DEVELOPMENT OF UNDERSTANDING AND SELF-

CONFIDENCE IN MATHEMATICS; GRADES 5–8

Markku S. Hannula, Hanna Maijala, & Erkki Pehkonen
Department of Teacher Education, University of Turku, Finland

This paper presents some preliminary results of the longitudinal aspect of a research
project on self-confidence and understanding in mathematics. We have collected a
survey data of 3057 fifth-graders and seventh-graders and a follow-up data of ten
classes (191 pupils) one and a half years later. The longitudinal data indicates that
the learning of mathematics is influenced by a pupil’s mathematics-related beliefs,
especially self-confidence. Pupils’ level of understanding fractions also influences
their developing understanding of infinity. These relationships between different
variables depend also on pupils’ gender and age.
INTRODUCTION
Pupils’ conceptions on themselves as learners are strongly connected with what kind
of general attitudes they have toward the discipline in question. Mathematics is a
highly valued discipline in school, and therefore, pupils experience success in
mathematics important. It has been observed, that pupils’ beliefs on mathematics and
on themselves as mathematics learners have a central role in their learning and
success in mathematics (e.g. Schoenfeld 1992). The importance of beliefs in
mathematics education is in concordance with the constructivist understanding of
teaching and learning. We understand beliefs as “an individual's understandings and
feelings that shape the ways that the individual conceptualizes and engages in
mathematical behavior” (Schoenfeld 1992, 358). Mathematical beliefs can be
divided into four main components: beliefs on mathematics, beliefs on oneself as a
mathematics learner/applier, beliefs on teaching mathematics, and beliefs on learning
mathematics (e.g. Lester et al. 1989).
Mathematics can be described as a combination of calculation skill and competence
in mathematical reasoning, but neither of these alone characterizes mathematics.
There is much research evidence that many pupils learn mathematics as a symbol
manipulation without meaning (e.g. Resnick & Nelson – Le Gall 1987).
Mathematical understanding can be distinguished from the neighbourhood concepts
‘skill’ and ‘knowledge’, for example as follows: Mathematical knowledge answers
the question ‘What’, and one may remember mathematical facts. Mathematical skill
answers the question ‘How’; which includes, for example, the traditional calculation
skill (procedural knowledge). Only mathematical understanding answers the ‘Why’ -
question; it allows one to reason about mathematical statements. These are
intertwined concepts, since understanding contains always knowledge and skill.
Another view perceives mathematical understanding as a process that is fixed to a
certain person, to a certain mathematical topic and to a special environment (Hiebert
& Carpenter 1992).

Proceedings of the 28th Conference of the International

Group for the Psychology of Mathematics Education, 2004 Vol 3 pp 17–24
 Several studies have shown that beliefs about oneself have a remarkable connection
with success in mathematics (e.g. Hannula & Malmivuori 1996, House 2000).
However, to establish a causal relationship between self-concept and achievement is
more problematic. In a literature review, Linnanmäki (2002) found out that in some
studies no evidence for causality could be found, in other studies evidence was found
for the causality from self-concept to achievement, while yet others found evidence
for an opposite direction. The seemingly contradictory results indicate a
developmental trend, where causality is mainly from achievement to self-concept
during the first school years, it changes into a reciprocal linkage for the latter part of
the comprehensive school, and in the upper secondary school level the causal
direction is from self-concept to achievement (Chapman, Tunmer & Prochnow 2000).
In her own study on self-concept and achievement in mathematics, Linnanmäki
(2002) found evidence for this developmental trend in mathematics for grade 2 to
grade 8 pupils. Looking at a more broadly defined concept, attitude, Ma and Kishor
(1997) synthesised 113 survey studies of the relationship between attitude towards
mathematics and achievement in mathematics. The causal direction of the
relationship was from attitude to the achievement. Although the correlations were
weak in the overall sample, they were stronger throughout grades 7 to 12, and in
studies that had done separate analysis of male and female subjects.
Gender differences favouring males in confidence in mathematics are well recorded.
Differences among teenagers have been reported, for example, by Bohlin (1994),
Hannula and Malmivuori (1997), Pehkonen (1997), and Leder (1995). Vanayan et al.
(1997) reported that already in grade 3 boys estimated themselves to be better in
mathematics than girls. In mathematics achievement the results on gender differences
are less clear. In IEA’s large international studies the gender differences previously
favoured boys (Husén 1967). In more recent studies the gender differences have
decreased and in many countries disappeared completely (Beaton et al. 1997).
However, robust gender differences are still found, for example, in some tasks on
infinity (Hannula et al. 2002) and fractions (Hannula 2003).
The focus of this paper is to reveal the development on pupils’ understanding and
self-confidence from grade five to grade eight. Furthermore, the most important
predictors of results are looked for.
METHOD
The study forms a part of a research project “Understanding and Self-Confidence in
Mathematics” financed by the Academy of Finland (project #51019). The project
contains a large survey with a statistical sample from the Finnish pupil population of
grades 5 and 7 with 150 school classes and 3057 pupils. The survey was implemented
fall 2001, and the information gathered was deepened with interviews and
observation in 10 classes at convenient locations. In spring 2003 the questionnaire
was administered a second time in these 10 classes. Altogether 101 pupils in the
younger sample and 90 pupils in the older sample have answered our questionnaire
twice. Because the number of classes in the longitudinal sample was small we need to

3–18 PME28 – 2004

 control for possible deviations before making generalizations. The questionnaire was
planned especially for the project. It contained, in addition to background variables,
19 mathematics tasks, estimations on success expectation and success confidence as
well as a belief scale (25 items); see for more details Nurmi et al. (2003).
For the analyses of the longitudinal data we recoded the summary variables into three
categories (lowest quartile, middle values, highest quartile) and used general linear
model multivariate analyses (GLM Multivariate). It provides regression analysis and
analysis of variance for multiple dependent variables by several covariates. The first
measures of variables were considered as the covariates and the second measures of
the same variables as the dependant variables. For each dependent variable, the
overall 2 (eta-squared) statistic is reported as a measure of the proportion of total
variability attributable to the covariates. Statistically significant (p < .01) 2 are also
reported separately for each pair of covariate-dependent variable as a measure of the
proportion of total variability attributable to the specific covariate.
ON RESULTS
In this paper we will report the results of the longitudinal data. There are many
published papers on some specific features of the initial results for the research
project: on infinity (Hannula et al. 2002), on confidence (Nurmi et al. 2003), and on
number concept (Hannula 2002, 2003).
Here we shall use three sum variables for success in mathematics test (fractions,
infinity, other tasks), and three sum variables for beliefs (self-confidence, success
orientation, defence orientation). The mathematics variables are based on the
analyses made in Hannula (2002). The two first ones (fractions, infinity) represent
our indicator for understanding, and the third variable (other tasks) consists mainly of
more computational tasks. Belief variables (self-confidence, success orientation,
defence orientation) were constructed with the help of factor analyses from the belief
scale (cf. Nurmi et al. 2003). The self-confidence factor consists of ten statements
that were adopted from the self-confidence subscale of Fennema-Sherman
Matehmatics Attitudes scales (Fennema & Sherman 1976). For background variables
we shall control the effects of gender and grade.
The longitudinal sample of the fifth-graders did not differ much from the overall
sample, if we look at the averages of all sum variables (there is only a small effect1
favouring focus classes in fractions). However, the seventh grade longitudinal sample
had better skills in mathematics (small to medium effects) and a slightly lower
defence orientation (small effect) than the full sample. The largest effect was found in
infinity, where the large sample of 7th graders had a mean score 4.7 (SD 2.9) while
the longitudinal sample had a mean score 5.9. In our interpretations of the results we
need to be aware of these deviations from the larger sample.

1
Here we use the d-value = |mean1 – mean2| / SD as a measure for difference and a convention established by Jacob
Cohen (Cohen, 1988) that sets norms for “small,” (d = 20) “medium,” (d = 50) or “large” effects (d = 80).

PME28 – 2004 3–19

 Development observed
We noticed a rapid development in all achievement variables from grade 5 to grade 6.
The development continued to be rapid in the domain of infinity after grade 7, but it
slowed down in other variables, probably due to a ceiling effect2 (Figure 1). In belief
variables, we saw decline in self-confidence and success orientation together with an
increase in defence orientation from grade 5 to grade 6 and from grade 7 to grade 8.
However, grade 7 measures differed from grade 6 measures to another direction.
Here we need to be aware of the differences between the younger and the older
sample (Figure 2).

12

10

8 5th grade
6th grade
6
7th grade
4 8th grade
2

0
Infinity Fractions Other tasks

Figure 1. Development in mathematics achievement (infinity, fractions, other tasks).

4
3,5
3
5th grade
2,5
6th grade
2
7th grade
1,5
8th grade
1
0,5
0
Self-confidence Success orientation Defence orientation

Figure 2. Development in belief variables (self-confidence, success orientation,

defence orientation).
Close to half of variation in mathematics achievement was predicted by achievement
in the previous test (Figure 3). Only a small part of this variation was attributable to
pupil’s achievement in the same topic in the previous test. Fractions was an important

2
The theoretical maximums for variables are 14 (infinity), 13 (fractions), and 12 (other tasks).

3–20 PME28 – 2004

 predictor for success in infinity and other tasks. For the belief variables we see that
self-confidence and success orientation were fairly well predicted by earlier beliefs
(Figure 4). Self-confidence was a more important predictor of these two variables.
Defence orientation seemed to be relatively unstable variable, and only 8 % of the
variation in the later test could be explained.

First measure 2 (%) Second measure Total

2 (%)
Infinity 7 Infinity 45
4 7

Fractions 17 Fractions 39

10 16
Other tasks 45
4
Other tasks

Figure 3. GLM Multivariate analyses of mathematics achievement

First measure 2 (%) Second measure total
2 (%)
Self-confidence 31 Self-confidence 35
12
16 25
Success o. Success o.

Defence o. 7 Defence o. 8

Figure 4. GLM Multivariate analyses of belief variables.

When beliefs and achievement were combined in one model, the explained variance
for fractions increased from 39 % to 46 %. Minor increase (1 – 4 %-units) was
observed for all other variables except success orientation. A statistically significant
new effect was found from self-confidence to fractions (2 = 6 %). When analyses
were made separately for boys and girls, and for the two age samples, we found some
variations in the model:
Infinity was more strongly predicted by fractions in older samples (2 = 13 % (girls),

2 = 15 % (boys)) and infinity predicted fractions among older girls (2 = 16 %).
Other tasks became a less stable variable in older sample.

Gender and age had an influence on the stability of beliefs. Self-confidence in the

younger sample was more stable among boys (2 = 11 % (girls), 2 = 40 % (boys)),
while in the older sample among girls ((2 = 37 % (girls), 2 = 25 % (boys)). Success

PME28 – 2004 3–21

 orientation was more stable in older samples (2 = 35 % (girls), 2 = 24 % (boys)) and
defence orientation was more stable in older boys’ sample (2 = 21 %).
Self-confidence was strongly predicted by infinity in older girls’ sample (2 = 36 %).

We made a more detailed analysis on the connection between fractions and infinity
because the connection seemed to be an important one, not easy to explain
theoretically, and because the underlying factor constructs were not very strong
(Hannula, 2002). We made a GLM multivariate analysis on task level, and found that
one fraction task (Figure 5) was by far the most important predictor, and that the
effect was on two of the three infinity tasks: “How many numbers are there between
numbers 0.8 and 1.1?” and “Which is the largest of numbers still smaller than one?
How much does it differ from one?” Looking at the correlations between the two
measures of these three variables we found the strongest correlations between the
first measure of the fraction task and the later measures of the infinity tasks.

0 1

Figure 5. The task 2c: mark 3/4 on the number line.

CONCLUSION
There is significant development during grades 5 to 8 in topics measured in the
mathematics test. Most notably the development is rapid both in the topics that are
covered in the curriculum (fractions), but also in topics that are not dealt directly with
in the curriculum (density of rational numbers). At the same time, there is –
somewhat paradoxically – a negative development in beliefs. In both samples self-
confidence and success orientation became lower in the second measurement, and
defence orientation increased. Confusingly, the beliefs of the older sample at the
beginning of the seventh grade were more positive than the results from the younger
sample at the sixth grade. Partially the difference can be explained by the differences
in the samples (the older sample deviating from the average towards more positive in
both achievement and beliefs). This somewhat odd difference in the 6th and 7th
graders’ beliefs might also be partially due to the effects of time of the measurement,
beliefs possibly declining by the end of spring term and increasing again by the
beginning of a new term (’a fresh start’).
Stability of the measured belief variables seems to be related to pupils’ gender and
age. Defence orientation is the least stable of the constructed belief variables, and we
might even question the validity and usefulness of the variable. However, there seems
to be a developmental trend for this orientation to become more stable among older
boys. Possibly defensive approach to mathematics is something that develops slowly
during school years and more typically for boys.
Mathematics achievement in this specific test can be predicted to a large extent from
the pupils’ past achievement in the same test. Most notably the pupils’ success in

3–22 PME28 – 2004

 fractions becomes an increasingly important predictor for future achievement in the
used tasks on number concept, as the pupils grow older. As a specific task, pupils’
ability to perceive a fraction as a number on a number line predicts their future
understanding of density of number line. This finding highlights the importance of
number line as a conceptual tool.
Self-confidence is another variable that seems to be an important predictor for future
development. A pupil’s self-confidence predicts largely the development of self-
confidence in the future, but also the development of success orientation and
achievement. A strong connection between self-confidence (and other beliefs on
oneself) and mathematical achievement has been found also in earlier research (i.a.
Hannula & Malmivuori 1997; Tartre & Fennema 1995).
Regarding the relationship between beliefs and achievement, our analyses suggest
that the main causal direction already from grade 5 onwards is from self-concept to
achievement. In the older sample we also found achievement in infinity to be a strong
predictor for the development of the girls’ self-confidence, which supports the
hypotheses of a reciprocal linkage. Like the results of Ma and Kishor (1997), also our
findings indicate that gender is an important variable in any analyses of causal
relationship between affect and achievement.
REFERENCES
Beaton, A. E., Mullis, I. V. S., Martin, M. O., Gonzales, E.Y., Smith, T. A. & Kelly S. D. I.
(1997). Mathematics Achievement in the Middle School Years: IEA’s Third International
Mathematics and Science Study. Chestnut Hill MA: Boston College.
Bohlin, C. F. (1994). Learning style factors and mathematics performance: Sex-related
differences. International Journal of Educational Research 21 (4), 387-398.
Chapman, J. W., Tunmer, W. E. & Prochnow, J. E. (2000). Early reading-related skills and
performance, reading self-concept and the development of academic self-concept: A
longitudinal study. Journal of educational psychology 67, 145-152.
Cohen, J. (1988), Statistical Power Analysis for the Behavioral Sciences, Academic Press,
New York, 2nd edn.
Fennema, E. & Sherman, J. A. (1976). Fennema-Sherman mathematics attitudes scales.
JSAS Catalog of Selected Documents in Psychology, 6, 31 (Ms. No. 1225).
Hannula, M. S. (2002). Understanding of Number Concept and Self-Efficacy Beliefs in
Mathematics. In P. di Martino (Ed.) MAVI European Workshop MAVI XI, Research on
Mathematical Beliefs, Proceedings of the MAVI-XI European Workshop, April 4-8,
2002, pp. 45-52. University of Pisa, Italy.
Hannula, M. S. (2003). Locating fraction on a number line. In N. A. Pateman, B. J.
Dougherty & J. Zilliox (Eds.) Proceedings of the 27th Conference of the International
Group for the Psychology of Mathematics Education, Vol 3, 17-24.
Hannula, M., & Malmivuori, M. L. (1997). Gender differences and their relation to
mathematics classroom context. In. E. Pehkonen (Ed.), Proceedings of the 21st
Conference of the International Group for the Psychology of Mathematics Education.
Vol. 3, 33-40.

PME28 – 2004 3–23

 Hannula, M. S., Maijala, H., Pehkonen, E. & Soro, R. (2002). Taking a step to infinity:
Student’s Confidence with Infinity Tasks in School Mathematics. In S. Lehti & K.
Merenluoto (Eds.) Third European Symposium on Conceptual Change – A Process
Approach to Conceptual Change, pp. 195–200. University of Turku. Dept Teacher
Education in Turku.
Hiebert, J. & Carpenter, T.P. 1992. Learning and Teaching with Understanding. In: In D.A.
Grouws (Ed.), Handbook of research on mathematics learning and teaching, pp. 65–97.
New York: Macmillan.
House, J. (2000). Student self-beliefs and science achievement in Ireland: Findings from the
third international mathematics and science study (TIMMS). International journal of
instructional media 27 (1), 107-115.
Husén, T. (1967). International Study of Achievement in Mathematics, vol II. Stockholm:
Almqvist & Wiksell, and New York: John Wiley & Sons.
Leder, G. (1995). Equity inside the mathematics classroom: Fact or artifact? In W. G.
Secada, E. Fennema & L. B. Adaijan (Eds.) New directions for equity in mathematics
education. Cambridge University Press.
Lester, F. K., Garofalo, J. & Kroll, D. L. (1989). Self-Confidence, Interest, Beliefs, and
Metacognition: Key Influences on Problem-Solving Behavior. In D. B. McLeod & V.
M. Adams (Eds), Affect and Mathematical Problem Solving, pp. 75-88. New York:
Springer-Verlag.
Linnanmäki, K. (2002). Matematikprestationer och självuppfattning, en uppföljninggsstudie
I relation till skolspråk och kön. [Mathematical proficiency and self-concept; a
longitudinal study with an attention to language and gender]. Turku: Åbo Akademi
University Press.
Ma, X. & Kishor, N. (1997). Assessing the relationship between attitude toward
mathematics and achievement in mathematics: A meta-analyses. Journal for Research in
Mathematics Education, 28 (1), 26-47.
Nurmi, A., Hannula, M. S., Maijala, H., & Pehkonen, E. (2003). On pupils’ self-confidence
in mathematics: gender comparisons. In N. A. Pateman, B. J. Dougherty & J. Zilliox
(Eds.) Proceedings of the 27th Conference of the International Group for the Psychology
of Mathematics Education. Vol 3, 453-460. University of Hawaii.
Pehkonen, E. (1997). Learning results from the viewpoint of equity: boys, girls and
mathematics. Teaching Mathematics and its Applications 16 (2), 58–63.
Resnick, L. & Nelson - Le Gall, S. 1987. Meaning construction in mathematical problem
solving. ERIC ED299132.
Schoenfeld, A. (1992), Learning to think mathematically: problem solving, metacognition
and sense making in mathematics, in A. D. Grows (Ed.), Handbook of research on
mathematics learning and teaching, 334-370.
Tartre, L. & Fennema, E. (1995). Mathematics achievement and gender: A longitudinal
study of selected cognitive and affective variables. Educational studies in mathematics
28, 199-217.
Vanayan, M., White, N., Yuen, P. & Teper, M. (1997). Beliefs and attitudes toward
mathematics among third- and fifth-grade students: A descriptive study. School science
& mathematics 97 (7), 345-351.

3–24 PME28 – 2004