A note on stimulus control shaping and one-trial learning in two- and three-year-old children.
Our laboratory is interested primarily in discrimination learning with human subjects who lack well-developed language and problem-solving skills (e.g., young children, adolescents and adults with severe intellectual handicaps). With respect to young children, discrimination research has focused mainly on questions about stimulus preferences and transfer of discrimination training from one problem to another (e.g., transposition) (cf. Stevenson, 1970). Few data are available that bear explicitly on the question of type and amount of training necessary to establish stimulus control on first exposure to a laboratory discrimination training regimen.
The study reported here was conducted originally to compare the effectiveness of different methods for teaching two- and three-year-old children to discriminate circles from ellipses. One method was a well-validated stimulus control shaping program that was developed many years ago by our research group (Sidman & Stoddard, 1966, 1967; see baseline training and ellipse fading program, below). The other methods relied on simple differential reinforcement of the final performance (i.e., trial-and-error procedures). As will be described, the children's unexpectedly rapid learning with the trial-and-error methods rendered the comparison planned originally less interesting. The data, however, bear formal reporting for two reasons. First, they raise important and rarely considered issues concerning discrimination learning in young children who are exposed for the first time to laboratory discrimination training procedures. Second, they suggest the need to take a closer look at variables that may have been underappreciated in our previous studies with the circle-ellipse program.
Fifty-nine typically developing children were brought to the laboratory by their parents in response to an advertisement in an urban university newspaper. The children's ages ranged from 2-0 (years-months) to 3-11.
Apparatus and General Procedures
In a quiet room, approximately 1.6 |m.sup.2~, the child sat before a stimulus display and response panel located at eye level. The panel consisted of a 3 x 3 matrix of 5-|cm.sup.2~ translucent keys, divided by 1.9-cm barriers. On each trial, stimuli were rear-projected onto the keys, with presentation controlled by motor-driven shutters. Only the eight outer keys were used, and a mask covered the center key. A press on a key operated a microswitch which then activated control circuitry in an adjacent room. When the child pressed the correct key, chimes rang and a reinforcer (an M & M candy or a token) was delivered to a tray at the child's left. A 1.5-s intertrial interval followed. Incorrect key presses had no effect, and every trial ended with a reinforced selection (correction procedure). Tokens were exchanged after the session for a toy. Stimulus presentations and the children's key selections (and other relevant events) were recorded on a 20-pen event recorder. (See Stoddard and Sidman, 1967, for complete apparatus descriptions and procedural details.)
Brightness discrimination. In the initial display, one key contained a black outline circle, 3.75 cm in diameter, on a yellow background; the other seven keys were dark. The key location of the circle changed unsystematically from trial to trial. This display continued until the child responded correctly on 9 of 10 consecutive trials with the last 5 correct. The task is essentially a brightness discrimination. Although a circle is present on the bright key, the child does not have to discriminate it; brightness is a sufficient basis for responding correctly.
Form: No form discrimination. In the next series of trials, the incorrect keys became gradually brighter in seven steps until all keys were equally bright yellow. The program gave one trial per step if the child made no errors. If the child did select one or more blank keys before finally selecting the key with the circle on it, the program backed up one step. The task may be characterized as a form:no form discrimination, because the child still need not discriminate the form as a circle.
The next stage examined what effects this baseline training would have on the children's acquisition of a circle-ellipse discrimination: to discriminate the circle from seven relatively flat ellipses (.53 ratio of minor to major axes), equal in width to the circle. Having learned to look for a form as the positive stimulus, would that constitute sufficient preparation that the children would then discriminate that form from others with little or no difficulty? The principal comparison was among test performances of groups that had one of three kinds of experimental history. One group received the ellipse fading program shown previously to be effective in teaching the circle-ellipse discrimination (Sidman & Stoddard, 1966, 1967). One group received the test only, with no fading or other supplementary experience. For the third comparison group, the form:no form baseline was maintained along with the test. All children who did not receive the fading program proceeded to their respective condition after a single correct form:no form trial, when all keys were equally bright.
Ellipse-fading program. This trial series began by introducing faint ellipses onto the blank keys. The ellipses then gradually became more distinct in 10 steps. Again there was one trial per step if all trials were correct. The same backup procedure as in brightness fading was in effect if a trial was incorrect. Nineteen children received this training.
Circle-ellipse test. All trials presented a circle and seven fully distinct ellipses, the terminal display in ellipse fading. The correction procedure and differential reinforcement were used in the test, but the backup procedure was not in effect. Twenty children participated in this condition, referred to as "test-only."
Form:no form baseline and circle-ellipse test. This series also presented circle-ellipse test trials, but each test trial alternated with a form:no form trial. The alteration began with a form:no form trial. If the test produced many errors, the easier, already learned task on every other trial might prevent systematic error patterns from developing (Sidman & Stoddard, 1966, 1967; Stoddard & Sidman, 1967). Other procedures were as in the test-only condition. Twenty children served in this condition.
All children learned the brightness discrimination. Of the 59 children, 50 met the learning criterion for the brightness discrimination within 10 trials, and 8 of the remaining 9 met the criterion within 15 trials. All also learned the form:no form discrimination, 28 with no errors at all, and an additional 16 with no more than 3 errors.
Table 1 Number of Children Who Met Circle-Ellipse Test Criterion after Indicated Number of Test Trials in Each Condition Number of Prior Test Trials Condition (N) 0 1 2 3 4 F(a) 1. Ellipse Fading Program 9 7 - - 1 - 1(b) 10 8 - 1 - 1 - Totals 19 15 0 1 1 1 1 2. Test Only 10 1 4 - 2 - 3 10 3 3 3 1 - - Totals 20 4 7 3 3 0 3 Alternate Form: No Form/Test 9 1 4 - 1 1 2 11 2 6 2 - - 1 Totals 20 3 10 2 1 1 3 Note. For each condition, data from two- and three-year-old children are shown in the upper and lower rows, respectively. a F indicates failure to meet the test criterion within six trials. b Met criterion on the next trial.
Table 1 gives summary data for the three groups. It shows the number of children who met a criterion of two successive test trials after the indicated number of test trials. (The probability of meeting criterion by chance is less than .02.) A child who made no errors at all is counted in the "0" column. A child who made one or more errors on the first trial but then responded correctly on the next two trials is counted in the "1" column, and so on. Any child who failed to meet the criterion within the first six test trials is counted in the "F" column. Totals-rows combine column data for two-(upper row in each condition) and three-year-old children (lower row in each condition).
Of the 19 children in the ellipse-fading program group, 18 rapidly learned the circle-ellipse discrimination. Within the test trials administered after the program, 15 children met criterion without an incorrect circle-ellipse TABULAR DATA OMITTED trial. These results were consistent with the previously reported effectiveness of this stimulus control shaping program (Sidman & Stoddard, 1966, 1967).
A total of 40 children were tested in the test-only and form:no form baseline conditions. Seven children met criterion without a single incorrect circle-ellipse trial. Seventeen more made errors on the first circle-ellipse trial but not on the second and third trials. More than half of the subjects, therefore, met criterion with one or fewer incorrect circle-ellipse trials. Of the remaining children, 10 met criterion and 6 did not.
Table 2 presents individual data for all subjects in each condition. The table shows (a) the total number of trials from all phases prior to the test, (b) the number of those trials that were incorrect, and (c) the number of responses to S- on each of the first six test trials for each subject in each condition. Underlining indicates the trials on which the criterion was met. Note that the median number of training trials prior to the test differs for the program group as compared to the other two groups (26 versus 21 and 18). This difference is attributed to the greater number of trials required to conduct the ellipse-fading for the program group. Could the program group's larger number of training trials (i.e., all types included) account for these subjects' superior performances on the test? A Pearson r correlation coefficient calculated between number of trials prior to the test, and number of test trials prior to criterion, for those subjects who met the criterion, suggested little relationship (r = -.14, p |is less than~ .33). However, correlating the data using a point-biserial correlation between continuous (number of trials prior to the test) and dichotomous (subjects who met criterion with no prior test trials vs. all others who did not) variables yielded a significant correlation (r = -.27, p |is less than~ .02). Though this correlation does not take into account group (condition) membership as a statistical variable, it does suggest that the number of training trials per se may have played a role in the superior test results of the ellipse-fading group.
Data from individual children revealed some remarkable results. One 24-month-old child (Subject TO-1) in the test-only group pressed ellipses four times on the first test trial and never pressed one again in 15 more trials. A 32-month-old child (TO-8) in the same group pressed an ellipse only once, on the first of 17 trials. In the form:no form baseline group, a 27-month-old child (AL-2) did the same thing, pressing an ellipse twice on the first trial and then demonstrating errorless performance thereafter. These data were not unusual, as the findings in Table 2 suggest.
Two-year-old children in the nonprogram groups had somewhat more difficulty than the three-year-olds. Fewer met criterion after zero or one incorrect trial(s) (10 in the combined groups versus 14 three-year-olds). Also, of the six children who failed, five were in the two-year-old range.
The results show that, after mastering the brightness discrimination and given a single trial (test only) or two trials (form:no form baseline) of circle vs. blank key training, many two- and three-year-old children were able to display the circle-ellipse discrimination, either immediately or after a single circle vs. ellipse training trial. The possibility of such rapid discrimination learning in such a high proportion of very young children is rarely reported and virtually never emphasized in a literature that perhaps too often focuses on learning curves. This focus may leave an incorrect impression that gradual learning is generally characteristic of children's initial response to laboratory discrimination training regimens. The present data clearly show that one-trial discrimination learning can be achieved in this population.
Viewing the study in toto, there were four training outcomes that merit separate discussion: (1) children who met the learning criterion with no prior test trials, (2) children who met criterion after only a single test trial, (3) children who met criterion after several training trials, and (4) those who failed to meet criterion. Each outcome will be discussed in turn.
(1) For children who met the learning criterion with no prior test trials, the presence vs. absence training was sufficient by itself to establish control by stimulus features that distinguish circles from ellipses (overall area, ratio of major to minor axis, etc.). These findings appear consistent with previous studies that have demonstrated acquisition of stimulus control following "interdimensional" training in which the positive (S+) and negative (S-) stimuli are from different stimulus dimensions (cf. Honig & Urcuioli, 1981; Mackintosh, 1977). The circle vs. blank key discrimination is such a procedure. Our study makes a contribution by showing that only very brief interdimensional discrimination training may be required to establish reliable control in some subjects. It also shows that interdimensional training can proceed virtually errorlessly under some circumstances.
(2) For children who met the learning criterion after only a single test trial, the brief interdimensional training did not suffice to establish reliable control by the circle-ellipse difference. They met criterion only after a single trial of "intradimensional" discrimination training that contrasted stimuli from the circle-ellipse dimension. For these children perhaps, the preceding interdimensional training may have established control by stimulus features common to both circles and ellipses (e.g., the presence of solid and/or curved lines). A question for further research is whether a higher proportion of subjects would display stimulus control by the circle-ellipse difference(s) if additional interdimensional training (i.e., more circle vs. blank key trials) is provided. The contrast to this possibility is that some subjects may require at least some intradimensional training to establish control by the circle-ellipse difference(s). This raises the possibility that for some young subjects, only the single or few intradimensional discrimination training trials--in the absence of any interdimensional training--would suffice to establish control by the circle-ellipse discrimination. Such a control condition should be incorporated into further research.
(3) For children who met criterion after making errors on more than |Incomplete Text In Original Publication~
|Printer friendly Cite/link Email Feedback|
|Author:||Stoddard, Lawrence T.; Serna, Richard W.; McIlvane, William J.|
|Publication:||The Psychological Record|
|Date:||Mar 22, 1994|
|Previous Article:||Darwinian functionalism: a cognitive science paradigm.|
|Next Article:||Paul Harkai Schiller: an introduction.|