ERIC EJ798634: Strengthening Scientific Verbal Behavior: An Experimental Comparison of Progressively Prompted and Unprompted Programmed Instruction and Prose Tutorials

2007·0.33 MB·English

by  ERIC

JOURNALOFAPPLIEDBEHAVIORANALYSIS 2007, 40, 179–184 NUMBER1 (SPRING2007)
STRENGTHENING SCIENTIFIC VERBAL BEHAVIOR:
AN EXPERIMENTAL COMPARISON OF PROGRESSIVELY
PROMPTED AND UNPROMPTED PROGRAMMED INSTRUCTION
AND PROSE TUTORIALS
DARREL R. DAVIS, DARREL E. BOSTOW, AND GUDMUNDUR T. HEIMISSON
UNIVERSITYOFSOUTHFLORIDA
Web-basedsoftwarewas usedtodeliverandrecordtheeffects ofprogrammed instruction that
progressivelyaddedformalpromptsuntilattemptsweresuccessful,programmedinstructionwith
one attempt, and prose tutorials. Error-contingent progressive prompting took significantly
longer than programmed instruction and prose. Both forms of programmed instruction
substantially increased the appropriate use of behavioral vocabulary during subsequent
interpretiveessays.Thesebehavioralgainsextendedtoadifferentsetting,suggesting thatmore
wasbeing learned thansimply howtoanswer programmed tutorial framescorrectly.
DESCRIPTORS: programmed instruction, Web interaction, constructed response, practi-
calapplication, behavioral vocabulary,prompting, computer instruction
_______________________________________________________________________________
Programmed instruction is the sequential may automatically arise even though they have
arrangement of reinforcement contingencies played no role in verbal instructional contin-
that cumulatively establish terminal repertoires gencies. Research that measures such general-
as well as their stimulus control. This term is ization is warranted, because its results may
most frequently associated with teaching verbal justify the extensive effort in (a) integrating
behavior (Skinner, 1957) but may broadly refer instructional learning principles, (b) acquiring
to all forms of instruction that are response a thorough understanding of the content, and
contingent and arranged in a cumulative se- (c) software programming necessary to deliver
quence. Programmed instruction involves care- programmed instruction.
fully crafted response–consequence interactions A recent renewal of interest has grown with
of a complexity that requires an extensive computer technology (e.g., Kritch & Bostow,
understanding of operant reinforcement con- 1998; Miller & Malott, 1997). Kritch and
tingencies, which may partially explain the Bostowconfirmedthatincreasingthedensityof
absence of its wide application in schools, required overt interaction improved the ability
universities, and commerce. oflearnerstoapplyrulestaughtbyprogrammed
The results of any form of instruction may instruction. The present research revisited the
generalize to noninstructional situations and relation between programmed instruction and
the generalization of tutorial performance to
may interlock with practical performances.
subsequent practical application, in this case,
Kritch and Bostow (1998) explored this kind
the nature of interpretive behavioral essays.
of generalization with respect to programmed
instruction of verbal behavior and discovered
that functional relations between repertoires METHOD
Participants and Class Context
We thank John Ferron for his helpful advice with
Onehundredeighteenundergraduatestudents
respecttothe statistical analyses inthis manuscript.
Reprints may be obtained from Darrel E Bostow, participated in the fall semester experiment, and
Department of Psychological and Social Foundations, 107 students participated in a replication in the
College of Education, University of South Florida,
following spring semester. These students came
Tampa,Florida33620(e-mail: [email protected]).
doi:10.1901/jaba.2007.93-05 from two courses with substantially overlapping
179
180 DARREL R. DAVIS et al.
behavioralcontent.Theywererandomlydistrib- end of the frame. Many frames contained an
uted into the experimental groups in the first abstract statement of a behavioral relation (i.e.,
experiment and were stratified by class and then a rule) and then an example. Redundancy
randomly distributed into experimental groups maintained high correct answer strength as the
in the replication. program introduced new examples. Frame step
Prior to the commencement of the first size was refined through reiterative field testing,
experiment, students read and took quizzes on such that the probability of answering correctly
Skinner’s Walden Two (1974), and either wasmaintainedatbetterthan.70perframeand
Skinner’s Beyond Freedom and Dignity (1971) generalization responses were intermittently
or Alberto and Troutman’s Applied Behavior required. In summary, the tutorial was direct
Analysis for Teachers (2003) depending on the and devoid of content not directly related to
class from which they were drawn. In the emission of correct answers, as suggested by
experimental procedures, all learners studied Holland’s (1964, 1967) early work. These
tutorials explaining the role of feelings in specifications define programmed instruction as
behavior analysis. These experiments were an used in the present research.
additional 1-week assignment in the first
semester’s experiment and were the initial Independent Variable (Treatment) Conditions
assignment in the next semester’s replication. Progressive prompting programmed instruction.
Theinstructionalcontentwasbrokenintoeight
Setting, Apparatus, Instructional Materials, sets (lessons), averaging 31 frames per set. In
and Procedure this condition, each frame was a screen pre-
Data collection occurred in a laboratory sentation, and content initially appeared with-
containing 20 computers connected to the out prompt letters (i.e., portions of the answer
Web. The first author supervised the laboratory word) surrounding the blanks. If the user’s
and used a standard set of vocal instructions to answer was correct, the program advanced to
maintain procedural integrity. the next frame. Answering incorrectly resulted
The 6,300-word eight-set tutorial about in seeing ‘‘incorrect’’ and ‘‘try again’’ below the
feelings was formatted into programmed in- content, and then one letter of the answer
struction (frames) and also into prose in appeared next to the blank in the re-presented
a scrolling Web page format. It covered frame. If the next attempt was correct, the
contrasts between respondent and operant program advanced the next frame. If not, an
conditioning, setting conditions, public and additional prompt letter appeared (alternately
private behavior, and the difference between and cumulatively) before or after the blank.
feelings and practical behavior. It explained This progressive prompting continued until
emotionsasby-productsofcontingenciesrather a final incorrect response resulted in the correct
than causes, and how feelings can be modified answer being displayed. An unseen software
by context management. clock timed each student interaction in all
Priming and thematic formal prompting conditions.A constantlypresent ‘‘progressbox’’
(Skinner, 1957) were used during programmed displayed the total number of frames in the set,
instruction. Generalization was induced with stated the current frame’s serial number, and
examples. No more than two or three sentences showed the current frame’s ‘‘contribution
appeared in one frame. Typically, frames remaining.’’ The latter datum began at 100%
contained one blank to which the user with the first attempt for a given frame. An
responded by typing a word or phrase. The incorrect response reduced the frame’s contri-
required answers did not appear elsewhere in bution to the terminal percentage correct for
the frame, and the blanks tended to be near the that tutorial. The reduction was based on the
PROGRAMMED INSTRUCTION 181
percentage of the total letters necessary for that the possibility of a ceiling that could artificially
frame’s correct answer, and fell to zero if the restrict the upper range of possible score
answer was eventually revealed. variation and with it the possible detection of
Traditional programmed instruction with differences between experimental effects.
one attempt. This condition was identical to The same answer words were occasionally
the progressive prompting condition, with the required in different items on the posttest.
exception that (a) the learner viewed a frame Analysis of this 30-item posttest, using the
without letter prompts and (b) the learner was results of the 118 participants from the first
given only one try before the answer was experiment, yielded a Kuder-Richardson in-
revealed. After the correct answer was supplied ternal consistency score of .84. This score
by the learner or viewed as a correction, the described the degree to which individual items
program advanced to the next frame. Also, this correlated with the remaining items and
frame’s ‘‘contribution remaining’’ box was suggested that separate item performances were
absent from the tutorial screen because a given under the control of similar variables.
frame’s contribution was either 100% or 0% Essays. Prior to the commencement of the
depending on the one possible attempt. various treatment conditions and then again
Scrolling prose condition. The same frame immediately following completion of the com-
contentappearedasparagraphsononescrolling puter posttest, the lab manager guided students
Web page without blanks. Software timing to a Web site on which they wrote brief essays
beganwith openingtheWebpage,andclicking in response to the following statements: (a)
on an ‘‘I’m done reading’’ button terminated ‘‘Explain the relationship of your feelings to
the timer. what you do.’’ (b) ‘‘Describe why you feel the
way you do.’’ (c) ‘‘Say what you would do to
Dependent Variables
create a world in which you will feel better.’’
Tutorial percentage correct scores. In the pro- The output from responses to these essays was
gressive prompting and traditional conditions, used to calculate the proportion of the total
individualframeperformancecontributedtoan number of learner-produced words that
accumulating percentage correct score that matched author-selected words. This measure-
became the final performance score at tutorial ment yielded scores that were defined by the
termination. In the progressive prompting authors as the degree of behavioral interpreta-
condition, individual frame contributions were tion present within student essays. Vocabulary
based on the remaining contribution at the gain scores were calculated by subtracting
point at which the student achieved frame preexperimental essay proportions from post-
advancement. experimental essay proportions.
Posttest. A 30-item posttest was delivered in
the same format as the traditional programmed
RESULTS AND DISCUSSION
instruction tutorials. Answers for blanks in the
tutorials were also answers in this posttest. Figure 1 presents fall semester tutorial time
However, text surrounding the posttest blanks score distributions, which indicate how long
was terse and devoid of thematic or formal students spent completing tutorial instruction.
prompting to assess more sensitively the A statistically significant difference among
standing strength of intraverbals (Skinner, groups was found, F(2, 113) 5 158.4, p ,
1957) and to yield scores distributed at the .001. The traditional PI (one-try) group spent
middle of a 0% to 100% range. Weakened three times as much time as did prose readers
intraverbal supplementation of this nature (whileachievinganaveragewithin-tutorialscore
(Skinner) was assumed likely to have reduced of 60%). The progressive prompting group
182 DARREL R. DAVIS et al.
devoted nearly four times as much time as the
prosegroupdidwhileachievingan84%within-
tutorial average during their tutorial sets.
Fall semester posttest results are presented in
Figure 2. Although little difference between the
two programmed instruction groups was ob-
served (Ms 5 54.3% and 53.9%, respectively),
both groups scored significantly (p , .01)
higher than prose readers (M 5 32%).
Figure 3 presents data on essay scores before
and after instruction. Again, a Tukey post hoc
analysis indicated that both programmed in-
struction groups differed significantly (p , .01)
from the prose group but not from each
other (progressive prompting M 5 11.4%,
SD 5 9.6%; traditional M 5 8.5%, SD 5
6.3%; prose M 5 4.2%, SD 5 6.3%).
The experiment was replicated in the follow-
ing spring with the same courses and similar
students; however, this time it occurred before
any course content had been studied. Group
assignment was random as in the fall, but also
stratified; students were distributed propor-
tionately by course. The resulting frequency
distributionsand groupdifferences (notshown)
obtained in this replication were similar to the
fall term.
Pearsonrcorrelationsforbothfallandspring
terms revealed that posttest and essay gain
scores were positively correlated with tutorial
time taken (r 5 .37, p , .01).
As in previous studies, programmed instruc-
tion produced greater learning gains in com-
parison with prose reading but required signif-
icantly more time. The higher within-tutorial
percentage correct rates produced by the pro-
gressive prompting procedure may maintain
students’ interest and help to maintain their
study behavior; however, further investigation
into the relation between within-tutorial scores
and the motivation to continue is warranted.
In summary, the present research confirmed
Figure 1. Frequency distributions of the time it took
the relation between the density of overt interac-
studentstocompletethetutorialtaskinthefallsemester.
Distribution bell curves were calculated using the mean, tions and both posttest and applied performance
standarddeviation, andrange ofeach distribution. that was previously observed by Kritch and
PROGRAMMED INSTRUCTION 183
Figure 3. Frequency distributions of the gain scores
Figure2. Frequencydistributionsofthescoresearned
calculated from subtracting initial essay scores from
on the computer posttest that immediately followed
posttutorial essay scores.
tutorial completion.
184 DARREL R. DAVIS et al.
Bostow (1998). In other words, the more satisfaction;botharepromisingareasforfurther
frequent the response-contingent program re- research.
quirement, the better the resulting performance.
Furthermore, the current results extended the
REFERENCES
generalityoftutorial-inducedchangesinbehavior
Alberto, P. A., & Troutman, A. C. (2003). Applied
to scientific concept usage (essays).
behavior analysis for teachers. Englewood Cliffs, NJ:
Response-contingent programmed instruc-
PrenticeHall.
tionprogramsthatuserigidframestimulioften Holland,J.G.(1964).Responsecontingenciesinteaching
result in substantially different specific frame machineprograms.JournalofProgrammedInstruction,
3,1–8.
performances,terminalperformancescores,and
Holland, J. G. (1967). A quantitative measure for
inclinationtocontinueworking.Theyarerarely programmed instruction. American Educational Re-
a perfect fit for specific learners. In the searchJournal,4,87–101.
experience of the present investigators, within- Kritch, K. M., & Bostow, D. E. (1998). Degrees of
constructed-response interaction in computer-based
tutorial scores correlate with subsequent perfor-
programmed instruction. Journal of Applied Behavior
mance and, incidentally, student complaints. Analysis, 31,387–398.
Further investigation of programs that adjust to Miller, M. L., & Malott, R. W. (1997). The importance
of overt responding in programmed instruction even
learners’ performances should be a fruitful area
with added incentives for learning. Journal of
of research because adjusting feedback could
BehavioralEducation,7, 497–503.
accommodate students with a broader range of Skinner, B. F. (1957). Verbal behavior. New York:
beginning skills, allowing them to achieve Appleton-Century-Crofts.
Skinner, B. F. (1971). Beyond freedom and dignity.
higher within-tutorial scores and potentially
Indianapolis:Hackett.
profit from greater motivation resulting from Skinner, B. F. (1974). Walden two. New York:
higher performance scores. The added prompt- Macmillan.
ingbuiltintoeachframebeforeacorrectanswer
Received July 15,2005
is either produced or shown may relate to
Final acceptance September 6,2006
terminal performance as well as to student Action Editor,Richard G.Smith