Baire classification of I-density and

I-approximately continuous functions.

Krzysztof Ciesielski, Department of Mathematics, West Virginia University, Mor-

gantown, WV 26506 1
Lee Larson, Department of Mathematics, University of Louisville, Louisville, KY
402922

Abstract
Let T O be the ordinary topology, T I be the I-density topol-
ogy and T D be the deep-I-density topology on the real numbers,
R. Any continuous function f : (R, T I ) → (R, T O ) is a Darboux
function of the first Baire class. Any unilaterally continuous function
f : (R, T D ) → (R, T D ) is a Darboux function of the Baire*1 class.

1. Introduction
The terminology and notation we use is standard. In particular, R denotes
the set of real numbers and N = {1, 2, 3, . . .}. For A, B ⊂ R, Ac stands
for complement of A, A4B for the symmetric difference of A and B and
dist(A, B) for the Euclidean distance between A and B. The natural topology
1
This author was partially supported by West Virginia University Senate Research
Grant.
2
This author was partially supported by a University of Louisville Arts and Sciences
Research grant.

Key Words: I-density topology, fine topologies, Baire classification

AMS Subject Classification: 26A03, 26A21

1
 2

on R is denoted by T O . Topological terms in which we do not specifically

state the topology concern the natural topology. For example, int (A) and
cl (A) stand for the interior and closure of A with respect to T O , respectively.
The oscilation of a function f at point x will be denoted by ω(f, x).
W. Wilczyński [11] introduced a topology on R which has many properties
in common with the ordinary density topology, except that it is based upon
category instead of measure. This topology, called the I-density topology, is
defined as follows.
Let I stand for the ideal of first category subsets of R. A Boolean function
P defined on R is said to be true I-almost everywhere (I-a.e.), if

{x : P (x) is false} ∈ I.

A sequence of functions fn : R → R converges (I) to a function f , if for each

increasing sequence {nj } of natural numbers, there is a further subsequence
{njk } such that fnjk converges pointwise to f , I-a.e. Notice that this defini-
tion mimics the standard definition of convergence in measure, or stochastic
convergence [2], with the exception that the ideal N of Lebesgue null sets is
replaced by I.
If A ⊂ R, then a point a ∈ R is an I-density point of A if χn(A−a)∩(−1,1)
converges (I) to χ(−1,1) , where χS is the characteristic function of the set S.
The set of all I-density points of the set A is written ΦI (A). It is obvious
from the above definition that

ΦI (A) = ΦI (B) whenever A4B ∈ I. (1)

A point a is an I-dispersion point of A, if a ∈ ΦI (Ac ).

Closely related to the notion of an I-density point is that of a deep-I-
density point. An I-density point a of the set A is a deep-I-density point of
A, if there is a closed set F ⊂ int (A) ∪ {a} such that a ∈ ΦI (F ). The set of
all deep-I-density points of A is denoted by ΦD (A). From the definitions, it
is obvious that for any set A ⊂ R,

ΦD (A) ⊂ ΦI (A). (2)

Denote the Baire subsets of R by B. A standard definition of the Baire

sets is the following [6]:

B = {G4I : G ∈ T O and I ∈ I}.

 3

From this definition, it is not hard to show that

 associated
 with every B ∈ B
is a unique open set B̃ such that B̃ = int cl B̃ and B = B̃4I for some
I ∈ I. Any open set G for which G = int (cl (G)) is called a regular open set.
In a sense, B̃ is the largest open set such that B can be written as B̃4I for
some I ∈ I.
The I-density topology [11, 12] is defined as

T I = {A ∈ B : A ⊂ ΦI (A)}.

Similarly, the deep-I-density topology [7, 12] is defined as

T D = {A ∈ B : A ⊂ ΦD (A)}.

It is clear from (2) and the definitions of T I and T D that

T O ⊂ T D ⊂ T I. (3)

It is known that these containments are proper. (The first inclusion is proper
by Lemma 2.4 [12, Theorem 2]. The second inclusion is proper because T D
is completely regular [7] while T I is not [10, Theorem 5].)
Using the three topologies, T O , T I and T D , there are nine different
definitions of continuity possible. For X , Y ∈ {I, D, O}, let

CX Y = {f : (R, T X ) → (R, T Y ) : f is continuous}.

For example, C OO is the set of all functions f : R → R continuous in the

ordinary sense. It is not hard to show that C OI = C OD consists exactly of
the constant functions [4]. It is also known that C DO = C IO , and these are
called the I-approximately continuous functions [7]. The remaining classes
with which this paper is concerned are the I-density continuous and deep-I-
density continuous functions, C II and C DD , respectively.
It is known that the following relationships hold

C OO ⊂ C IO and C II ⊂ C DD ⊂ C DO = C IO . (4)

Moreover, the inclusions in (4) are proper and these are the only inclusions
between those classes [4].
All the continuity and density definitions given above can be restated
in more-or-less obvious ways in one-sided versions. For technical reasons it
 4

is often more convenient to work with one-sided density or continuity. For

example, to show that a point a is an I-density point of a set A, it is often
easier to establish that it is both a left and right I-density point. Such simple
technical extensions to the definitions will be used without further comment.
Let D stand for the functions f : R → R with the Darboux (intermediate
value) property, B1 the functions of the first Baire class and B1∗ the functions
in the Baire*1 class; i.e., the class of all functions f : R → R with the property
that for every perfect set P there is its nonempty portion Q = P ∩ (a, b) such
that f restricted to Q is continuous [8].
In Section 3 it is shown that the one-sided I-density continuous functions
are in D ∩ B1 . Section 4 has as its main purpose a proof that C DD ⊂ D ∩ B1∗ .
But first, the next section is devoted to presenting some technical lemmas
which are needed in the later sections.

2. Some Technical Lemmas

In this section some technical lemmas are presented which are used in the
later sections. The first of these is a restatement of the definition of I-density
points [12].

Lemma 2.1. x ∈ ΦI (A) if, and only if, for every increasing sequence {nm }
of natural numbers there is a subsequence {nmp } such that
 c
[ \   c
(−1, 1) ∩  nmp (A − x) = (−1, 1) ∩ lim inf nmp (A − x) ∈ I.
p→∞
q∈N p≥q

The next lemma is a dual version of Lemma 2.1.

Lemma 2.2. Let B ⊂ R and x ∈ R. The following are equivalent:

(i) x is a I-dispersion point of B;

(ii) for every increasing sequence {nm } of natural numbers there exists a
subsequence {nmp } such that

lim χ(nmp (B−x))∩(−1,1) = 0, I − a.e.;

p→∞
 5

(iii) for every increasing sequence {nm } of natural numbers there exists a
subsequence {nmp } such that
\ [  
(−1, 1) ∩ nmp (B − x) = (−1, 1) ∩ lim sup nmp (B − x) ∈ I.
p→∞
q∈N p≥q

In the definition of I-density given above, the divergent sequence of “mul-

tipliers”, {nk }, is limited to the positive integers, rather than arbitrary se-
quences of positive numbers diverging to infinity. This restriction is removed
by the following theorem.

Theorem 2.3. Let B ∈ B and x ∈ R. The following statements are

equivalent to each other.
(i) x is an I-dispersion point of B;

(ii) for every divergent increasing sequence {tn } of positive numbers there
exists a subsequence {tnk } such that
 
lim sup tnk (B̃ − x) ∩ (−1, 1) ∈ I;
k→∞

(iii) for every divergent increasing sequence {tn } of positive numbers there
exists a subsequence {tnk } such that

lim sup (tnk (B − x)) ∩ (−1, 1) ∈ I.

k→∞

Proof. The equivalence of (iii) and (i) is proved by W. Poreda, E. Wagner-

Bojakowska and W. Wilczyński [10, Corollary 1]. Thus, (ii) is equivalent to
the fact that x is an I-dispersion point of B̃. But the last fact is equivalent
to (i) by (1).

The next lemma and its corollary provide a tool for constructing sets
which are in T I and T D . They are similar to theorems originally proved by
W. Poreda, E. Wagner-Bojakowska and W. Wilczyński [10, Theorem 1] [12,
Theorem 2]. To state them we need first the following definition.
S S
We say that any of the sets n∈N (an , bn ) or n∈N [an , bn ] is a right interval
set of a point a ∈ R if an+1 < bn+1 < an < bn for n ∈ N and limn→∞ an = a.
In the case when a = 0 we simply say that it is a right interval set.
 6

S
Lemma 2.4. If E = n∈N [an , bn ] is a right interval set such that

(i) limn→∞ (bn − an )/an = 0; and

(ii) limn→∞ bn+1 /an = 0,

then 0 is an I-dispersion point of E. In particular, E c ∈ T D .

Proof. It follows immediatelly from the proof of [12, Theorem2].

From the lemma above we obtain easily the following corollary. (Compare
also, [1, Lemma 3].)
S
Corollary 2.5. If n∈N [an , bn ] is a right interval set with

(bn − an )
lim = 0,
n→∞ bn
then there exists an increasing sequence {nm }m∈N of natural numbers such
that 0 is an I-dispersion point of
[
[anm , bnm ].
m∈N

Finally, the following example provides a way to construct functions which

are well-behaved in the ordinary sense, but are not well-behaved with regard
to density continuity.

Example 2.6. There is a monotone continuous function f : R → R which

is not in C II or C DD .
S
Proof. To construct such a function, let E = n∈N [an , bn ] be as in Lemma
2.4 and let D = n∈N [cn , dn ], where [cn , dn ] = [bn+1 , an ] for each n ∈ N. Then,
S

Dc ∈/ T I . It is also easy to see, that by decreasing the intervals [an , bn ], if

necessary, that E c ∈ T D .
Define the function f by letting f (x) = 0 whenever x ≤ 0, f (cn ) = an
and f (dn ) = f (bn ) for all n. Make f piecewise linear between the points on
which it has already been defined. Then f −1 (E c ) = Dc . So, f ∈ / C II ∪ C DD .
 7

3. I-approximately Continuous Functions

In this section it is shown that C IO ⊂ D ∩ B1 . The following lemma is used
in the proof.

Lemma 3.1. If f is right I-approximately continuous at each of its points,

a ∈ R and A = {x : f (x) > a}, then int (A) is dense in A.

Proof. It may be supposed without loss of generality that a = 0. Let

An = {x : f (x) ≥ 1/n} ∈ B and let A = n∈N An . If x ∈ Ãn , then, by
S

Lemma 2.1, x is an I-density point of An , as An 4Ãn ∈ I. The definition of

right I-density continuity shows that f (x) ≥ 1/n. It follows from this that
Ãn ⊂ An . Therefore, G = n∈N Ãn ⊂ n∈N An = A. To see that G is dense
S S

in A, let x ∈ A and choose n ∈ N such that 1/n < f (x). Then x is a right
I-density point of {w : f (w) > 1/n} and it is apparent that x must be a
limit point of Ãn . From this, it follows that G is dense in A.

Theorem 3.2. Every right I-approximately continuous function is of the

first Baire class.

Proof. Poreda, Wagner-Bojakowska and Wilczyński [9] proved a slightly

weaker version of this theorem for two-sided I-approximate continuity. The
following alternative proof is presented here because it is somewhat shorter
and the result is a little sharper.
Let f be right I-density continuous on R. It suffices to show that {x :
f (x) ≥ 0} is a Gδ set. To do this, for each p ∈ N, let Up = {x: f (x) > −1/p}
and, for p, q, r, k ∈ N, define
( ! )
k−1 k
A(p, q, r, k) = x ∈ R: , ∩ r (Up − x) 6= ∅ (5)
q q

and q
\
A(p, q, r) = A(p, q, r, k). (6)
k=1

It is easy to see that each A(p, q, r) is an open set.

Next, define \ \ [
U= A(p, q, r). (7)
p∈N q∈N r≥q
 8

It is clear that U is a Gδ set. It will be shown that U = {x: f (x) ≥ 0}.

To show that U ⊂ {x: f (x) ≥ 0}, fix p ∈ N and let
\ [
Vp = A(p, q, r).
q∈N r≥q

Suppose that x ∈ Vp . For each q ∈ N there is an rq ∈ N, rq ≥ q, such

that when 1 ≤ k ≤ q, then
!
k−1 k
, ∩ rq (Up − x) 6= ∅.
q q
From this and Lemma 3.1 it is apparent that
!
k−1 k
, ∩ rq (int (Up ) − x) 6= ∅ for k = 1, 2, . . . , q. (8)
q q
Let {rqi }i∈N be an increasing subsequence of {rq }p∈N and put ni = rqi for
i ∈ N. From (8) it follows that j∈N nij int (Up ) is a dense open subset of
S

(0, 1) for every subsequence {nij }j∈N of {ni }i∈N . Therefore,

lim sup nij Up ∩ (0, 1)
j→∞

is a residual subset of (0, 1). It follows, by Lemma 2.2, that x is not a

right I-dispersion of Up and the right I-density continuity of f shows that
x ∈ {x: f (x) ≥ −1/p}. Thus, Vp ⊂ {x: f (x) ≥ −1/p} and
\
U= Vp ⊂ {x: f (x) ≥ 0}.
p∈N

To show that {x: f (x) ≥ 0} ⊂ U let us fix x such that f (x) ≥ 0 and
p, q ∈ N. It must be shown that there is an r ∈ N, r ≥ q, such that
x ∈ A(p, q, r). If not, for every r ≥ q there must be an integer kr , with
1 ≤ kr ≤ q, such that
!
kr − 1 kr
, ∩ r (Up − x) = ∅.
q q
There must exist an increasing sequence of natural numbers ri such that
kri = k for some 1 ≤ k ≤ q. This gives
!
k−1 k
, ∩ ri (Up − x) = ∅
q q
 9

for all i so that for any subsequence {rij } of {ri }

!
k−1 k
lim inf rij (Up − x) ∩ , = ∅.
j→∞ q q

Therefore, x is not a point of right I-density of Up . But, this is impossible

because f (x) ≥ 0 and f is right I-approximately continuous at x.
Therefore U ⊃ {x : f (x) ≥ 0} and consequently U = {x : f (x) ≥ 0},
which finishes the proof of the theorem.
The following corollary is immediate.

Corollary 3.3. If f ∈ C IO , then f is continuous in the ordinary sense on a

dense Gδ subset of R.

Corollary 3.4. C IO ⊂ D ∩ B1 .

Proof. Since sets which are open in the I-density topology must be
bilaterally c-dense in themselves, this is an immediate consequence of the
preceding theorem and Young’s criterion. (See Bruckner [3].)

4. I-density Continuous Functions

The goal of this section is to prove that C DD ⊂ D ∩ B1∗ . To do this, the
following definition and lemma are needed [5, Lemma 29.1].
A partition of a set E is a pairwise disjoint family Π = {Ei : i ∈ Λ}
S
such that i∈Λ Ei = E. Note that any partition Π can be associated with a
function F : E → Λ such that F (x) = F (y) if, and only if, x and y belong to
the same Ei ∈ Π. Conversely, any function F : E → Λ determines a partition
of E.
For a set A and n ∈ N define

[A]n = {B ⊂ A: card(B) = n} .

If Π = {Ei : i ∈ Λ} is a partition of [A]n , then a set H ⊂ A is homogeneous for

the partition Π if, for some i ∈ Λ, [H]n ⊂ Ei . That is, all n-element subsets
of H are in the same piece of the partition Π.
 10

Lemma 4.1. (Ramsey’s Theorem) If n, k ∈ N, then every finite partition

Π = {E1 , E2 , . . . , Ek } of [N]n has an infinite homogeneous set. In other
words, for every F : [N]n → {1, 2, . . . , k} there exists an infinite H ⊂ N such
that F is constant on [H]n .
Theorem 4.2. C DD ⊂ D ∩ B1∗ .
Proof. Assume to the contrary that for some perfect set P the set

Z = {x ∈ P : f |P is not continuous at x}
is dense in P .
We will construct sequences: {xn }n∈N of points of P , {(an , bn )}n∈N of
open intervals, {Jn }n∈N of compact intervals, and {In }n∈N of open intervals
having the same midpoint as the corresponding Jn , and contained in that
corresponding Jn . The construction is inductive, and aimed at having all the
objects obtained satisfy the conditions (a) through (f) listed below.
−1 (A) stands for B̃, where B = f −1 (A).
For the reminder of this proof let fg
Start by choosing x0 ∈ Z, (a0 , b0 ) = (x0 − 1, x0 + 1) and I0 = J0 = ∅.
Assume that for all n ∈ N and all i ∈ N, i ≤ n, it holds that:
(a) f (xi ) ∈ Ii ;
(b) Ji−1 ∩ Ji = ∅ and, for i > 2,
1
|Ji | ≤ min{dist(Jk , Jk+1 ): k ∈ N, k < i − 1};
3
(c) |Ji | < ω(f |P , xi ) and 0 < |Ii | < 2−i |Ji |;
(d) xi ∈ (ai , bi ) ∩ Z ⊂ [ai , bi ] ⊂ (ai−1 , bi−1 ) and |bi − ai | < 2−i ;
(e) for every k ∈ N, 2i ≤ k < 4i ,
!
1 g k k+1
 
−1
f (Ii ) − xi ∩ i , i 6= ∅;
bi − x i 4 4

(f ) for every x ∈ [ai , bi ] and every k ∈ N, 2i−1 ≤ k < 4i−1 ,

! !
1 
−1 (I
 k k+1
fg i−1 ) − x ∩ i−1 , i−1 6= ∅.
bi−1 − x 4 4
 11

Let us present the inductive construction. Assume it is done for some

n ≥ 0. We will show the next step. Start with condition (f). If n + 1 = 1,
(f) is void and can be ignored by defining U = R. Otherwise, by (e), the set
( ! )
1  g k k+1
   
Uk = x: f −1 (In ) − x ∩ n , n 6= ∅
bn − x 4 4

contains xn for every k ∈ N, 2n ≤ k < 4n . It is also not difficult to see that

the sets Uk are open. Therefore
\
U= Uk
2n ≤k<4n

is also open and contains xn . It is easy to see that condition (f) is satisfied
for x ∈ U .
Now, find
y ∈ P ∩ f −1 (Jnc ) ∩ ((an , bn ) ∩ U ) .
The existence of such a y is guaranteed because U is open, xn ∈ U and (c).
If y ∈ Z, let xn+1 = y. Otherwise f |P is continuous at y. In this case, the
fact that Z is dense in P and U is open guarantees the existence of

xn+1 ∈ P ∩ f −1 (Jnc ) ∩ ((an , bn ) ∩ U ) ∩ Z.

Since f (xn+1 ) ∈ / Jn and xn+1 ∈ Z, there exists a small interval Jn+1 cen-
tered at f (xn+1 ) satisfying conditions (b) and (c). Choosing In+1 centered
at f (xn+1 ) of length
|Jn+1 |
2n+2
guarantees (a),
 (b), and (c).
Defining a0n+1 , b0n+1 to be centered at xn+1 and such that
h i 1
a0n+1 , b0n+1 ⊂ (an , bn ) ∩ U and b0n+1 − a0n+1 <
2n+1
h i
guarantees (d) and (f) for the interval a0n+1 , b0n+1 . However, it still must
be shown that condition (e) is satisfied. This is done by choosing interval
(an+1 , bn+1 ) ⊂ (a0n+1 , b0n+1 ).
 12

Note that xn+1 is an I-density point of f −1 (In+1 ). Therefore, by Lemma

2.1, there exists an increasing sequence {ni }i∈N of natural numbers such that
the set   
S = lim inf ni fg −1 (I ) − x ∩ (−1, 1)
n+1 n+1
i→∞
is residual in (−1, 1). Define
 
−1 (I
Wi = ni fg n+1 ) − xn+1 .

The set
+∞
[ \
Wi
r=1 i≥r

is residual in (−1, 1). In particular, for every k ∈ N, 2n+1 ≤ k < 4n+1 ,

 
+∞
!
[ \ k+1 k
 Wi  ∩ n+1 , n+1 6= ∅.
r=1 i≥r 4 4
nT o
The sequence i≥r Wi is increasing. Thus, there is an r0 ∈ N such that
r∈N
!
k+1 k
Wi ∩ n+1 , n+1 6= ∅
4 4
for every i ≥ r0 and k ∈ N, 2n+1 ≤ k < 4n+1 . But

−1 (I
 1 
−1 (I

Wi = ni fg n+1 ) − xn+1 = 1 fg n+1 ) − xn+1 .
xn+1 + ni
− xn+1
Define (an+1 , bn+1 ) as
1 1
 
xn+1 − , xn+1 + ,
ni ni
where i ≥ r0 and [an+1 , bn+1 ] ⊂ [a0n+1 , b0n+1 ]. The desired condition (e) is
satisfied. This ends the inductive construction.
It will now be shown how the conclusion of the theorem follows from the
construction.
Let \ \
{x} = [an , bn ] = ([an , bn ] ∩ Z) .
n∈N n∈N
We will show that f is not deep-I-density continuous at x. To be more
specific, we will find a sequence {ni }i∈N such that
 13

S
(1) f (x) is a deep-I-dispersion point of i∈N Ini , and

(2) x is not a deep-I-dispersion point of f −1 (

S
i∈N Ini ).

We will first show x is not an I-dispersion point of f −1 ( i∈N Ini ) for every
S

sequence {ni }i∈N .

Let {ni }i∈N be any increasing sequence of natural numbers. By the defi-
nition of x, condition (f) implies
!
 
−1 (I ) − x
 k k+1
tn fg n ∩ n, n 6= ∅
4 4

for every k ∈ N, 2n ≤ k < 4n , where tn is defined as 1/(bn − x). Note

that sequence {tn }n∈N is increasing and diverging to ∞. Thus,h the
i open
k k+1 1
set Un = tn f −1 (In ) − x intersects every interval 4n , 4n ⊂ 2 , 1 . This
g

implies that for every increasing sequence h{ni }ii∈N of natural numbers and
for every s ∈ N the set i≥s Uni is dense in 21 , 1 . Hence,
S

   
[
−1 
lim sup tnij fg Inij  − x ⊃ lim sup Unij 6∈ I
j→∞ i∈N j→∞

for every subsequence {nij }j∈N of {ni }i∈N . Thus, by Theorem 2.3(ii), x is
not an I-dispersion point of f −1 ( i∈N Ini ).
S

Let us now turn to the proof of condition (1). We must find an increasing
sequence {ni }i∈N of natural numbers such that f (x) is a deep-I-dispersion
S S
point of i∈N Ini . The set i∈N Ini is open. Therefore, it suffices to find a
sequence {ni }i∈N such that f (x) is an I-dispersion point of i∈N Ini . For the
S

sake of simplicity, let us assume that f (x) = 0.

There are two cases to consider.
Case 1o . There exists an increasing sequence {ni }i∈N of natural numbers
such that the Jni are pairwise disjoint.
By taking a subsequence of {ni }i∈N , if necessary, it may be assumed that
[
Jni
i∈N

is either a right or left interval set. For simplicity, assume it is a right interval
set.
 14

Let Jni = [ci , di ] and Ini = (αi , βi ). Then

f (x) = 0 < di−1 < ci < αi < βi < di

for all i. Condition (c) states that

βi − αi |Ini | 1
= < ni .
di − ci |Jni | 2

Let zn be the common center of In and Jn , for n ≥ 0. Then

βi − αi βi − αi βi − αi βi − α i
lim ≤ lim ≤ lim = 2 lim = 0.
i→∞ βi i→∞ zni i→∞ zn − ci i→∞ di − ci
i

By Corollary 2.5 choose a subsequence of {ni }i∈N with the desired properties.
Case 2o . There is no pairwise disjoint subsequence {Jni }i∈N of the se-
quence {Jn }n∈N .
Let us first consider the subsequence {J2n+1 }n∈N , indexed by the odd
numbers, of the sequence {Jn }n∈N . Define a partition function F : [N]2 →
{0, 1} by

F ({n, m}) = 1 if, and only if, J2n+1 ∩ J2m+1 6= ∅.

By Lemma 4.1 (Ramsey’s Theorem) there exists an infinite homogeneous

subset {ni }i∈N of N; i.e., the sequence {ni }i∈N of natural numbers such that
for some k ∈ {0, 1}, F ({ni , nj }) = k for all positive integers i 6= j. But k = 0
would contradict the definition of the case 2o , which is currently considered.
Thus k = 1; i.e.,
J2ni +1 ∩ J2nj +1 6= ∅ (9)
for all nonnegative integers i 6= j.
Now let us repeat the Ramsey-type argument, which was used above, for
the even-numbered counterparts of {J2ni +1 }i∈N . Define G: [N]2 → {0, 1} by

G ({i, j}) = 1 if, and only if, J2ni ∩ J2nj 6= ∅.

By Lemma 4.1 (Ramsey’s Theorem) there exists a subsequence {nis }s∈N of

{ni }i∈N such that
J2nis ∩ J2nit 6= ∅ (10)
 15

for all nonnegative integers s 6= t, while condition (9) is still preserved, or

more precisely
J2nis +1 ∩ J2nit +1 6= ∅ (11)
 
for s 6= t. Define ε = dist J2ni0 , J2ni0 +1 . By (b), ε > 0. Moreover, by (b),
(10) and (11)
ε
[    
B0 = J2nis ⊂ x: dist x, J2ni0 <
s∈N
3

and
ε
[    
B1 = J2nis +1 ⊂ x: dist x, J2ni0 +1 < .
s∈N
3
Hence
ε
dist(B0 , B1 ) ≥ > 0.
3
Note that [
S0 = I2nis ⊂ B0
s≥0

and [
S1 = I2nis +1 ⊂ B1 .
s≥0

Thus dist(S0 , S1 ) > 0, which implies that either

dist(f (x), S0 ) > 0

or
dist(f (x), S1 ) > 0.
This clearly means that f (x) is an I-dispersion point of either S0 or S1 .
This finishes the proof of Theorem 4.2.
Since every function in B1∗ is continuous in the ordinary sense on a dense
open set, the following corollary is obvious.

Corollary 4.3. If f belongs to C DD or C II , then there is a dense T O -open

set G such that f |G is T O -continuous.

Theorem 4.4. The spaces C DD and C II equipped with the topology of uni-
form convergence, are of the first category in themselves.
 16

Proof. This will only be proved for the class C DD as the other case is
essentially the same.
Let {In }n∈N be the sequence of all open intervals with rational endpoints
and let Cn be the family of all deep I-density continuous functions that are
continuous on In in the ordinary sense. By Theorem 4.2, C DD = n∈N Cn .
S

Also, it is evident that the sets Cn are closed in the topology of uniform
convergence. Finally, for any function g ∈ Cn and any of its neighborhoods
U it is easy to slightly modify the function f from Example 2.6 in such a
way that
f ∈ U ∩ (Cn \ C DD ).
Thus, the sets Cn are nowhere dense.

References
[1] V. Aversa and W. Wilczyński. Homeomorphisms preserving I-density
points. Boll. Un. Mat. Ital., B(7)1:275–285, 1987.

[2] Heinz Bauer. Probability Theory and Elements of Measure Theory. Holt,
Rinehart and Winston, Inc., 1972.

[3] A. M. Bruckner. Differentiation of Real Functions. Lecture Notes in

Mathematics 659. Springer-Verlag, 1978.

[4] Krzysztof Ciesielski and Lee Larson. Some examples concerning I-

density continuity. (submitted).

[5] T. Jech. Set Theory. Academic Press, 1978.

[6] K. Kuratowski. Topology, volume 1. Academic Press, 1966.

[7] E. Lazarow. The coarsest topology for I-approximately continuous func-

tions. Comment. Math. Univ. Caroli., 27(4):695–704, 1986.

[8] R. J. O’Malley. Baire*1 Darboux functions. Proc. Amer. Math. Soc.,

60:187–192, 1976.

[9] W. Poreda, E. Wagner-Bojakowska, and W. Wilczyński. A category

analogue of the density topology. Fund. Math., 75:167–173, 1985.
 17

[10] W. Poreda, E. Wagner-Bojakowska, and W. Wilczyński. Remarks on I-

density and I-approximately continuous functions. Comm. Math. Univ.
Carolinae, 26(3):553–563, 1985.

[11] W. Wilczyński. A generalization of the density topology. Real Anal.

Exch., 8(1):16–20, 1982-83.

[12] W. Wilczyński. A category analogue of the density topology, approx-

imate continuity, and the approximate derivative. Real Anal. Exch.,
10:241–265, 1984-85.