Intensity and Desire to Learn
French [see, Gardner, Cltment, Smythe, and Smythe 1979; Gardner and
Smythe 1981; Gliksman, Gardner, and Smythe 1982].)14
1nitially, it had been intended to employ the Motivational intensity and Desire to Learn
French scales. However, a close examination of these scales led to the decision to use an
instrument designed expressly for the specific target population. See Ely ( 1984:28-32) for
further discussion. (In comparing this scale with the pilot questionnaire scale, note that items
3 and 5 of the original scale were those deleted [cf. Ely 1984:31].)
E1.v
11
12
Language k a r n i n g
Classroom Participation
The classroom behaviors hypothesized to predict proficiency were those
representing students self-initiated utterances in Spanish. As discussed
earlier, a number of studies have examined voluntary participation in the
language class (Day 1984; Gliksman, Gardner, and Smythe 1982; Naiman,
Frohlich, Stern, and Todesco 1978; Seliger 1977). In the present
14
Language Learning
of the utterances noted by B; coder B had done the same for 92 percent of
the utterances noted by A. A reliability check was made of the specific
codings: of the utterances coded as volunteer or general solicit by the other
coder, coders A and B agreed with 99 and 97 percent of the codings,
respectively.
Proficiency
Three proficiency measures were employed: Oral Fluency; Oral
Correctness; and Written Correctness. Oral Fluency and Oral Correctness
were assessed by means of a story-retelling task (see Oller 1979). The
stories were developed by the investigator, in consultation with the teachers
participating in the research. Lists were compiled of all grammatical
structures and lexical items in the textbook chapters studied during the
quarter. From these lists, two stories were written, one for Levei I (first
quarter students) and one for Level 2 (second quarter students). The
content of the two stories was similar, relating a series of events common
among college students. The stories were recorded by a female native
speaker of Spanish. These stories were relatively short (103 words for Level
1; 122 words for Level 2); however, to help relieve students possible
concern about their ability to remember the story, a handout was provided
with picture frames showing the sequence of events.
Each class performed the task in the university language laboratory
during class time. Students sat in alternate booths and listened to and
retold the story using individual tape recorders. Students were not allowed
to take notes and were told that memorization of the story was not
necessary. The stories were presented twice, following four conversational
questions (designed to help the students become accustomed to the
speakers voice).
The tape recordings were transcribed by a native speaker of Spanish.
These transcripts were the basis for determining an individuals Oral
Fluency and Oral Correctness. A high Oral Fluency score reflected a
relative absence of self-interrupted elements. Interrupted elements fell into
two categories: fragments and disfluent words. Fragments were phonemes
This is probably a conservative estimate of the overall reliability of this process, since the
coders commented that the interaction was unusually fast-paced during the final class period.
These stories, as well as the examinations used for Written Correctness, are not
reproduced here because of the possibility of their future use by the university.
or syllables which did not constitute words. Disfluent words were words
(either isolated or in phrases) which did not convey a complete thought, but
which were interrupted when the speaker began to frame the same idea in
an alternative manner or began to express a new idea.
The transcripts were marked for Oral Fluency by the investigator.
Reliability was measured by comparing practice transcripts marked
separately by the investigator and a bilingual Spanish speaker: there was an
average of 98 percent agreement (97 and 99 percent) with the other person
regarding utterances marked as disfluent. A students Oral Fluency score
was calculated by adding the number of fragments and disfluent words,
and then dividing by the number of fluent words.
Oral Correctness was assessed by marking errors in morphology, syntax,
and lexical choice in the transcripts. In order to prevent Oral Correctness
from being confounded with Oral Fluency, the transcripts were first
marked for fluency; only non-disfluent elements were marked for
correctness. Oral Correctness was judged by two individuals who had
recently taught first year courses in the Spanish program at the university.
The judges had two tasks: first, to mark the transcripts, indicating and
correcting mistakes; and second, to assign a score (a count of the number of
pertinent errors). Errors were ignored if they did not represent language
elements covered in the textbook chapters (see Ely 1984 for scoring
guidelines). A students Oral Correctness score consisted of an average of
the judges error counts divided by the number of fluent words. The degree
of agreement between the judges was assessed by means of a Pearson
correlation coefficient calculated for all of the scores assigned: the
correlation was .98.
The third proficiency score, Written Correctness, was the numerical
grade on the regular final written examination. The examinations focused
on grammar and included opportunities for using language in the context
of connected discourse. These final examinations covered the same
textbook chapters used in the story-retelling task. Each level had one
examination, which was prepared by a group of teachers. The exams were
graded by the teachers; the grading was checked by the investigator and,
where necessary, the number of points taken off for a particular type of
error was equated across exams (see Ely 1984). A modified form of splithalf reliability was calculated (with totals for odd and even numbered
sections constituting the two halves) and stepped up using the SpearmanBrown prophecy formula. The estimates of reliability for the Level 1 and 2
examinations were .94 and .91, respectively.
16
Language Learning
FINDINGS
Preliminary Analyses
After the final questionnaire was administered, two items were found to
have low corrected item-total correlations, Risktaking item 4 and
Sociability item 5." These two items were omitted from the scales in
subsequent analyses.
The scales were subjected to a reliability assessment before use in the
causal analysis. The Cronbach alphas for the scales were: Strength of
Motivation, .86; Language Class Discomfort, .79; Language Class
Risktaking, .65; Language Class Sociability, .66; and Attitude toward the
Language Class, .86.2032'
For all the variables in the model, the data were examined for significant
differences in the means. The one significant difference among the means
for the six classes was for classroom participation. (As discussed below,
differences attributable to class were controlled for in the prediction of
Classroom Participation.)
An additional analysis was carried out in order to investigate the factors
which may underlie the variables of Language Class Discomfort, Language
Class Risktaking and Language Class Sociability. All the items on the three
scales were subjected to a single factor analysis. The items were found to
load on three factors (see Ely, 1984 for a more detailed presentation and
discussion).
I9A reexamination of the results for the pilot questionnaire showed these two items to have
had the lowest corrected item-total correlations for their respective scales (as detailed in Ely
1984).
*?he levels of reliability for Risktaking and Sociability were lower than expected; it may be
noted, however, that they compare favorably with reliability coefficients for certain other
personality, motivation, and cognitive style scales used in research with L2 learners. See, for
example, Gardner and Smythe ( 198 I ) and Lalonde (1982). as well as Budner (1962, used in
Naiman, Frohlich, Stern, and Todesco 1978). In general, the reliability figures for the final
questionnaire scales were lower than those of the pilot study; this may have been due to the
addition of foil items, the shortening of the scales, differences in the samples, etc.
"In a subsequent study with a larger sample of first year university Spanish students (125
subjects), a considerably higher Cronbach alpha of .74 was obtained for Language Class
Risktaking. A report of that study, which had a different overall purpose from that of the
present research, is in progress.
17
"In the study mentioned in the previous note, the corresponding R 2 figures were .37 for
Language Class Risktaking and . I I for Language Class Sociability.
18
Language Learning
"As discussed in Ely (1984), the dependent measure represented a log transformation of
Classroom Participation.
19
Table I
Summary of srepwise regression of Classroom Parriciparion
Variables
in equation
beta
Risktaking
Class*
.39
0.40
13.92
2.91
<.01
.02
beta to
enter
F to
enter
-0.19
-0.1 I
0.07
0.07
0.06
0.06
-0.05
-0.02
2.66
1.10
0.50
0.50
0.34
0.34
0.20
0.04
.lI
.30
.48
.48
.56
.56
Variables not
in equation
Discomfort
Yrs. Hi. Sp.
Att. Lang. C.
Aptitude
L a n g u v e Back.
Str. of Motiv.
Concern Grade
Sociabilitv
-.21
-_
14
.I4
.01
.02
.I9
-.04
.ox
.65
.83
N = 75
5 Contrasts: d.f. 5. 68
R = 30
not predictors; there was, however, a significant effect for Class. The
equation explained 30 percent of the variance in students voluntary
participation in the Spanish classroom.
Antecedents of Proficiency
For each of the three proficiency measures, a separate regression analysis
for Level 1 and Level 2 was carried out. The results for the story-retelling
task (Oral Correctness and Oral Fluency) represent 63 of the 65 students
(two outliers had been omitted) who were present on the day of
administration: 27 at Level 1 and 36 at Level 2. The Written Correctness
results represent a total of 72 students: 32 at Level 12 and 40 at Level 2 (see
below for comments on group sizes).
The beta coefficients for significant predictors in all six proficiency
regressions (three proficiency measures at two levels) are presented in
Table 2 (and, together with tests for level comparability, discussed in Ely
1984). Oral Correctness was predicted by Classroom Participation (Level
I), Concern for Grade (Level 2), and Language Learning Aptitude (Level
2). Oral Fluency was predicted by Language Background (Level 1) and the
number of Years of High School Spanish study (Level 2). Written
Correctness was predicted by Strength of Motivation (Level 2) and
20
Language Learning
Table 2
Standardized regression coefficients of significant predictors ofproficiencv
Level I
0r.C.
Classrm. Partic.
Aptitude
Str. of Motiv.
Concern Grade
Yrs. Hi. Span.
Language Back.
Or.F1.
Level 2
Wr.C.
0r.C.
.31
.31
0r.FI.
Wr.C
.40
.4 1
.39
.40
.38
.45
R2
.I6
(probability level to enter set at .05)
.20
.I4
.25
.I5
.30
DISCUSSION
As hypothesized, Language Class Risktaking was found to be a positive
predictor of students voluntary Classroom Participation. Language Class
Discomfort influenced Classroom Participation only indirectly, through
its negative effect on Language Class Risktaking. The apparent lack of a
significant effect of Language Class Sociability on Classroom
21
Table 3
Correlations of predictors with criterion measures
~
Level I
Classrm. Partic.
Aptitude
Str. of Motiv.
Concern Grade
Yrs. Hi. Span.
Language Back.
Level 2
0r.C.
Or.Fl.
Wr.C.
0r.C.
0r.Fl.
Wr.C.
.40
.I4
.I6
.00
-. I I
.I0
.09
.I6
.12
.08
.22
.45
.I8
.02
.29
.I2
.39
.25
.05
-. 10
.08
.39
.36
.I9
-.05
-.I I
.37
.07
-.03
-. 18
-.08
.I4
.I0
.26
.38
-.oo
Participation could have been due to the nature of language use in the first
year classes. It may be that Language Class Sociability has greater
importance at subsequent levels of language study, when increased
linguistic skills permit a higher degree of interaction.
The fact that Oral Correctness, but not Written Correctness, was
influenced by Classroom Participation may be attributable to differences
in the cognitive skills involved in the two tasks. Thus, performance on the
written examinations may have benefited from the use of L2 skills and
learning strategies not specifically practiced in voluntary oral
participation. Alternatively, the findings may indicate that the real-time
oral test reflected the effect of (relatively) unmonitored (Dulay, Burt, and
Krashen 1982) or automatic (McLaughlin, Rossman, and McLeod 1983)
language performance developed through classroom interaction.
In the case of Oral Fluency, the influence of former language learning
experiences appeared to outweigh the other hypothesized predictors.
However, perhaps the relative effect of such experiences would be
attenuated at later stages of L2 learning, with on-going classroom
interaction becoming increasingly important.
The results of the study should be interpreted in the light of several
potential limitations. First, the separate level regressions predicting
proficiency were based on sample sizes smaller than are normally desirable.
This served to increase the possibility of Type 11errors; that is, effects that a
larger sample would have revealed were not detected. The magnitude of
causal relationships in the population was thus probably underestimated
by the results of these proficiency regressions.
In addition, a number of alternative models incorporating the variables
used in this study could be proposed. However, due to the lack of
additional independent data for cross-validation of the empirically derived
22
Language Learning
results, it was necessary to limit the investigation only to the model which
had been specified a priori on the basis of theory and previous research. 24
Third, while path analysis procedure is designed to test causal relations
among variables, a cross-sectional, correlationally-based study is not
capable of definitively proving causality. For this, it would be necessary to
conduct research following an experimental paradigm or using repeated
measurements obtained over time. 2 5
It is felt that situation-specific scales such as those employed in this
research may offer a viable alternative to the use of global psychological
measures. However, it should be noted that the instruments developed for
this study are regarded as exploratory in nature; further research is planned
which would continue to refine the operationalization of these constructs.
Finally, in evaluating the present findings, concerns regarding the study
of affective states through self-report (Oller 1981, 1982; Oller and Perkins
1978a, 1978b) should be taken into consideration, as should responses to
these concerns (Gardner 1980; Gardner and Gliksman 1982; Mischel
1981a; Upshur, Acton, Arthur, and Guiora 1978).
The population addressed in this study was that of university foreign
language students in the United States. Research should be carried out to
assess the applicability of the findings to other groups of second language
learners. There may be differences, for instance, between university and
secondary school students. In certain secondary school settings, peer
approval may be extremely important, resulting in levels and effects of
Language Class Discomfort, Language Class Risktaking, and Language
Class Sociability that differ from those among college students. In addition
to extending the present research to different age groups, it will be of
interest to conduct similar studies in bilingual, ESL, and EFL settings.
Classroom teachers may find in these results an indication of the
importance of Language Class Risktaking and Classroom Participation in
various language learning situations. However, the fact that a negative
causal relationship was found between Language Class Discomfort and
Language Class Risktaking suggests that simply exhorting students to take
24
In addition, a number of variables not included in the present study could be explored in
future research. For example, it may be useful to incorporate test anxiety (Madsen 1982) in
the model.
25
It should also be noted that path analysis models rest upon a number of assumptions
regarding the relationship of the variables in the population (Gallini 1983; see Ely
198498-99).
more risks and participate more may not be effective. Apparently, before
some students can be expected to take linguistic risks, they must be made to
feel more psychologically comfortable and safe in their learning
environment. To this end, classroom teachers may wish to devise and test
the relative effectiveness of various strategies for lessening Language Class
Discomfort. As students come to feel more secure, they can then be
encouraged to assume a more active role in the classroom.
REFERENCES
Beebe, L.M. 1983. Risk-taking and the language learner. In Classroom Oriented Research in
Second Language Acquisition, eds. H.W. Seliger and M.H. Long, pp. 39-65. Rowley,
Mass.: Newbury House.
Bem, D.J., and A. Allen. 1974. On predicting some of the people some of the time: the search
for cross-situational consistencies in behavior. Psychological Review 8 I :506-520.
Bem, D.J., and D.C. Funder. 1978. Predicting more of the people more of the time: assessing
the personality of situations. Ps.vchological Review 85:485-501.
Brown, H.D. 1980. Principles of Language Learning and Teaching. Englewood Cliffs, N.J.:
Prentice-Hall.
Budner, S. 1962. Intolerance of ambiguity as a personality variable. Journal of Personality
30:29-50.
24
Language Learning
Mischel, W. 198 Ib. Personality and cognition: something borrowed, something new? In
Personality. Cognition and Social Interaction. eds. N. Cantor and J. Kihlstrom, pp. 3-19.
Hillsdale, N.J.: Erlbaum.
Mischel, W., and P.K. Peake. 1982. Beyond deja vu in the search for cross-situational
consistency. Psychological Review 89:730-755.
Morris, L. W. 1979. Extraversion and Introversion: An Interactional Perspective. New York:
Halsted Press.
Naiman, N., M. Frohlich, H.H. Stern, and A. Todesco. 1978. 7he Good Language Learner.
Research in Education Series No. 7. Toronto: The Ontario Institute for Studies in
Education.
Oller, J.W. Jr. 1979. language Tests at School. London: Longman.
Oller, J.W. Jr. 1981. Research o n the measurement of affective variables: some remaining
questions. In New Dimensions in Second Language Acquisition Research, ed. R. W.
Andersen, pp. 14-27, Rowley, Mass.: Newbury House.
Oller, J.W. Jr. 1982. Gardner on affect: a reply to Gardner. hnguage Learning 32:183-189.
Oller, J.W. Jr., and K. Perkins. 1978a. A further comment on language proficiency as a source
of variance in certain affective measures. Language Learning 2k417-423.
Oller, J.W. Jr., and K. Perkins. l978b. Intelligence and language proficiency as sources of
variance in self-reported affective variables. Language laarning 28:85-97.
Rossier, R.E. 1975. Extraversion-introversion as a Significant Variable in the Learning of
English as a Second Language. Ph.D. Dissertation, University of Southern California.
Seliger, H.W. 1977. Does practice make perfect?: a study of interaction patterns and L2
competence. Language Learning 27:263-278.
Upshur, J.A., W. Acton, B. Arthur, and A.Z. Guiora. 1978. Causation or correlation: a reply
t o Oller and Perkins. Language Learning 28:99-104.