The maze used in the experiments reported in this dissertation is referred to as a “unit-alike” maze. This means that the layouts (“units”) of alleys from the beginning to the first choice-point (junction), and from each choice-point to the next, are all alike. Also the layouts at all choice-points are alike. In the form used, to avoid the blind, a left choice is required at all junctions except the last (referred to as the exceptional choice-point), where the rat must turn right to reach the food box. By dropping in an appropriate wall section, any junction before the last can be made the starting point.
In most of the experiments, the rat was started each day the same number of times from each of 3 entrance points, such that he must run through 2, 3, or 4 units before reaching the exceptional choice-point. The order of using these different entrance points was varied according to a systematic schedule. Each experiment was concerned with a specific sort of floor over which the rat ran. In each experiment, the intent was to continue practice until no rat could show further progress.
The results showed that the level of performance to which the rats could attain depended upon the type of floor used. No floor was found on which there was no evidence of learning. The lowest level found was with a floor of suspended wire screen with the walls carried from a superstructure and not resting on the floor; but a floor made of a 9-inch deep bed of fine sand gave results which were not much higher. The highest level was obtained with a floor of smooth concrete. Floors made of 1.5-inch and 3.5-inches of sand gave intermediate results. Various tests indicated that the cue was not olfactory.
The experiment was designed to show kinesthetic influence even in the situation where the rat is started in varied order but equally often each day from the 3 entrance points 2, 3, and 4 units before the exceptional junction. In the first part of the experiment for 10 days, each rat ran 2 trials per day from each of the above 3 entrance points. In the second part, for a like period, each rat ran the same schedule of trials, except that 2 trials from the entrance point only 1 unit from the exceptional choice-point were interspersed each day among the standard set of runs. In tabulating results, those trials from the 1-unit-distant entrance-point were omitted from the calculations. Such tabulations showed clearly that the added trials in the second part changed the performance on the standard trials to give more errors in the earlier blinds with less errors in the later blinds and less correct runs.
I wish to make grateful acknowledgement of my indebtedness to Dr. John F. Shepard, who suggested this problem and gave unfailing advice and assistance in carrying it out.
…Historical Survey: The Sensory Control of the Maze: The question of the sensory cues used by the rat in maze learning has not yet been settled. That the problem is an important one will be denied by no one who desires the highest possible precision in this valuable instrument for the analysis of animal and human behavior. Many of the discrepancies in the earlier work with the maze performed at different laboratories can be reduced to discrepancies in the degree and kind of sensory control used in the experiments. Certainly a common criticism still offered of a new piece of work with animals is that the control of the maze situation was inadequate.
…Statement of Problem:
…The question arises: Do rats learn kinaesthetically on sand floor mazes because cues from the floor and other parts of the environment are lacking, or simply because such cues are reduced enough so that kinaesthesis is easiest? In other words, can rats be forced to learn the unit-alike maze in terms of local cues, when kinesthetic cues are made unreliable? The present experiment seeks to answer this question; specifically, it proposes:
-
To subject the sand-floor maze to a more rigorous test, by determining whether rats can learn this maze in the absence of reliable kinesthetic cues.
-
To study the influence of various degrees of floor control upon the Learning of the unit-alike maze.
[Experiment 1 replicated 1931 on sand removing floor cues]
…Experiment 2: The 1.5-Inch Bare Sand Floor: Since Experiment 1 showed that rats obtained a cue from the rubber cloth cover of the sand floor maze, a floor of bare sand offered the possibility of controlling the floor cue.
…Experiment 3: The 3.5-Inch Sand Floor: To test the second hypothesis mentioned above, that rats gain a cue from some sub-stratum of the floor which is not affected by disturbance of the surface, a completely new maze was built. A heavy wooden platform supported by steel rollers was constructed, large enough to support the whole maze. A bed of sand 3.5-inches deep covered with a thin black cloth was laid over this platform, and the unit-alike maze was built on top of all. This maze had the advantage of being movable in the maze room to test the effect of general environmental cues as opposed to cues within the maze.
…It is evident from this test that the maze responses are primarily dominated by cues arising from within the maze itself. The influence of cues arising from the sand of the floor was now studied. Table 17 & 18 show the results of 2 tests made more than a week apart, in which the exceptional unit and the one preceding it (C and D) were torn down, the sand beneath them thoroughly stirred and interchanged, and the walls then replaced. On both tests there is a decided disturbance, as indicated by the decrease in right turns at unit D and the increase in these turns in adjacent units.
…The fact that various alterations of the sand floor of the maze rapidly lost their power to disturb the maze-habit led to further experimentation.
…Experiment 4: The 9-Inch Sand Floor Maze: The previous experiments with sand floor mazes seemed to show that: (1) most serious detriment to the maze habit was produced by disturbing the sand floor; (2) the effect of these disturbances rapidly diminished when repeated. To explain this fact, one theory was that the rats began using cues derived from the deep layers of the floor when cues from the surface layers were made unreliable by being frequently altered. To test this theory and to make one last attempt at constructing a sand floor which would not give differential cues, a unit-alike maze was built which had a sand floor 9-inches deep. The bed for this floor was made by elevating the walls of the platform used to support the maze in Experiment 3. The set-up of the maze was identical in every respect with that of the preceding experiment except that the sand floor is nearly 3× as deep.
…Table 20 shows the degree of final mastery of this maze, based on the last 99 trials of each rat. In general, the scores of the rats are considerably reduced over those made on the 3.5-inch sand floor…The average success on the 9-inch sand floor is 53.4% of correct responses to the exceptional unit, while on the 3-inch floor it was 65.2% and on the 1.5-inch floor 66.6%. The decrease in score is about 12%. Moreover, the drop is not due to the total failure of a few rats; every animal made a lower score on the 9-inch sand floor than on the 3-inch floor.
…Experiment 5: The Concrete Floor: A floor of solid concrete was next to be tested. The floor of a small room on the ground level was chosen for the purpose. The concrete was smoothed off with a carborundum brick, and a unit-alike maze was set up from universal sections, directly on this floor.
…All the other rats discriminate the exceptional unit from the standard units with great success, averaging 87.5% correct responses in this unit.
…Experiment 6: The Wire-Mesh Floor: The final type of floor used in this investigation was composed of a strip of ordinary galvanized window screen, suspended at its two ends and stretched tightly across a frame which held it about 3-inches from the floor of the maze room. The screen was thus subjected to constant tension along the long axis of the maze. Beneath the screen, supporting sticks ran cross-wise of the maze, spaced exactly 1 maze-unit apart. The maze walls did not rest on this screen floor, but were suspended about 0.5-inches above it.
…Table 22 shows the success achieved by the rats in their last 99 trials in this maze. Their average score as measured by the number of correct responses in unit D was only 46%, as compared with 53% on the 9-inch sand floor, 65% on the 3.5-inch sand floor, 66% on the 1.5-inch sand floor, and 81% on the rubber cloth floor.
…Discussion: The primary conclusion to be drawn from the results of the experiments reported here is that mazes of unit-alike pattern differ in difficulty according to the type of floor used in their construction, when other factors are held constant. The difference in difficulty between the various floors used in this investigation is shown by Table 23.
…If we take the percentage of perfect runs made on each floor as a criterion, we see that the scores range from 87.29% for the concrete floor maze to 46.3% for the wire screen floor. This means that perfect runs were made almost twice as frequently on the concrete floor as on the wire mesh. The order of difficulty of the various floors according to this criterion is: (1) concrete, (2) rubber-covered sand, (3) 1.5-inch sand, (4) 3.5-inch sand, (5) 9-inch sand, (6) wire-mesh floor.
…However, the problem of the nature of the floor-cue is little advanced by inspection of our data. The fact that the two least difficult floors for the rats, the rubber-covered sand and the concrete floors, were the smoothest, hardest, and least rich in tactile differences to the human, seems evidence that the cue is probably not tactile in nature. Shepard has obtained more conclusive evidence to this same end. He has made the suggestion that the cue may be of resonance character, resulting from vibrations of auditory frequency. These data do not support or oppose the resonance view. It might have been supposed that the wire-screen floor would be rich in such resonances and thus easily learned, which was not the case. On the other hand, the various floors of bare sand might be thought to be lacking in auditory resonance, and to damp resonances from layers below. (This was the opinion which first led to the construction of sand floor mazes.) Some support for the resonance view may be found in the fact that sand floors are difficult to learn in terms of floor cues, and that the difficulty increases with the depth of the sand.
1929 found in his work that alterations of the sub-surface layers of the floor produced more serious deterioration of the unit-alike maze discrimination than changes of the surface floor covering. He has suggested therefore that resonances from below the surface may be used as cues. If this is true for the maze floors used in this investigation, it suggests an explanation for the poor scores made on the wire-screen floor and on the 9-inch sand floor. The very deep sand would be more effective in masking any sort of cue from below than the shallow sand. Likewise, the wire screen floor, suspended as it is from the two ends, would be likely to communicate nothing from below the surface, while the intrinsic vibrations of the screen may have been nearly identical in all parts. In honesty, it must be admitted that such discriminations would require a degree and kind of auditory acuity which is completely incomprehensible to the human. Yet their existence and use in at least one maze situation, that of Shepard, has been demonstrated.
…In further experimentation along these lines, it is suggested that a better means of controlling the factor of kinaesthesis would be to use a maze composed of so many units that the exceptional one could not be located accurately on the basis of kinesthetic cues. This method would avoid any confusion resulting from finding the exceptional unit sometimes earlier, sometimes later in the kinesthetic series, as in our experiment. 1933 used such a maze, and found that a unit-alike maze containing 8 units could not be learned kinaesthetically, because with this number of units the rat cannot make fine enough kinesthetic discriminations to distinguish 7 units from 8.
It is suggested that a positive test of the use of floor cues of an auditory-resonant nature (as distinguished from tactile cues) could be obtained by causing the maze floor to vibrate continuously, at constant or variable rates. Such floor cues could be either masked or emphasized by this means, and the resulting effects upon maze performance analyzed. [Did they ever do this? By Raphelson’s description, the floor-cue maze work would have continued for another 5–10 years before Shepard ceased work & began retiring post-1945. Presumably 1959 would describe any “positive tests” with active sound/vibration generation, if one could get a copy of it…]
…Conclusion: Rats were trained to run a maze composed of a number of units identical in pattern, of which one unit required a reaction opposite to the others. The use of kinesthetic cues was restricted by starting the rats in the maze at 3 different points according to a pre-arranged schedule. 6 types of floor construction were at different times used in this maze situation: a sand floor covered with rubber-sheeting; floors of bare sand having depths of 1.5-inches, 3.5-inches, and 9-inches, a concrete floor, and a floor of suspended wire screen. Tests were made to determine the nature of the sensory cues governing the responses to the various units of the maze. Under these conditions, the following conclusions seem justified:
-
Rats can achieve partial mastery of a unit-alike maze pattern in the absence of reliable kinesthetic cues.
-
Within the limits of this experiment, cues from the floor of the maze are more important than other sensory factors in determining the rats’ responses in the maze.
-
When other factors are held constant, unit-alike mazes vary in difficulty for rats according to the floor on which they are constructed.