Wolach2004 PDF
Wolach2004 PDF
Wolach2004 PDF
ALINA TARLEA
reverse Stroop effect). It is possible that a reverse Stroop effect can be ob-
tained in a numerical Stroop effect situation.
One could predict that participants who count digits in a stimulus item
when the digit in a stimulus matches the digit count for the stimulus item
should count the digits faster than participants who count symbols or letters
of the alphabet. Such a finding would indicate a facilitation effect for match-
ed digits. Facilitation effects have been obtained in experiments on the color
Stroop effect (e.g., Sichel & Chandler, 1969; Dyer, 1973).
The hypotheses for the present study are that (1) Participants will count
symbols and letters of the alphabet in the same amount of time when a
given trial consists of 32 symbol presentations, e.g., -, + + +, ==, :, or 32
presentations of letters of the alphabet (e.g., AAA, BB, CCCC, DDD). (2)
Participants will take longer to read symbols than to read letters of the
alphabet when the same symbols and letters of the alphabet used to test the
first hypothesis are used to test this hypothesis. (3) A numerical Stroop ef-
fect will be obtained as subjects take longer to count mismatched digits, e.g.,
111, 33, 444, 2, than when they count the symbols or letters of the alphabet
in the first two hypotheses. (4) A reverse numerical Stroop effect will be ob-
tained as participants will take longer to read digits when digits are mis-
matched, e.g., 111, 33, 44, 2, than when they read letters of the alphabet. (5)
A facilitation effect will be obtained as participants will take less time to
count matched digits, e.g., 1, 22, 333, 4444, than when they count symbols
or letters of the alphabet. (6) A reverse facilitation effect will be observed as
participants will take less time to read matched digits, e.g., 1, 22, 333, and
4444, than to read letters of the alphabet.
Participants
Four students, two male and two female, from an undergraduate Ex-
perimental Psychology class were participants. Their mean age was 19.5 yr.,
with a SD of 2.4. The students received credit toward their grades for partic-
ipating in the experiment.
Apparatus
Participation required a home computer that ran under Windows 95 or
higher. The participants installed two computer programs, Stroopl and
Stroop2, on their home computers.
Procedure
Participants were instructed to alternate running computer programs
Stroopl and Stroop2 for a total of four sessions with each program. Partici-
pants were required to wait at least one day after running a program before
running the other program. All four participants completed eight sessions
within two weeks.
NUMERICAL STROOP EFFECT 71
of the four conditions appeared at the left of the screen in a column. The
caption on the "Begin" Command Button changed to "End." When the
participant completed counting the characters for all 32 stimulus items, he
clicked on "End." Then the instructions for the next trial appeared on the
screen along with the "Begin" Command Button. In addition to randomiz-
ing the four conditions within each of the three Daily Trial Blocks, the rows
for the 32 stimulus items in a given condition were randomized before the
condition was presented. For example, any of the 32 rows in a condition
could be the first row presented for that condition. All 32 rows for a condi-
tion were presented each time the condition occurred.
The computer recorded the time to count the items in the 32 rows for
each of the 12 daily trials and placed the 12 times in a file. The time for a
trial was the time from when the "Begin" Command Button was clicked un-
til the "End" Command Button was clicked.
When the second computer program (Stroop2) was run, the procedure
was the same as the procedure for computer program Stroopl with one ex-
ception. The three instruction sets for computer program Stroop2 required a
subject to "Read the" characters or digits as opposed to "Count the number
of" characters or digits.
RESULTS
A Randomized Block factorial analysis of variance with four factors of
interest was performed. The first factor, Count or Read, had two levels,
Count Characters in Each Row versus Read Characters in Each Row. The
second factor was Blocks of Daily Trials. The first block included the first
day of training with program Stroopl and the first day of training with pro-
gram Stroop2. The second block included the second day of training with
program Stroopl and the second day of training with program Stroop2, etc.
The third factor was blocks of trials within a daily session, Daily Block 1,
versus Daily Block 2, versus Daily Block 3. The fourth factor was for Condi-
tions 1 through 4 (Symbolic versus Alphabetic Letters versus Matched Digits
versus Mismatched Digits).
All statistically significant results (alpha level of .05) are presented in
this Results section. The Mean Square Error term for all F ratios in the anal-
ysis of variance and all simple main effects tests was 9.25 with 285 dlf. This
error term provides a measure of variability for the analysis of variance. The
values for Partial Eta Squared (PES), the NonCentrality Parameter (NC),
and Observed Power (OP) are provided after each statistically significant F
ratio from the randomized block analysis of variance. Values for Partial Eta
Squared and NonCentrality Parameter provide information about the effect
size. The values for Observed Power provide an estimate of power for a
given F ratio.
NUMERICAL STROOP EFFECT 73
TABLE 1
TIME(SEC.) TO COUNTOR READ STIMULUS
AVERAGE ITEMSAND
STANDARD DEVIATIONS
FOR EACHCONDITION
Condition Count Read
- - - - - - - -
M SD M SD
Symbols 16.0 2.5 22.5 6.3
Alphabet 16.1 2.1 17.9 3.9
Matched Digits 13.6 2.5 16.5 4.5
Mismatched Digits 18.0 1.4 19.8 3.4
Simple main effects tests were performed on each row and each column
of the interaction for Count Characters in Each Row versus Read Characters
in Each Row by Conditions 1 through 4. All six simple main effects tests
were statistically significant. The four simple main effects tests for Count
versus Read were Count (M = 16.0 sec.) versus Read (M = 22.5) Symbols
(F,,*, = 215.03, p < .05); Count (M = 16.1 sec.) versus Read (M= 17.9 sec.) Al-
74 A. H. WOLACH, ET AL.
phabetic Letters (F,,2,,= 16.66, p < .05); Count (M= 13.6 sec.) versus Read
(M= 16.5 sec.) Matched Digits (F,,,,,= 46.05, p < .05); Count (M= 18.0 sec.)
versus Read (M= 19.8 sec,) Mismatched Digits (F,,2,5=17.84, p<.05). The
four statistically significant simple main effects tests in conjunction with the
means in this paragraph indicate that participants took longer to read than
to count in Conditions 1 through 4.
The two simple main effects tests for Conditions 1 through 4 were
Count (Mean Symbols = 16.0, Mean Alphabetic Letters = 16.1, Mean Match-
ed Digits = 13.6, Mean Mismatched Digits = 18.0, F,,,,, = 17.06, p < .05); Read
(Mean Symbols =22.5, Mean Alphabetic Letters = 17.9, Mean Matched Dig-
its = 16.5, Mean Mismatched Digits = 19.8, F,,,5 = 34.27, p < .05).
Tukey HSD procedure was used to make pairwise comparisons for the
two simple main effects tests (Count and Read) for Conditions 1 through 4.
Pairwise comparisons for counting produced the following statistically signif-
icant differences in mean times: Count Symbols (iM=16.0 sec.) versus Count
Matched Digits (M= 13.6 sec., q,,, = 5.65); Count Symbols (M = 16.0 sec.)
versus Count Mismatched Digits (M= 18.0 sec., q,,, = 4.41); Count Letters of
the Alphabet (M= 16.1 sec.) versus Count Matched Digits (M= 13.6 sec.,
q,,, =5.88); Count Letters of the Alphabet (M = 16.1 sec.) versus C o ~ n Mis-
t
matched Digits (M = 18.0 sec., a,, = 4.18); Count Matched Digits (M = 13.6
sec.) versus Count Mismatched Digits (M = 18.0 sec., q,,,= 10.06).
The comparisons with Tukey procedure indicated that participants took
less time to count digits when the Digit and Number of Digits in a row were
the same than in any of the other three conditions (Count Symbols, Count
Alphabetic Letters, and Count Mismatched Digits). This series of findings
indicates that a numerical Stroop effect was found for counting. Participants
took more time to count digits when the digit and number of digits in a row
were not the same when compared with any of the other three conditions.
This finding indicates that the conflict in the digit and number of digits for
a stimulus item increased the time it took to read the digits in a stimulus
item.
Painvise comparisons for reading produced the following statistically
significant differences in mean times: Read Symbols (M= 22.5 sec.) versus
Read Alphabetic Letters (M = 17.9 szc., q,, = 10.34); Read Symbols (M = 22.5
sec.) versus Read Matched Digits (M= 16.5 sec., y,,,= 13.52); Read Symbols
(M=22.5 sec.) versus Read Mismatched Digits (M= 19.8 sec., q2,=6.03);
Read Alphabetic Letters (M = 17.9 sec.) versus Read Mismatched Digits (M =
19.8 sec., q,,, = 4.32); Read Matched Digits (M = 16.5 sec.) versus Read Mis-
matched Digits (M = 19.8 sec., g,,5 = 7.50).
DISCUSSION
A numerical Stroop effect was demonstrated for subjects who counted
characters. That is, when the digit in a stimulus item and the number of dig-
NUMERICAL STROOP EFFECT 75
its in the stimulus item were mismatched, the subjects took longer to count
the mismatched digits than to count letters of the alphabet -or symbols.
When the digit in a stimulus item and the number of digits in the item
matched, the subjects counted the digits faster than counting letters of the
alphabet or symbols (facilitation effect). Sichel, et al. (1969), Dyer (1973),
and Hintzman, Carre, Eskridge, Owens, Shaff, and Sparks (1972) demon-
strated facilitation effects when participants stated the colors of words for
colors printed in compatible colors. This finding is also consistent with find-
ings from the Francolini and Egeth study (1980) that showed facilitation
occurred when participants counted matched digits.
The time to count symbols was comparable to the time to count letters
of the alphabet. This demonstrates that participants had not learned to at-
tach numbers to consecutive letters of the alphabet before the experiment
was performed.
Since participants had not learned to attach numbers to letters of the
alphabet, the alphabet condition was an appropriate control condition for
the reading task. The symbol reading task was not an appropriate control
condition for the reading task because names for symbols were on the aver-
age longer (and sometimes multisyllable words) than words for characters in
the other three conditions (short one-syllable words such as "onc" or "A").
When participants read stimulus items, they took longer to read symbols
than letters of the alphabet. The difference in the time it took to read in the
Symbol and ~ l ~ h a b e tCharacters
ic conditions was probably related to differ-
ences in average length of words for the two conditions.
Reading stimulus items always took longer than counting the same stim-
ulus items. When the digit in a stimulus item and the number of digits in
the item were mismatched, the subjects took longer to read the mismatched
digits than to read letters of the alphabet. This reverse Stroop effect is at
variance with Stroop's failure (1935) to find a reverse Stroop effect. That is,
Stroop found no difference in time to read mismatched items as compared
to reading control items. The Stroop data showed trends in the direction of
the significant finding for reading that was obtained in the present study.
Stroop (1935) performed a third experiment in which he showed a transient
reverse Stroop effect. Stroop found the reverse Stroop effect when subjects
read words for the names of colors printed in mismatched colors after eight
days of practice with naming the colors in which mismatched words for the
names of colors were printed. One can posit that the numerical reverse
Stroop effect in the present experiment was caused by practice with count-
ing stimulus items on alternate days. However, Blocks of Daily Trials was
not statistically significant in the present experiment. In addition, Blocks of
Daily Trials did not interact with any of the other factors in the present ex-
periment. These findings do not suggest that practice led to increased diffi-
76 A. H. WOLACH, ET AL.
culty for reading mismatched digits. It is possible that a reverse Stroop ef-
fect is more likely to occur with numerical tasks than color tasks.
The present study included conditions comparable to all conditions
used in Stroop7sstudy (1935). The present study did not include conditions
that can be compared to the specific conditions used in Windes's study
(1968). Windes7s reading conditions required subject to read (say) the char-
acter that was in a stimulus item. For example, a subject was required to say
2 when he observed 222. The reading conditions in the present study re-
quired a subject to read all characters in a stimulus item. A subject would say
2 three times when he saw 222. Since the comparisons Windes made were
between reading and counting conditions, Windes did not use conditions
comparable to the conditions in the Stroop study or in the present one.
The participants in a given condition were timed across a series of items
in the Stroop study (1935) and the present study. Reaction times were taken
for individual stimulus presentations in the Windes study (1968). The
Stroop study and the present study used a within-subjects design to com-
pare conditions. Windes used a between-subjects design. Windes's mismatch-
ed condition included the following stimulus items: 1, 11, 111, 2, 22, 222, 3 ,
33, and 333. Note that some of the items (1, 22, and 333) were not mis-
matched. The Stroop experiment and the present experiment did not use
matched items in the mismatched condition.
Stroop-like interference has been explained by the allocation of atten-
tion to one of two dimensions, relevant or nonrelevant. Suppose a partici-
pant in an experiment has to count the number of digits in a stimulus item.
Further suppose the stimulus item is 3333. The appropriate response on the
relevant dimension is "four." The inappropriate response on the irrelevant
dimension is "three." A traditional view of the Stroop effect assumed inter-
ference effects occur because of different responses to automatic (direct pro-
cessing) and controlled (indirect processing) processes (Cohen, Dunbar, &
McClelland, 1990). Reading (irrelevant dimension) is presumed to be the au-
tomatic process and counting (relevant dimension) is presumed to be the
controlled process in a numerical Stroop effect experiment.
The Stroop effect may have occurred in the present study because auto-
matic processing (reading digits) interfered with controlled processing (count-
ing digits). The reverse Stroop effect may have occurred because the con-
trolled process (counting digits) interfered with the automatic process (read-
ing digits).
A facilitation effect occurred when matched digits were read or count-
ed. When the controlled and automatic processes were consistent with each
other, the two processes worked together to facilitate performance.
The theoretical model of Cohen, et al. (1990) considers automaticity of
cognitive processes in terms of a continuum based on strength of processing.
NUMERICAL STROOP EFFECT 77