Human Factors in a Computer Assisted Foreign Language Environment: The Effects of Gender, Personality and Keyboard Control

Lydie E. Meunier,

The University of Tulsa

Abstract:
This multi-method research examines the effects of gender, personality and keyboard control on foreign language learning and interaction patterns of 60 intermediate French students in a computer assisted language learning environment at the university level. This study includes ten female dyads (pairs), ten male dyads, and ten mixed-gender dyads. Results of the Analyses of Variance reveal that learning achievement and interaction patterns at the computer are more strongly related to personality differences and keyboard control than to gender differences, although males and females still perform differently, yet not quite as expected. The most detrimental factor in this computer based French activity was shown to be the inability for linear learners, whether males or females, to adjust to the nonlinear format of foreign language hypertexts.

KEYWORDS

Computer assisted language instruction, gender differences, personality differences, keyboard control, verbal interaction, MBTI, linear and nonlinear learners, hidden agenda, socialization process.

GENDER DIFFERENCES AND COMPUTER USE

The Socialization Process

Several studies investigating gender differences in the use of computers revealed that males tend to be more interested in computers than females and that males use computers more than females (Collis 1985a; Collis 1985b; Fetler 1985; Fisher 1984; Adam and Bruce 1993; Murray 1993). Other studies indicate that a preference for computer

47

use, or lack of it, stems from socialization which takes place outside schools (Yelloushan 1989; Henwood 1993; Kirk 1992). For instance, parents are more likely to buy a computer and video games for their sons than for their daughters (Levin and Gordon 1989). Several studies also note that sex differences in computer use are engendered by the media which tend to advertise computer use essentially as a male activity (Forsyth and Lancy 1989; Jones 1987; Sanders 1985; DiMona and Herndon 1994); these studies found that companies selling video games essentially target a male audience by designing aggressive and competitive games that often deal with war, killing, car races, medieval-style combats, etc. According to DiMona, et al. (1994), boys are favored in commercials because of studies conducted in masculine psychology and summarized as follows: When boys associate a product with the presence of girls, they lose their interest in buying this product because they think “it is for girls” therefore a degrading product (p. 489). Likewise, the advertising for both hardware and software in computer magazines -meant for an adult audience- essentially depicts males actively involved in problem-solving situations, while females (when present) are confined to secondary roles such as typists (Forsyth and Lancy 1989; Jones 1987; Sanders 1985; DiMona, et al. 1994). The conclusions that can be drawn from these several studies are these: Western societies display a cultural bias in favor of males as users of the computer, while societal reinforcement for female interest in computers is lacking.

Hidden Curricula in Schools

Considering these detrimental effects of socialization on girls before they enter school, a logical question is whether teachers provide an equal opportunity for both genders to use computers. Unfortunately, research suggests that the bias is evident in the classroom too. Educators exhibit a strong hidden curriculum with tacit values that favor and encourage males in their computer expertise (Culley 1988; Fennema 1987; Levin and Gordon 1989; Smithson 1990; Stable 1990; Watson 199 1). Because boys come to school more computer literate than girls, in mixed-gender schools the boys have a skill advantage, and their more competitive attitudes have a negative effect on girls in computer laboratories (Stable 1990; Levin and Gordon 1989; Smithson 1990). Classroom observations of computer assisted lessons in mixed-gender schools have revealed that educators make very little effort to counteract the tendency of boys to elbow out girls on theirway to get to the newest computer equipment (Culley 1988). Siann and Macleod (1986) observed that in computer based group activities, females showed lower results when working with males even though the same females had no disadvantage in similar tasks when working individually or in same-gender groups. Observations of interactions in mixed-gender groups show that males compete for control of the computer keyboard and thus hinder females during the learning process: A positive

48

correlation was found between time spent on the keyboard and the results on achievement tests Palton 1990; Siann, et al. 1986; Underwood and McCaffrey 1990; Webb 1985).

However, both Webb (1985) and Azmitia (1987) have reported that control of the keyboard is not necessary for understanding and that learning can also take place by watching other people. Other studies indicated that when educators give equal attention to males and females, no gender differences are reported (Feldmann, Fish, Friend and Bastone 1991; Forsyth, et al. 1989). Feldmann, et al. (1991) observed that gender differences are important in some social contexts but not in others, and the social climate is different when either males or females comprise a large majority in a class. Feldmann, et al. also suggest that humanities courses using computer based tasks might be considered less biased than traditionally male-oriented courses, which implies that academic disciplines may also affect classroom communication and gender interaction.

Gender Differences in Computer Based Humanities Courses

During a computer based English course, Cummings (1985, p. 157) made a negative judgment of female students who were perceived as talkative and off-task because they had less need of silence than male students to develop their inner thoughts while working on English syntax. Males were described as thoughtful, analytical and on-task because they were more silent than females. Interestingly, in this study, Cummings considered males better than females on the premise that their silence evinced the “appropriate analytical skill” required for computer tasks.

In a study conducted with ESL students (Abraham and Liou 1991), differences in gender-specific behavior were incidentally noticed and interpreted as follows: “In the mixed pairs, the female assumed the role of typist, perhaps encouraging the male to dominate the discussion and decision making” (Abraham, et al. 1991, p. 93). This study suggests that some computer based activities in mixed-gender groups may actually trigger socially acquired behaviors.

Underwood, et al. (1990) examined interaction styles in mixed- and single-gender pairs during a computer based English course. Their study was designed to measure the interactional effects on learning taking place in the zone of proximal development during computer based communicative activities. Students had three sessions: (1) an individual CAI session, (2) a cooperative CAI session, and (3) another individual CAI session. The effect of cooperative work was measured by comparing the results of session 3 with those of session 1. The results indicated that both types of single-gender pairs improved individual performance in session 3, but mixed-gender pairs did not show any improvement of individual abilities. The authors explained the flat performance of the mixed-gender groups by stating that partners had difficulty

49

cooperating and that females tended to be dominated by their male partners who competed for keyboard control.

The hypothesis of Feldman, et al. (1991), that computer based humanities courses may be less biased than courses traditionally seen as more male-oriented, is not supported by these studies. Interestingly, in computer based humanities courses, the poor performance of female students seems to be related more to interaction styles and personality differences than to their level of computer literacy.

SUMMARY

Past studies have indicated that males tend to be dominant in a computer assisted environment and that in -mixed-gender situations their competitive attitude has a negative effect on the females in that environment. This has shown to be the case in computer assisted instruction across the curriculum, humanities classes included. Past research in computer assisted humanities courses taught in middle schools also suggests that interaction styles and personality differences seem to be additional variables influencing computer assisted learning. To date, no study has been conducted to investigate the effects of both gender and personality differences in computer based foreign language group activities at the university level, hence the research project outlined below.

PRESENT STUDY

Purpose

The purpose of this research is to investigate the effects of gender- and personality differences in computer based foreign language group activities at the university level. Based upon the survey of current research, the following four research questions were considered:

1. Do males and females perform differently regardless of dyad types?

2. Do males and females perform differently in mixed- vs. same-gender dyads?

3. Is gender a stronger factor than personality in predicting language learning and interaction patterns?

4. Does keyboard control have an effect on language learning?

50

Sample Population

The participants were sixty students (N = 60) enrolled in intermediate (second year) French classes at a major research institution. Contact forms were distributed, inviting students to participate in an experimental computer based foreign language activity. On these forms, no information was provided about the actual purpose of the research. Only experimental procedures were detailed. The forms clearly stated that students with no computer background would be offered a pre-training session on the use of Macintosh computers. All students were given a $ 1 0 incentive for volunteering two hours of their time. I deemed this monetary incentive was necessary because only a few instructors agreed to integrate their students’ participation into class work. Therefore, another form of extrinsic motivation had to be offered to select a sample that would be as representative as possible of the sampling frame. Students interested in participating in the study were asked to sign the contact form and to indicate available hours. Of the 364 students enrolled in second-year French, 137 students volunteered. An equal number of mates and females were needed in order to conduct the study. Some students had to be eliminated on the basis of their education background: All students needed to be graduates of American high schools to eliminate possible confounding variables due to differences in the socialization and education process as practiced in other countries. Other volunteers had to be eliminated on the basis of their age to control for the age factor: All students retained for this study were considered traditional age students (19 – 25). From the remaining pool of 115 volunteers, a stratified sampling design was adopted. Volunteers were categorized according to the following: (1) gender, (2) computer background (only two students did not have any and needed computer pre-training), and (3) time availability. Gender and time availability became the basic categories retained for the random pairing of students to partners of the same or opposite sex. The two students with no computer background were included in the final sample. I was aware that the lack of computer background for these two students could have been the source of a possible confounding variable. However, the final results indicated that the learning achievement and interaction patterns of these two students were consistent with other students of the same gender and personality type, and that computer pre-training was successful.

Design

This study included ten female dyads (or pairs), ten male dyads, and ten mixed-

gender dyads. All 60 subjects were asked to participate in pre-test sessions. Pretest measures consisted of the Myers Briggs Type Indicator (MBTI) personality profile test, and a software-based pretest (both tests are detailed in the “instrumentation” section). Subjects then played the French version of a computerized mystery game, Carmen San Diego, in dyads, and performed a post-test. Carmen San Diego involves (1) French reading skills in French, (2) decision making skills, and (3) computer manipulation.

51

The experiment took place entirely in a computer laboratory. In addition to the experimental manipulation, a descriptive component was integrated into the design to examine interaction patterns by means of video taping.

Instrumentation

To establish the participants’ personality profile, the MBTI was selected because it has been used previously as an adequate, valid, and reliable instrument by other L2 researchers (Ehrman and Oxford 1988, 1995; Ehrman 1994, Oxford 1990; Carrell and Monroe 1993). The MBT1 instrument was originally designed according to Jung’s psychological theory advocating that personality variation relies on orderly and systematic dynamics. These orderly dynamics are described below across four different psychological traits, each trait defined in terms of dichotomies:

Extraversion (E) or Introversion (I)

This factor expresses a person’s preferred attitude toward the world. Extraverts (E) are stimulated by people and things, whereas introverts (I) are stimulated by inner thoughts and reflections.

Sensing (S) or Intuition (N)

This factor indicates the perception mode, that is, how individuals receive and process information. Sensing (S) types learn best through their ‘Use of sight, hearing, touch, smell, and taste. Intuitives (N) use their sensory perception less and rely more on imagination and supposition to guide information gathering.

Thinking (T) or Feeling (F)

Decision making, according to Jung, occurs with either thinking M or feeling (F). Thinkers (T) prefer to use logic and analysis to decide. Feelers (F) prefer to use subjective values to arrive at decisions.

Judgment (J) or Perception (P)

This dimension refers to lifestyle. judging G) types prefer an orderly lifestyle, both in their values or practical organization, and get satisfaction out of finishing things. Perceiving (P) types are flexible, adaptable and are feeling comfortable in unstructured situations (Barrett, Sorensen, and Hartung 1987, 1-2).

52

In addition to the MBTI, students took software-based pre-and post-tests which evaluated their language-learning achievement in this experiment. Both pre-

and post-tests measured two competencies: (1) French vocabulary pertaining to descriptions of people, and 2) French vocabulary pertaining to geographical knowledge. The pretest procedure was meant to control for preexisting language differences across students by establishing every student’s initial reading proficiency. The results of the post-test were compared to those of the pretest to gauge the instructional efficiency of the language software across genders, personalities and dyad types. I postulated that the CALL treatment would help develop a more automatic reading process during the post-test and that a more controlled and slower reading process would be used during the pretest. In these tests, the participants had to pretend to be detectives reading reports on the whereabouts and descriptions of criminals, a task similar to that required by the software. Written instructions given in English were read to the students before the test was administered. The whole test was limited to 20 minutes, with 1 0 minutes for each of the two parts. In order to avoid a ceiling effect, the tests contained more questions (total = 130) than students could possibly answer in 20 minutes. Students were not expected to complete all items; however, they were advised to work efficiently and accurately, and were asked to complete as many items as possible within the time allotted. Pretests and post-tests contained identical items, but questions were scrambled and reorganized for the post-test to minimize transfer of learning from pretest to post-test.

Carmen San Diego was selected for the CALL treatment because it is a mystery game that falls into the category of “collaborative” software (Meunier 1992; Wyatt 1987) defined as computerized open-ended and discovery activities which motivate group discussions prior to making branching decisions. Because one important aspect of this research study is the description of group dynamics, I came to the conclusion that Carmen San Diego was ideally adequate. (This hypothesis was confirmed by the results obtained through a pilot test used to determine the suitability of the instrumentation.) Prior to the 30-minute treatment, I explained and demonstrated the game to students. After the demonstration, I asked participants to play in dyads for 30 minutes, to share the keyboard and mouse, and to collaborate as a team on solving mysteries.

In addition to the experimental treatment, I observed and investigated interactional and conversational patterns. The 30-minute treatment session was videotaped to increase the accuracy of analysis and facilitate interpretation. To help reduce the inhibiting effects of video taping on behavior, a camera was set up with no operator. A tiny wireless FM microphone was used to record the conversation of both partners. The participants’ use of slang and some curse words was taken as a good indicator that they were not disturbed by the recording equipment.

53

Measurement

For each of the four psychological constructs contained in the MBTI, that is Extraversion (E) – Introversion (@; Sensing (S) – Intuition (N); Thinking (1) – Feeling (F); and judging U) – Perception (P); detailed in the “instrumentation” section, each subject favors one pole over the other, as illustrated below. All psychological constructs are measured through a series of multiple choice questions with different weights assigned to each answer on the basis of prediction ratios (0, 1, or 2). Points for each pole are then totaled, and the strongest pole determines the psychological preference of the testee within each psychological construct. For example, if a subject had the following final scores, the personality profile of that person would be INFP:

E — I S – N T – F J – P

8 22 12 13 2 18 1 28

Because the number of subjects per personality profile was too narrow, computation lacked statistical distribution. To overcome this problem, the computation was run using function pairs for interaction profiles (IP, IJ, EP, EJ) and for learning profiles (ST, SF, NF, NT) as independent variables, two subgroups of the MBTI personality instrument (Briggs Myers and McCaulley 1992). The participants’ learning profile was considered the independent variable in computations dealing with learning achievement, and the interaction profile was deemed the independent variable in studying interactional behaviors during the experiment.

After the personality profiles were established, the pre- and post-tests were scored. The learning achievement of each participant was measured by subtracting pre- from post-test results.

Once the participants’ gain scores were calculated from pre- and post-tests, I analyzed and classified all conversations — on the 15 hours of video tapes — that took place at the computer according to three categories: (1) task statements, (2) management statements, and (3) social statements. All statements in each of these categories were added and then compared across genders, personality profiles, and dyad types: FF (female only), FM (mixed-gender), MM (male only). A second rater coded 10% of the transcripts according to the given criteria in order to provide a sample for inter-rater reliability. The correlation coefficient (r = .95) obtained by applying a Fisher’s R to Z procedure was significant (p < .001) and indicated very little variance in the rating procedure.

I analyzed the video tapes a second time to measure keyboard control. Keyboard control was measured in seconds with stopwatches then averaged for each gender, personality profile, and dyad type. In this study, every mouse or keyboard manipulation was considered an instance of “keyboard control.” Because Macintoshes

54

lend themselves to greater keyboard participation by the person sitting on the right of the computer (the mouse is attached to the right of the Macintosh), the position of subjects at the computer also had to be included as an additional variable while computing data (in spite of the fact that the mouse was placed in front of the keyboard, between the participants of each dyad).

STATISTICAL PROCEDURES

I chose t- tests to analyze gender differences in learning achievement. ANOVA was used to (1) investigate the relationships between dyad types and genders (independent variables) and test results (dependent variable); (2) examine whether gender or personality was a stronger predictor of language learning, interaction patterns, and keyboard control; and (3) measure the effect of keyboard control on language learning.

DATA ANALYSIS AND RESULTS

Results of General Comparisons between Males and Females

The difference between post- and pretest scores was first established for all males and females, regardless of dyad type. Results were subdivided into three categories: (1) overall learning, (2) learning of vocabulary pertaining to descriptions of people, and (3) learning of vocabulary pertaining to geographical facts. The results of the t-test performed on the overall learning quota indicated that the difference between males and females was not statistically significant (Table 1). The rate of overall learning was then broken down into (1) vocabulary pertaining to geographical facts, and (2) vocabulary pertaining to story characters. Males outperformed females numerically in acquiring vocabulary dealing with geographical facts (Table 1). Females outperformed males with differences that were statistically significant (Mean cliff. = 6.73, t = 2.34; p = .02 <.05) in the acquisition of vocabulary pertaining to story characters.

N

Mean

Std. Err.

t Value

p Value

Overall F Learning

30

18.00

1.34

Overall M Learning

30

15.00

1.32

1.29

.20

F Learning of vocab./geography

30

15.62

1.33

M Learning of vocab./geography

30

17.63

1.74

-.91

.36

F Learning of vocab./people description

30

12.99

1.79

2.34

.02

M Learning of vocab./people description

Table 1: General comparisons between male and female learning

(regardless of dyad type)

55

Effect of Dyad Type

Gain scores illustrated in Table 2 show no statistical significance in the differences computed through ANOVA. Clearly, neither males nor females are significantly affected by group types. Results indicate that in mixed-gender dyads, females tend to acquiesce to male factual preferences by learning more vocabulary pertaining to geographical facts and by lowering their interest in vocabulary related to story characters. On the other hand, males tend to assert themselves -with geographical facts in mixed-gender dyads, and there is no adjustment on their part to female interests, 1 e. story characters. Relatively lower scores of males in MM groups for the acquisition of geographical facts may stem from a lack of cooperation between males or from a lack of motivation on their part. One of my colleagues suggested that the phenomenon of greater male learning in mixed groups could also result from males showing off for females. However, video analysis displayed neither verbal nor behavioral clues supporting this hypothesis. The following analysis will show that personality differences actually account for this phenomenon (see discussion).

Overall Learning

Vocab./Geo

Vocab./People