Chapter Three

An Investigation of Learning Styles in General Chemistry Students


RESEARCH DESIGN

The Purpose

In order to evaluate the model of learning based on Dr. Carl Jung's theory of personality type, it is necessary to appraise how well the descriptors fit students, and whether modification of instruction or study techniques may improve student achievement. It is of interest whether the model may assist in determining which students do well in chemistry, and which students may be assisted by differences in presentation of content and concepts. This chapter will describe the research conducted in efforts to evaluate the model.


The Subjects

The subjects were all students of Clemson University. During the Fall semester, 1994, the pilot study was conducted with students regularly enrolled in Psychology 101. Phase one of the research was conducted during Spring semester, 1995, with students regularly enrolled in the second semester of the general chemistry sequence, CH102 . Phase two, the replication, was conducted during the Fall semester, 1995, with students regularly enrolled in the first semester of the general chemistry sequence, CH101. Since the phase two course precedes the phase one course, there is no chance that a student in phase one would also be enrolled in phase two. Since all students were regularly enrolled in the courses and sections, they represent a highly self selected population. Additional demographic data were collected with the cognitive profile inventory, and are reported in chapter 4.

Pilot Study

A pilot study was conducted to evaluate the experimental instrument (Appendix A) which is a modification of one used by Hansen, Silver and Strong (Silver, 1980). The first part of the instrument consisted of twenty sets of four words each, with the word order randomized within each set. Each set of four words occupied a horizontal row of type. The subject was instructed to rank the words in each set in order of personal preference. "Not your parent's, not your high school guidance counselor's, not your friends', your own preference." The students were further instructed that there are no right or wrong answers, and no word is to be considered a better or poorer choice than any other word. There is no judgment attached to any answer. An example uses vanilla, chocolate, strawberry and pistachio ice cream flavors. For myself, I would choose chocolate first, strawberry second, vanilla third, and I would only choose pistachio if there were no other flavor available. Subjects assign a 5 to their first choice, a 3 to their second choice, a 1 to their third choice, and a 0 to their last choice. In the example, I would give chocolate a "5," strawberry a "3," vanilla a "1" and pistachio a "0." They were told that the more honest they were in their choices and rankings, the more useful and valuable the resulting information would be to them.

After completing the rankings, the subjects were instructed to turn to part two. Part two had the same sets of words, but here the words were ordered into columns associated with the Jungian quadrants SF, ST, NF, and NT. The subjects were then instructed to transfer the number assigned to each word in part one, to the same word in the same group of words in part two. The randomization of word order in part one prevented choices in ranking to be influenced by recognition of patterns by column. As students transferred their numbers, many commented on beginning to see a pattern emerge of 0's, 1's, 3's, and 5's in certain columns. After all rank numbers had been transferred, the subject summed the columns, and then summed the totals for all four columns to confirm a grand total of 180. A total of other than 180 indicated an error in either ranking, transferring, or addition, and an opportunity to find and correct the error. Part three consisted of a diagram of an x,y axis placed in a square (Figure 1). The quadrants are bisected with diagonals. The upper right quadrant is labeled SF, the upper left is ST, the lower left is NT and the lower right is NF. The subjects were invited to transfer their numbers to the quadrants. The origin, center, of the axis is 0. Each corner is 100. Subjects were instructed to find and mark that point on each bisector equivalent to the sum for that column. For example, a sum of 50 for SF would be half way between the corner and the center of the upper right quadrant. The four points were then connected to form a quadrilateral. That figure represented the subject's cognitive profile. It emphasized that every learner has some area in each quadrant. One or more quadrants may have been dominant, but there was some area in each. Upon forming their profile, it was natural for subjects to be curious about what it meant. At this point in administration, explanation was given of the diagram and the significance of the quadrants, with time for questions and comments from the subjects.

Approximately 100 Psychology 101 students were administered the instrument, and were subsequently asked to evaluate the directions, specific vocabulary, and format. The pilot instrument had been slightly modified from the Hansen Silver Strong Learning Inventory (Silver, 1980) for agreement of case and tense of words in groups, and in physical arrangement of words on the page as presented to the student. Some words were replaced with others that were closer to Jung's original model and/or more familiar to the student subjects. Student comments revealed confusion about directions and dislike of turning pages repeatedly in order to transfer ranking numbers. These students reported they had little difficulty with the vocabulary of the instrument. Any questions which related to meaning of words required a definition in the appropriate sense of the word in context of the choices presented in the set, and a completely non-judgmental definition, conveying no sense of the administrator's preference, approval or disapproval. Subsequent to this research, the instrument was again modified for a more diversified population to include appropriate definitions with every word in part one.

Following the pilot study some specific words were changed as to form of speech in order to be more appropriate to ranking. The instrument was modified to a format which could be used with a Scantron form (Appendix B) which is familiar to the students. This format eliminates page turning back and forth, is easily administered, evaluated, and processed with a large population. Scantron administration also permitted evaluation of the profile to be selectively revealed to subjects at the appropriate time, based on control or treatment group membership. The quadrant profile numbers, sums in the pilot study version, were calculated as percentage prior to being returned to the subjects. It was determined that an expectation of 25% in each quadrant is natural, and values greater or less than 25% in a quadrant is easier to understand as dominant or sub-dominant by the students. This replaced the "total of 180" basis of the pilot study pencil and paper version of the inventory instrument.

Institutional Review Board Approval

Application was made to the Institutional Review Board for approval of this research. Approval was granted with modifications as required by the committee to the Informed Consent Form (Appendix C). The full study, including replication, is number 95-039.

All students taking Chemistry 102 in the Spring semester of 1995 and all students taking Chemistry 101 in the Fall semester of 1995 were administered the modified learning style evaluation test instrument and relevant demographic questions (Appendix B).

Students in the Chemistry 102 treatment group attended an instructional session early in the semester in which they were given their individual learning style profile in printed form and verbal interpretation of the material. Study techniques specific to the individual learning styles were described. Grades on periodic exams given during the course were evaluated by group, experimental vs. control, and no difference was found after exam two. Since students had only been exposed to the study techniques information during a single meeting early in the semester, it was thought that they might not have had adequate exposure to the techniques. Mid semester, they were given the study technique information again, in printed form (Appendix D). Students in the control group were given the same information, individual printed learning style profiles and study techniques, after the final exam in the course. Additionally, students in the experimental group of the CH102 part of the study, Spring semester 1995, responded to questions regarding "fit" of the descriptive information they were given about their learning profile results, i.e., whether they felt the description of their profile was accurate in describing themselves (Appendix E).

All students in the CH101 part of the study, Fall semester of 1995, responded on their final exam to questions regarding what study techniques they use (Appendix F). Since the population of Clemson University General Chemistry students is a self selected group, it is possible that these successful students already have determined what study techniques work for them, and we may be telling them what they already know.

Application for approval of the study was made and approval of the study was granted by the Institutional Review Board, as was approval for continuation of the study in replication. Initial approval identified the research as #93-034 and was granted in December 1993.

Research Questions

The following questions guided the research:

RQ 1. Does knowledge of learning profile and appropriate study techniques improve achievement in General Chemistry?

Analysis of the data included analyses of variance for differences between overall means and distributions of scores, overall and as demographic groups on the variables of gender, major, student status (freshman, sophomore, junior, senior) and ethnicity in a factorial design. A null hypothesis of no difference between treatment and control was tested overall and for each listed demographic group.

RQ 2. Which, if any, of SAT scores, math and/or verbal, normalized high school rank in class, or predicted GPA, may predict achievement in General Chemistry?

SAT math, SAT verbal, normalized high school rank in class, and predicted GPA was provided by Institutional Research. Regression analysis was performed on these data to determine predictor value of each of these scores for final grade in General Chemistry CH102. Previous studies reported in the literature (Andrews, 1979; Ozsogomonyan, 1979) cite SAT math as the best predictor of success in chemistry. Since the population under study is self selected as Clemson University students taking general chemistry, it is desirable to reaffirm this result with this population. The null hypothesis of no difference in predictor value of the variables was tested.

RQ 3. When a predictor variable, as identified in question two, is used as a covariate to control for performance as a factor, does knowledge of learning profile and appropriate study techniques improve achievement in General Chemistry?

The strongest predictor found in the regression was used as a covariate for further ANOVA analyses, to determine whether the student's ability level may significantly skew the efficacy of learning style study skills on final grade. The null hypothesis of no difference between treatment and control groups will be tested.

RQ 4. Do some types of learners do better in General Chemistry than other types of learners?

RQ 5. Does knowledge of learning profile and study techniques make more of a difference in achievement in general chemistry for some learning profiles than for others?

Final grade in CH101 or 102 was evaluated based on relative strength in each quadrant of the Jungian Profile. The null hypothesis of no difference in final grade by cognitive profile, overall and by group, will be tested.

RQ 6. How well did students in the CH 102 portion of the study feel the profile described themselves?

Students in the experimental group were asked on their final exam to respond to additional questions in survey form, using a Likert scale (e.g. A = all the time, B = most of the time, C = frequently, D = occasionally, E = never) relating to their use of the study techniques taught to them, or perception of fit of the profile descriptors (Appendix E).

RQ 7. Do students use study techniques appropriate to their learning profile, as measured by self reported data? Do those students in the experimental group, who were taught appropriate study techniques, use them to a greater extent than those in the control group?

Students in both the experimental and control portions of the CH 101 portion of the study were asked about the study techniques they use, in additional survey questions added to their final exam (Appendix F). Since the population of students at Clemson University who are taking general chemistry is already a population of successful students, it may be that these students have already discovered what works best for them.

All students taking CH 102 or CH 101 each semester under study were given 2 points extra credit for participating. Students were told that they must actively participate, attend meetings, etc. in order to receive the two points, but since final grade is a variable under investigation, all students will receive the two points. The study was replicated, as performed with Spring `95 CH102's, with all students taking CH101 in the Fall semester of 1995. The Fall replication was modified to include the use of additional printed materials at the beginning of the study for the treatment group to reinforce the study technique instruction. These additional materials included the same material specific to learning styles under the Jungian model, and appropriate study techniques as was given to the CH102 treatment group at mid-semester (Appendix D). The CH102 phase of the study initially included approximately 1000 students and the CH101 replication an additional 1200 students at the beginning of the semester.


Statistical Analyses of the Data

Data for phases 1 and 2 were gathered by questionnaire administered the first day of class on 957 CH 102 students during the Spring semester of 1995 (phase 1) , and 1183 CH101 students during the Fall semester of 1995 (phase 2). These results were taken by Scantron, tabulated initially in Excel on a Power Macintosh ® 7100/80AV, and ported to the Clemson University mainframe. Most of the analyses were accomplished on the Clemson mainframe, a bank of 4 HITACHI Model 3090 machines. The statistical package used was SAS ® , Proprietary software release 6.07 TS305 licensed to Clemson University, site 0001151001, from the SAS Institute Inc., Cary, NC, USA.

Group Assignment

Frequencies were tabulated in SAS ® for gender, section, major, year and ethnicity by a function identified as odd or even last digit of the student ID number. Chi Square analyses resulted in probabilities of .87 or higher for every breakout, showing all demographics to be random by this function. Students were then assigned to either the control group or the treatment group by odd or even last digit of their student ID number. In most cases, this number is the student's social security number with the exception of foreign students who do not have a US social security number. These students are assigned a similar number by the university. In any case, the student ID number on file with the University is the number used for purposes of group assignment. This was a convenient system to use, since the number is recorded on all Scantron forms used, and routine quizzes and exams in the course are tabulated on Scantron forms by student ID number.


Demographics

Frequencies were tabulated in SAS using Proc Freq with by variables identified for gender, major, year, and ethnicity as identified by the student on the questionnaire, and by group. Each student's cognitive profile was calculated from the student's ranking of 20 sets of 4 words per set, by preference (Appendix B ). The words were presented to the student in random order, and sorted in code for tabulation in proper sequence. A checksum was calculated for each student, to correct for student errors in ranking procedure, and this checksum was used to calculate the percentage preference for each quadrant of the individual profile. Each quadrant is recorded in the student cognitive profile as a percentage, the four quadrants summing to 100%.

Analyses for the Research Questions

Univariate evaluations were performed on the distribution of quadrant scores, and by general linear models procedure (Proc GLM) in SAS. The means for each of the four quadrants ranged from 24.6 to 26.0, with standard deviations from 6.6 to 8.4. All quadrant scores were normally distributed, with W: Normal of 0.98 for all quadrants. The means and standard deviations were used to determine descriptive intervals for quadrant scores.

Frequency count procedures within SAS (Proc Freq) with Chi square analyses were run on student quadrant dominances by demographic variables. A quadrant value was considered very low if 0% to less than or equal to 10 %; low if greater than 10% to less than or equal to 20%; mid to high if greater than 20% to less than or equal to 30%; and very high if greater than 30%.

ANOVA, MANOVA, MANCOVA and Correlation analyses were accomplished with the General Linear Models , Proc Univariate with normal plot routines, and means testing under standard SAS code practices, including Tukey evaluation of least significant differences between multiple groups, alpha=0.05, and least squares means with standard error and PDIFF options where SAT math scores were used as a covariate. These were used to analyze differences in grades between and within subgroups, by demographics and by cognitive dominances. All relevant results of all evaluations are published in chapter four of this dissertation.