PERSISTENCE AND SELF-EFFICACY IN PROVING

Annie Selden John Selden

New Mexico State University New Mexico State University
aselden @math.nmsu.edu [email protected]

We first discuss our perspective on, and three useful actions in, proving that depend on
persistence. Persistence is important for successful proving because it allows one to “explore”,
including making arguments in directions of unknown value, until one ultimately makes
progress. Persistence can be supported by a sense of self-efficacy, which is “a person’s belief in
his or her ability to succeed in a particular situation” (Bandura, 1995). We then examine actions
needed for a successful proof construction of a theorem given to mid-level undergraduates in a
transition-to-proof course. We contrast those actions with the actual actions of a mathematician
proving the same theorem. Then, after a brief discussion, we end with some teaching
implications.

Keywords: advanced mathematical thinking, proof, affect

Introduction
In this paper we first discuss our perspective on proof construction. Then, drawing on
observations from a multi-year teaching experiment, we point out three aspects of proof
construction that appear to be difficult to teach. We suggest that the three teaching difficulties all
arise partly from a need for students to have a kind of persistence which depends on students
having a feeling of self-efficacy. That is, the type of teaching difficulty depends, not on the
particular mathematics concerned, but on students’ views of their own abilities to work with it.
We next illustrate the way that self-efficacy and persistence are valuable by discussing the proof
construction of a specific theorem that most of our students could not prove and describe the
remarkable persistence shown by one mathematician’s in proving it. Then, after a brief
discussion, we end with some teaching implications.

An Action Perspective
While we view proofs as texts written in a certain genre (Selden & Selden, in press), we view
constructing a proof as a sequence of actions (Selden, McKee, & Selden, 2010). Some of these
are physical, such as writing a line of the proof, and some are mental, such as focusing on the
conclusion or “unpacking” its meaning. Such actions are taken in response to certain kinds of
(autonomously recognized) situations in a partly constructed proof. With practice, the links
between some repeatedly occurring proving situations and the resultant actions will become
automated, thus reducing the burden on working memory in future proof construction. Many
such actions are not recorded in the final written proof. Perhaps this is partly why it is difficult
for students to mimic a given proof, or even parts of it, when constructing another proof.

Three useful actions in proof construction

In several iterations of teaching a second-year university transition-to-proof course in a
modified Moore Method way (Coppin, Mahavier, May, & Parker, 2009; Mahavier,1999), proof
construction situations calling for one of the following three actions regularly occur. (a)
Exploring. In constructing a proof, one may understand both what is to be proved and what is

1
available to use without having any idea of how to proceed. In such situations, one might
reasonably try to prove any plausible result, including results of unknown value. However, few
students show any evidence of doing this, perhaps lacking confidence in their own ability to use
whatever new they might prove. (b) Reworking an argument in the case of a suspected error or
wrong direction. In constructing a proof, one may come to suspect one has made an error or is
arguing in an unhelpful direction. An appropriate response would be to rework part of the
argument. However, many students appear to be reluctant to do this, perhaps because they lack
confidence in their own abilities to produce something new and better than before. (c) Validating
a completed proof. Upon completing a proof, one should read it carefully for correctness from
the top down, checking whether each line follows from what has been said above. Few students
appear to do this, perhaps because they do not think that they are able to find errors in their own,
just completed, arguments.
Some student errors may depend on a wrong belief about mathematics or logic or on a
misinterpretation of a definition. Such errors can be pointed out and an explanation can be
provided by a teacher. However, the above three actions in proof construction are not about
correcting an error, but about habitually acting appropriately in particular situations. They seem
to depend on students’ views of their own abilities, that is, on a sense of self-efficacy and
persistence. We suspect that encouraging this kind of appropriate behavior may require some
kind of teaching beyond explaining errors.

Self-efficacy
Self-efficacy is “a person’s belief in his or her ability to succeed in a particular situation”
(Bandura, 1995). Also, “The most effective way of developing a strong sense of self-efficacy is
through mastery experiences,” and that performing a task successfully strengthens one’s sense of
self-efficacy. Short of such success, “Seeing people similar to oneself succeed by sustained effort
raises observers’ beliefs that they too possess the capabilities to master comparable activities to
succeed.” (Bandura, 1994).
According to Bandura (1994), individuals with a strong sense of self-efficacy: (a) view
challenging problems as tasks to be mastered; (b) develop deeper interest in the activities in
which they participate; (c) form a stronger sense of commitment to their interests and activities;
and (d) recover quickly from setbacks and disappointments. In contrast, people with a weak
sense of self-efficacy: (a) avoid challenging tasks; (b) believe that difficult tasks and situations
are beyond their capabilities; (c) focus on personal failings and negative outcomes; and (d)
quickly lose confidence in personal abilities.
Bandura’s ideas “ring true” with our past experiences as mathematicians teaching courses by
the Moore Method. Classical Moore Method (advanced undergraduate or graduate) courses are
taught from a brief set of notes consisting of definitions, a few requests for examples, statements
of major results, and those lesser results needed to prove the major ones. Exercises of the sort
found in most textbooks are largely omitted. In class meetings, the professor invites individual
students to present their original proofs and then very briefly comments on errors. Students are
typically forbidden to read anything on the topic or to discuss it with anyone other than the
professor. Once students are able to successfully prove the first few theorems, they often
progress very rapidly in their proving ability, even without apparent explicit teaching, and even
when subsequent proofs are more complex. Why should this be? We conjectured then, and also
now, that students obtained a sense of self-efficacy from having proved the first few theorems
successfully, and this helped them persist in explorations and re-examinations needed to prove

2
subsequent theorems. That is, they learned to gather as much information as they could and to
explore various ideas, whether or not they could initially “see” their usefulness.
Similarly, in discussing geometry conjectures, de Villiers (2012) has pointed out places
where a novice might

lose hope of getting anywhere as it’s not obvious from the start this will lead somewhere
useful. However, students should be encouraged to persist [emphasis ours] … and not so
easily give up … One might say that a distinctive characteristic of good mathematical
problem solvers [and provers] are that they are ‘stubborn’, and willing to spend a long time
attacking a problem from different vantage points, and not easily surrendering. (p. 8)

However, dogged persistence in a single direction, without accompanying monitoring and

control, can lead to what Schoenfeld (1985, p. 316) has called mathematical “wild goose chases.”
Further, the American mathematician Steven Krantz (2012) has written, in his book on
mathematical maturity, that

at least as important [as being smart] is perseverance or tenacity. Mathematics can easily be
discouraging. … you may find that you are stuck for a goodly period of time. … You either
do not know enough, or do not know the right things. This is why tenacity is important. You
must have adequate faith in yourself [sense of self-efficacy] to know that you can battle your
way through the problems. … Not unrelated to the idea of tenacity is the property of being
comfortable with delayed gratification. … Once you are challenged to generate your own
proofs and counterexamples, you are frequently at odds, and often frustrated. … The unifying
theme for dealing with the need for tenacity and the need to deal with delayed gratification is
self-confidence [sense of self-efficacy]. You need to believe in your own abilities, and you
need to believe that you can actually do this work. (pp. 97-99)

Thus, it would seem that a sense of self-efficacy, that is, a belief in one’s ability to succeed
on a particular kind of task, enables one to persist despite frustrations or wrong paths and that
this is an important part of doing mathematics, and in particular, of proving original results.
Indeed, we suspect that it is an important part of much creative cognition in general.

A Hypothetical Proof Construction

In this section we describe hypothetical proving actions for a theorem chosen from a set of
notes used in our one-semester, 3-hour per week, mid-level undergraduate transition-to-proof
course. In this course, students present in class their proofs of theorems from the notes and
receive substantial criticisms and advice. There is no textbook and there are no lectures. The
notes include theorems about sets, functions, real analysis, and abstract algebra, as well as
definitions and requests for examples. In this setting, the difficulty of a students’ proof
construction depends greatly on the relationship between the student’s previous experiences and
the notes. The hypothetical proof construction we describe is for one of the more difficult
theorems in the notes and occurs near the end of the algebra section: Theorem: If S is a
commutative semigroup with no proper ideals, then S is a group. To date, after three iterations of
the transition-to-proof course, only one student (of 50) has persisted in proving the theorem
correctly (albeit over the summer holiday). Thus, one can conclude that constructing a proof of

3
this theorem presents quite a challenge, although a modes change in the notes could make it
much easier.
The needed background information is very small. A semigroup is a nonempty set S with an
associative binary operation that we will write multiplicatively as xy for elements x and y of S.
Associativity means that for all elements x, y, and z of S, (xy)z = x(yz). S is commutative means
that for all elements x and y of S, xy = yx. If A and B are subsets of S, AB is the set of elements
ab, where a and b are elements of A and B respectively. In this setting, a nonempty subset I of S
is an ideal of S provided SI is a subset of I. Such an ideal is called proper in S provided it is not
all of S. In this commutative setting, S is a group if it has two additional properties. First, there
must be an “identity” element e of S so that for any element s of S, es = s. Second, given any
element s of S there must be an “inverse” element s' of S so that s's = e.
In proving the above theorem, it is easy to see that if I is an ideal of S, one can conclude I is
not proper, that is, I = S. What is not so easy is trying to construct an ideal that “looks” different
from S, and what that might have to do with producing an identity element e of S and inverses, in
order to prove that S is a group. Since there is nothing else to work on, one must persist in trying
to find an ideal of S without any idea of whether, or how, that might be helpful. It turns out that
some students rather quickly think that if s is any element of S then S{s} (also written Ss) might
be an ideal of S, and hence equal to S. Once the idea has been articulated, it is not so hard to
prove that Ss is an ideal. But how might the set equation Ss = S help in proving that S is a group?
Nothing in the students’ notes says anything about solving equations in semigroups. However, if
t is also an element of S, the above set equation means that there must be an element x of S so
that xs = t. That is, the equation xs = t can always be solved for x. It turns out that one can use the
solvability of this equation in several ways to collect information which, for many students, is of
unknown utility. Nevertheless this information, once collected, can be organized to show the
existence of an identity element and inverses in S. To do this requires both persistence and a
willingness to obtain whatever results, in the form of equations, that one can without knowing
whether those results will ultimately be helpful. (See the proof in the Appendix.)
We could have added two easily proved lemmas to our course notes that would have made
the proof of the above theorem much easier. However, the purpose of the course is to learn to
construct a variety of hard, as well as easy, proofs, and having relevant experiences is important
in developing the students’ ability to do so. We view learning to persist in “exploring”
mathematical situations by obtaining “whatever one can get,” even without knowing its ultimate
usefulness, as an important part of developing students’ proving abilities.
While the proof of the above theorem calls for persistence and exploration, proving in
general can call on a whole “tool box” of knowledge and abilities, such as the use of proof by
contradiction or mathematical induction, or looking for inspiration by proving easier theorems,
perhaps by adding a hypothesis such as finite-dimensional or finite. However, discussion of such
topics is beyond the scope of this paper.

A Mathematician’s Proof Construction

Savic (2012a, 2012b) investigated nine mathematics professors’ proving using tablet PCs
with screen capture software, as well as Livescribe pens and special paper, so that they could
take the devices home and construct proofs in a naturalistic setting (without the time constraints
and influences of an interview setting). All nine mathematicians’ writing and speaking was
recorded with time and date stamps. Several of the mathematicians acknowledged getting
“stuck” on the above Theorem 20: If S is a commutative semigroup with no proper ideals, then S
4
is a group, in a short set of notes containing only the material on semigroups. However, none
gave up, as most students might, but persisted. One mathematician proved it the next day and
another proved it after taking a break for lunch. Later, in a focus group interview, the professors
indicated several ways they have of getting “unstuck” in their own research. These included
getting up and walking around or doing something else for a while, as well as strengthening the
hypotheses in order to prove an easier conjecture. It seems clear these mathematicians took the
construction of the proofs in the semigroups notes as a positive challenge and had a sense of self-
efficacy. Apparently this provided the motivation to persist, a crucial component of their success.
Several of the nine mathematicians volunteered that the material was both accessible and
unfamiliar. However, they were unaware of the origin of the notes, that is, that they came from
the course described above. Perhaps for this reason, several mathematicians attempted to
construct counterexamples to some of the theorems. In attempting to prove Theorem 20, all nine
mathematicians at some point considered “principal” ideals, a concept not in the notes, when
considering the ideal Ss, where s is an element of S. This probably comes from remembering
facts about ideals in rings; however, our students could not have had such memories as the
course notes did not cover rings, and this course is a prerequisite for abstract algebra which
would cover rings. Note that Ss = S is one of two key ideas in proving Theorem 20--ideas
without which it is difficult to make progress. (See the proof in the Appendix.)

Dr. G’s Construction of a proof of Theorem 20

Below we describe most of the work that one of the nine mathematicians, Dr. G, did when
attempting to prove Theorem 20, which he eventually did successfully. Our description is a
synopsis of the transcripts of Dr. G’s speaking and writing while he worked alone using a
Livescribe pen and special paper that recorded his writing and speaking with time and date
stamps. (See Savic, 2012b for the original.)
As seen below, Dr. G took a remarkably meandering path as he explored how to prove
Theorem 20. His various “twists and turns” are indicated in bold typeface. Dr. G started at 7:02
a.m. by considering the statement of Theorem 20, but decided to think about it and have
breakfast. At 8:07 a.m., he returned from a walk and realized that gS (where g is an element of
S) is an ideal, so gS = S. He then thought about inverses and struckthrough his entire
previous argument. At 8:09 a.m., he noted that he needed an identity element which is not
given. At 9:44 a.m., he became suspicious that Theorem 20 might not be true, but noted that he
had few examples which might show that.
At 9:48 a.m., Dr. G started “tossing around” the idea that a [commutative] semigroup with
no proper ideals must have an identity, in which case, he could show it is a group. However, he
didn’t see why S should have an identity. He began to think that translating by a fixed
element [an idea not in the notes] would move every element, which would mean there was no
identity. Consequently, he then began to look for a counterexample. By 9:50 a.m. he neither
saw how to prove Theorem 20 nor how to find a counterexample.
He then looked ahead to Question 22, the final task in the notes, which has three parts that
ask whether certain semigroups are isomorphic. He saw how to answer that and then looked at
Theorem 21: A minimal idea of a commutative semigroup is a group. He thought that he could
probably prove that, so he went back to Theorem 20. By 9:51 a.m. Dr. G recalled that he had
earlier rejected Theorems 3, 9, and 12 of the notes and also did not believe that there are
unique minimal ideals. By 9:53 a.m., he recalled that he had not been told any of the theorems
were false and looked at the non-negative integers under multiplication. He saw that {0} is a

5
minimal ideal and noted that the non-negative integers under multiplication do not form a group.
He thought that this was a counterexample to Theorem 21, but had interpreted Theorem 21
incorrectly – something he later discovered and fixed.
At 9:54 a.m., he started actually answering Question 22. By 9:58 a.m., he had answered its
three parts correctly. At 9:59 a.m. Dr. G. took a break to think about Theorem 20 and at 10:08
a.m. he again attempted a proof of it. This time he saw that for a S, there is e S so that ae =
a and saw that e is “acting like …a right identity on a. Now why does it have to act that way on
[an arbitrary] b?” By 10:12 a.m. he found e' so that be' = b, but that didn’t help since he
couldn’t show that e = e'. Then at 10:13 a.m. he saw that there is an f so that b = af, and then
by 10:14 a.m., he saw that be = afe = aef = af = b. At 10:15 a.m., he saw that e is the identity
element. By 10:18 a.m., he had used a similar technique to show S has inverses and is thus a
group.
Perhaps the most important thing about the above description of Dr. G’s work is what is not
there. There is no evidence that Dr. G thought there was anything wrong with having gone in all
of those unhelpful directions or with having thought that some theorems were false, that he later
discovered were true. What seemed to matter to him was the generation of ideas. If those ideas
resulted in errors, one fixed them and learned from them. He exhibited remarkable persistence
and a willingness to try argument directions that he clearly didn’t know ahead of time would be
helpful, and he altered directions when the need arose. It seems clear that Dr. G had the needed
persistence, which was probably supported by a sense of self-efficacy with respect to his own
mathematical research.

Discussion
We are not the first to have considered the effect of affect and self-efficacy on university
students’ and mathematicians’ proving or problem-solving success. Iannone and Inglis (2010)
found “a positive correlation between [first-year U.K. undergraduate] students’ perception of
their abilities at producing proofs and their actual proof production performance, in line with
general literature about self-efficacy and mathematical ability.” (p. 5) In their study of
mathematicians’ problem solving, Carlson and Bloom (2005) concluded that the
mathematicians’ effectiveness “appeared to stem from their ability to draw on a large reservoir of
well-connected knowledge, heuristics, and facts, as well as their ability to manage their
emotional responses [italics ours].” Also, in a study of non-routine problem solving, McLeod,
Metzger, and Craviotto (1989) found that both experts (research mathematicians) and novices
(undergraduates enrolled in beginning mathematics courses), when given different experience
appropriate problems, reported having similar intense emotional reactions such as frustration,
aggravation, and disappointment, but the experts were better able to control them. This suggests
that the mathematicians in the latter two studies had a mathematical self-efficacy belief that
allowed them to persist.

Teaching Implications
It seems to us that in order to do things that require persistence and exploration, a student is
likely to need to believe that he or she can personally benefit from his or her persistence or
exploration. That calls for a self-efficacy belief, which in turn calls for, perhaps numerous,
successes in what is to be done. We see this as inherently difficult to arrange in traditional
university classes that generally involve students understanding a wide range of concepts and
theorems fairly quickly.

6
 In order to maximize students’ opportunities to experience successes in various aspects of
proving, it would be good to have students constructing their own proofs as early and often as
possible. However, this is often delayed in many transition-to-proof courses by an initial rather
formal treatment of logic. While correct logic is essential for proving, we think its early and
abstract treatment can be replaced by explanations of the relevant logic when needed in the
context of the students’ own work (in a “just-in-time” manner). This is possibly because logic,
beyond what most people already know as “common sense,” actually occurs fairly rarely in
student-constructed proofs (Savic, 2011).
While teaching logic can be integrated into discussing students’ proofs and need not delay
the start of their constructing proofs, there is an aspect of proving that is not usually explicitly
taught and that could be very helpful in facilitating students’early successes. There is a
relationship between the structure of a proof, the logical structure of the theorem being proved,
and the theorems and definitions used in constructing the proof. This relationship appears as part
of the final written proof, but it can be isolated and considered first. One can write it first leaving
blank spaces for the remaining work. Selden and Selden (1995) have referred to this as a proof
framework.
In constructing a proof framework, one writes as much as possible of a proof before
attempting to generate the ideas needed to complete it. One writes the hypotheses first, leaves a
blank space for the body of the argument, and then writes the conclusion. Typically one then
“unpacks” the meaning of the conclusion, and inserts the beginning and end of that part of the
proof into the earlier blank space. One proceeds as far as possible in this way. This exposes the
“real problem” to be solved. After the framework has been constructed, one can then try to
generate the original ideas needed to complete a proof. We feel teaching students to build proof
frameworks allows them to experience early successes. However, occasionally constructing a
proof framework helps only a little in obtaining the final proof; this is the case for Theorem 20
discussed above. For other theorems, such as the theorem that states that the sum of two
continuous real functions is continuous, constructing a proof framework can be very helpful
(Selden & Selden, 2013).

References
Bandura, A. (1995). Self-efficacy in changing societies. Cambridge: Cambridge University Press.
Bandura, A. (1994). Self-efficacy. In V. S. Ramachaudran (Ed.), Encyclopedia of
human behaviour (Vol. 4, pp. 71-81). New York: Academic Press.
Carlson, M. P., & Bloom, I. (2005). The cyclic nature of problem solving: An emergent multi-dimensional problem-
solving framework. Educational Studies in Mathematics, 58, 45-75.
Coppin, C. A., Mahavier, W. T., May, E. L., & Parker, G. E. (2009). The Moore Method: A pathway to learner-
centered instruction (MAA Notes No. 75). Washington, DC: MAA.
de Villiers, M. (2012). An illustration of the explanatory and discovery functions of proof. ICME-12 Regular
Lecture. Available online at: http://www.icme12.org/upload/submission/1893_F.pdf
Iannone, P., & Inglis, M. (2010). Self efficacy and mathematical proof: are undergraduate students good at assessing
their own proof production ability? Proceedings of the 13th Annual Conference on Research in Undergraduate
Mathematics Education. Available online at
http://sigmaa.maa.org/rume/crume2010/Archive/Iannone%20&%20Inglis.pdf.
Krantz, S. G. (2012). A mathematician comes of age. Washington, DC: MAA.
Mahavier, W. S. (1999). What is the Moore Method? PRIMUS, 9, 339-354.
McLeod, D. B., Metzger, W., & Craviotto, C. (1989). Comparing experts’ and mathematics novices’ affective
reactions to mathematical problem solving: An exploratory study. In G. Vergnaud (Ed.), Proceedings of the
13th Conference of the Internationa. Group for the Psychology of Mathematics Education (Vol. 2, pp. 296-
303). Paris: PME.

7
Savic, M. (2012a). What do mathematicians do when they have a proving impasse? Proceedings of the 15th
Conference on Research in Undergraduate Mathematics Education. Available online.
Savic, M. (2012b). Proof and proving: Logic, impasses, and the relationship to problem solving. Unpublished
doctoral dissertation. New Mexico State University, Las Cruces. Available on at www.milossavic.com.
Savic, M. (2011). Where is the logic in student-constructed proofs? In S. Brown, S. Larsen, K. Marrongelle, & M.
Oehrtman (Eds.), Proceedings of the 14th Conference on Research in Undergraduate Mathematics Education
(Vol. 2, pp. 445-456). Portland, OR: SIGMAA/RUME. Available online.
Schoenfeld, A. H. (1985). Mathematical problem solving. San Diego, CA: Academic Press.
Selden, A., McKee, K., & Selden, J. Affect, behavioural schemas, and the proving process. International Journal of
Mathematical Education in Science and Technology, 41(2), 199-215, 2010.
Selden, A., & Selden, J. (in press). The genre of proof. In M. N. Fried & T. Dreyfus (Eds.), Mathematics and
Mathematics Education: Searching for Common Ground. New York: Springer, Advances in Mathematics
Education series, end of 2013. Available as Tennessee Technological University Mathematics Department
Technical Report No. 2013-1, which can be downloaded from http://www.tntech.edu/math/reports/.
Selden, A., & Selden, J. (2013). Proof and problem solving at university level. In M. Santos-Trigo & L. Moreno-
Armella, (Guest Eds.), The Mathematics Enthusiast, Special Issue: International Perspectives on Problem
Solving Research in Mathematics Education (Vol. 10, 1&2, pp. 303-334).
Selden, J., & Selden, A. (1995). Unpacking the logic of mathematical statements. Educational Studies in
Mathematics, 29(2), 123-151.

Appendix

Theorem 20: If S is a commutative semigroup with no proper ideals, then S is a group.

[Comments on the flow of the proof are made in bold typeface, like this.]

Proof: Let S be a commutative semigroup with no proper ideals. Because S is non-empty, there is
an sS.
[First we show Ss is an ideal; so Ss = S.]
Suppose t S(Ss), then t can be represented as t = s'' (s's), where s'', s'S. But now t = (s''s')s, so t
Ss. Thus S(Ss) is contained in Ss, so Ss is an ideal of S. But S has no proper ideals, so Ss = S.
[Next, we note that, for any y, s S, there is xS so that y = xs; this is then used three
times.]
Now, if y  S, then y  Ss. Thus y = xs, where x S. That is, for any elements y and s of S, the
equation y = xs can be solved for x S.
[Next, we show there is an identity element, e.]
Now for u S there is eS so that eu = ue = u. Let wS. Then there is a z  S so that w = zu. So
ew = e(zu) = e(uz) = (eu)z = uz = zu = w. Thus e is an identity for S.
[Finally, we show that every element, p, has an inverse, p'.]
Now let p  S. Then there is p' S so that e = p'p. That is, every element of S has an inverse in S.
Thus, S is a group. QED