The fact that a general stimulus must be used in con- nection with the specific stimuli has already been stated. In using this method it is always necessary to have the apparatus so arranged that the positions of the two specific stimuli may be interchanged at will. This is necessitated experimenter sees to it that in no two successive trials is the same key the one to be operated. He is^ further, able to push back out of sight any number of keys and thus to present the subject with as few as one or as many as twelve. Let us assume that in any given ex- periment the observer decides that the key the fourth from the left shall always be the correct one. It then becomes the task of the sub- ject of the experiment to suit his reactions to the number chosen by the experimenter. Only if he discovers the " guiding idea " of the ex- perimenter can he succeed, trial after trial, in pushing the right key at first. It is obvious that both Yerkes's and Hamilton's methods serve to illustrate general reactive tendencies rather than to analyze reactions minutely and carefully. The methods are intended to bring into clear light those modes of responding to a given situation Avhich are characteristic of different types or conditions of living beings, and thus to furnish a basis for a profitable comparison of reacti^-e tend- encies. (See Hamilton. G. V., "A Study of Trial and Error Re- actions in Mammals," Jour. Animal Beh.. 1911, 1, 33. In this de- scription we have quoted largely from Yerkes's " The Study of Human Behavior." Contribution from Psychopathic Hospital. Boston. Massa- chusetts. No. 25. 1913. This article appears also in Science, 1914, CONTROL BOX 63 by the fact that the animals very easily form position habits, i.e., go repeatedly to the same side regardless of the stimulus to be found there. These so-called position habits are not always easy to detect. The animal often learns to respond to the rhythm of the shift in the position of the stimuli. The experimenter has always to be on his guard against the use of secondary criteria on the part of the animal. In control tests the series of shifts to be made on any given day should be determined by throwing a die. Since the highest number which can be thrown is six, the animal will not be forced to go to the one or the other side too often in succession. It is of the utmost importance for the experi- menter to be out of the sensory range of the animal during all the control tests. An error which occurs by reason of the experimenter's presence might almost be called the ^' Clever Hans error." With slight modification of the above description of an experiment in vision the method may be adopted for use with olfactory, auditory, and cutaneous stimuli.
Control box. — Much time has been spent in devising a box for controlling the animal which, without too much modification, can be used wdth all the animals likely to be used in laboratory w^ork. The box shown in Fig. 1 is the most serviceable one w^e have at the present time for ex- periments upon the sensitivity of animals to light.
The box we employ is 94.5 cm. in length by 74 cm. in width by 25.5 in depth. It is divided into a home compartment, H, 31.5 cm. in length by 25 cm. in width, and a response chamber, C, 40.5 cm. in width, which is partially separated into two smaller compart- ments by a partition. The length of the partition is 25.5 cm. The distance from the door into the response chamber to the stimulus patch is 69 cm. The distance between the stimulus patches is 25 cm. The dimensions of the face of the stimulus patch are 3 cm. by 5 cm. The other divisions of the box are not important. It can readily be seen that it offers two alleys which lead into food compartments, F and Fj. Two doors, D and Di, permit one to confine the animal in the chosen food compartment. By a special mechanism the doors can be closed from without the apparatus by pulling the knobs, K and Kj. Two doors lead from F and Fj into H. It is thus possible t® work with the animal without touching it and without being seen by it during its reaction. By means of a special signaling device the movements of the animal can be recorded without the necessity of watching the animal. Two miniature lamps blackened, except 64 APPARATUS AND METHODS for an opening about 1 mm. in diameter, are connected with the platforms, S, Si and 83. When the animal crosses S and goes into the response chamber both lamps are caused to wink (break circuit). If then he crosses the right hand platform, S2, the right lamp goes out until " choice " is completed and the animal goes into E,. If the animal goes across Sj, the left lamp goes out, etc. In the drawing the stimulus light falls upon the right stimulus patch, X, conse- :Ui Fig. 1, Control Box H, home box; C, response chamber; F, Fj, food compartment; X, Xii stimulus patches; S^ Sg, platforms causing lights L, L, to wink; S, platform causing both lights to wink simultaneouslj^; D, Di, doors leading to food compartment; K, Kj, knobs controlling these doors.
quently food may be obtained by passing around R through the open door Di. It will be noted that the door D is closed. If the animal makes the wrong turn it must pass Sj around into alley L; there finding the door closed, it must retrace its steps and pass S2 and into R through D. The apparatus as a whole can be shifted so that the light can be made to fall either upon X or Xj at will. The box is shown with punishment grill in each compartment.
PAWLOW'S METHOD 65 PAWLOW'S METHOD 65 When made according to the above specifications the box is suitable to carry out experiments upon the following ani- mals: pigeons and other birds of like size, half-grown chicks, young monkeys (small species), guinea pigs, rabbits, rats, and squirrels. It enables one to control the factor of smell; to prevent the animal from returning to the home box; from getting access to the food if an error has been made; and to take a record of the time elapsing between the exit from the home box and a right or wrong choice. These are essential points in every method of physical con- trol. When the box is to be used in connection with the apparatus shown in Fig. 3 and Fig. 6 the end bearing the stimulus patches is replaced by one in which two circular openings, 15 cm. in diameter and 25 cm. between centers, have been cut (the height of the center of these openings above the floor is 10 cm.).
For the auditory and olfactory work slightly different types of control boxes are demanded. Fig. 12 and Fig. 14 show respectively the forms which we have used.
Pawlow's method. — While properly belonging among physiological methods, Pawlow's salivary secretion method has been widely used in animal behavior, by Kussian students especially. This method is used to determine the efficiency of animals' receptors. The possibility of using it depends upon the fact that when certain animals (the dog has been largely used) are stimulated by chemical processes (the specific stimuli for secretion, food, etc.) a secretion of saliva occurs. The chemical property of food, which acts directly upon the receptors in the mouth, thus starting the flow of saliva, is the essential *' property." This property produces an " unconditioned reflex." In addition to the essential property of food there are certain non-essential properties, such, e. g., as its color, brightness, smell, etc., which may, under certain conditions, produce conditioned reflexes.
In actual practice the method is worked as follows: a healthy animal is chosen — one in which salivary secretion is abundant. The duct of one of the salivary glands is ex- posed on the outer surface of the cheek and a salivary 66 APPARATUS AND METHODS B Fig. 2. Pawlow's Experiments The upper left hand corner, X, shows the graduate attached directly to a small funnel fastened over the duct of the salivary gland. The lower part of the drawing shows Nicolai's method. The dog is held in position by comfortable bands. The tube, T, runs directly from the funnel over the salivary gland to the recording apparatus, Y. Careful records are made at R. S shows the position of the stimulus (color, sounds, etc.) and F the food dish. At A, B, and C are shown three sets of records: A, the number of drops, the total amount of the secretion and the temporal relations when food is placed in the mouth; B, when the food is sensed by sight, smell, etc., C, when the color green is shown after the conditioned reflex has been established. (After Nicolai.)
PAWLOW'S METHOD 67 fistula is formed. After this heals a small glass funnel or metal canula is attached to the opening of the gland.
There are three methods of making determinations: (1) As the secretion flows from the tube into a graduate the drops are counted individually and the total flow of saliva read from the scale of the graduate; (2) the saliva is al- lowed to flow into a graduate tube attached directly to the funnel over the duct and the amount of secretion deter- mined by reading the scale of the tube. Under these condi- tions the tube must be replaced after each experiment and carefully cleansed. This method is shown at X in Fig. 2. (3) A metal canula is inserted in the duct of the gland and connected by rubber tubing with the small glass tube. The saliva drops from the tube directly upon the lever of a Marey tambour (Y in Fig. 2). As the drops fall upon this lever a record is made upon a smoked drum, R. From this record the experimenfer may read the quantity of the secretion in drops or the total amount of secretion from the graduate, and also the temporal distribution of the drops. This latter method is the one used by Nicolai.
The animal is next subjected to a course of training. If light response is sought, an " association " between the given light and food is established by exposing the eye of the animal to the light and then immediately feeding him. This light stimulus in time causes a flow of saliva. A con- ditioned reflex has thus been established. Suppose it is now desired to test the animal's sensitivity to differences in wave-length. A green light, e.g., is exposed and food is given. This is repeated constantly until the reflex is established. Control tests are then introduced — white light or some other color is substituted for the green light. If the reflex occurs only (or mainly) when the green is ex- posed and inhibited when the other stimuli appear, we have presumably just ground for assuming sensitivity to wave-length differences.
It is quite clear that Pawlow's method, in theory at least, is designed to give the behavior student the same set of facts as the " discrimination method " (better sensory habit method) now so widely used. As a matter of fact it 68 APPARATUS AND METHODS has nothing like the general range of usefulness of the method first described. In the first place there are only a few animals which can be successfully experimented upon in this way. While the dog lends himself very readily to such a type of experimentation, it is very difficult to see how the method could be worked upon the primates. The monkey, unless severely trussed, would not allow the canula to re- main in place very long. In the case of birds, fish, reptiles, and amphibia, the use of the method is out of the question. Furthermore, the flow of saliva in many of the small ani- mals is not sufficiently great for the method to be used suc- cessfully. In addition to these differences, which are de- pendent upon the anatomical and physiological nature of the animal, there are several inherent difficulties in the method. In the first place, these reflexes tend to disappear after the animal has been subjected to the same stimuli many times. In the second place, these reflexes have not the precision-like character which the students of Pawlow at first maintained. It is quite unusual for secretion to be inhibited entirely by the confusion stimuli. Positive re- sults obtained by this method are certainly of value. It is doubtful if behavior students, in testing for sensitivity to various stimuli, will be satisfied with obtaining negative results.
Methods dependent upon instinctive response. — There is no doubt but that sensory habits arise much more rapidly where the food or other general stimulus is made to convey also the specific stimulus. As an example of the use of this method we cite the experiments in which Shepherd colored cubes of bread with different aniline dyes and soaked the cubes to be responded negatively to with quinine. Modifi- cations of this method appear in the work of Hess and of Katz and Revesz. In the study of the chick, e.g., red-dyed rice grains are fastened to the floor, while the green-dyed grains scattered among these are left free. In determining the limits of sensitivity in the red and in the violet food grains are illuminated by spectral rays from above. The animal is supposed to peck at grains which reflect rays to which its retina is sensitive. In the case of those animals which either collect or disperse when light falls upon them.
CRITICISM OF METHODS 69 we may use a similar method. With fish which collect in the light it is possible to start with rays from the middle region of the spectrum and then gradually to lengthen or shorten the wave-length of the incident beam and thus de- termine the limits of sensitivity in the red and in the violet, as shown by the fact that the animals swim about " aim- lessly " as they do in darkness when the wave-length is such that it no longer offers stimulation (p. 341). In work of this character no food is employed.
Control and auxiliary methods. — In all sensory habits it is necessary to show definitely to what stimulus the ani- mal is responding. In the case of two or multiple stimuli methods the animal may be responding only to one stimulus. If this directive stimulus is removed the habit breaks down. If one or all of the other stimuli are removed no break- down in the habit occurs. In order thoroughly to control sensory habits it becomes constantly necessary to be able to remove or add stimuli. Besides the normal methods of controlling the stimulating factors in the environment, we have another very serviceable one; that of removing the sense organs not employed specifically in the task set the animal. Further discussion of control methods is given on Criticisms on the methods for determining the sensi- tivity of receptors. — General criticisms of the exact meth- ods which have recently been employed have been urged on account of the slownaess with which habits arise. It some- times requires as many as 500 to 600 trials to train an animal to respond positively to the brighter of two lights and negatively to the darker in the apparatus we recom- mend for such work on p. 78. On account of this diffi- culty we have been criticised for setting the animal a task so far out of line with its everyday adjustments. The in- vestigators who criticise the more rigorous and exact methods on this score have had resort to direct food methods. As we have already stated, the habits arise very quickly under such conditions. Unfortunately the more rapid method does not yield results which can be inter- preted. In the case of dyed food we have no large control 70 APPARATUS AND METHODS over the range of wave-lengths and intensities. If the food is directly illuminated we have no control over the relative absorptive powers of the food stuffs for the different wave- lengths. ]\Iany other objections to direct food method may be urged. While we admit that the exact methods tax the time and the patience of the experimenter greatly, yet their use is certainly to be recommended in all cases. It is to be hoped that some means of hastening the speed of the formation of sensory habits will shortly be found. At present the following devices are being tried out: (1) Attempts to increase the stimulating effect of the one or the other stimulus. E.g., we confront the animal with red and green and put a rotating sector in the path of the red. The sector is rotated so slowly that the red is made to flicker. If the animal, on his preliminary trials, is held by the flickering light and tends to seek it, we make red the positive or food color. If, on the other hand, another type of animal is frightened by the flickering red, we make green the positive or food color. Gradually the sector is made to rotate more and more rapidly until the flicker disappears. We are then ready to make our con- trol tests.
(2) Attempts to increase the stimulating effect of one stimulus over the other by making the one differ from the other in several particulars. We may be working upon the animal's ability to respond to differences in the in- tensity of two white lights. Finding that the discrimina- tion arises slowly, we make the two stimuli differ in size and in form as well as in the brightness and then gradually eliminate all differences except that of intensity.
(3) We may use only one stimulus and get the animal to respond either positively or negatively to that. We then very gradually introduce the second stimulus.
III. Apparatus for Obtaining Specific Stimuli Apparatus for obtaining monochromatic light. — The apparatus for obtaining monochromatic light is somewhat complicated and expensive. If there were any other way APPARATUS IN VISION 71 of testing", in a satisfactory way, the color responses of ani- mals, we certainly should not go to the spectrum for our stimuli. Investigators have tried various other methods but with such indifferent success that it seems now all but a waste of time to attempt to use any but spectral light. Without going into details it seems worth while to sketch the apparatus which we have found serviceable and then to take up the technique in presenting these stimuli. This will be done in the fewest possible words and without enter- ing into needless technicalities. In order to obtain spectral bands relatively pure and of the greatest possible intensity, a prism spectrometer should be used. The prism should be large in size. The intensity of the light, too, is depend- ent upon the size and speed of the lenses. We have usually chosen second-hand compound photographic lenses (por- trait lenses) of large aperture (4 inches) and relatively short focus (8-18 inches). We sketch below (Fig. 3) the simplest apparatus which will care for all the color work which is likely to be done in any of the behavior labora- tories.
-n iTRa!
WHITE BEAMl STIMULUS PATCH ^ — -^ ' i TO ANIMAL Fig. 3. Ground Plan of Monochromatic Light Apparatus 72 APPARATUS AND METHODS ( 1 ) Sources: For testing differential sensitivity, limits of spectral sensitivity, etc., where high intensities are needed in control work, the automatic arc light has proved the most serviceable source. In determining thresholds, difference limens, etc., the Nernst filament has been most frequently employed. It is possible now to have made a tubular tungsten source in a nitrogen-filled bulb which has a very much greater intrinsic brilliancy than the Nernst. When these can be more conveniently obtained, they should prove more satisfactory in every way than the Nernst.
(2) Desckiption of Spectrometer: The condensing lens, Cd, Fig. 3, gives an image of the source upon the slit in the collimator, Si. The collimator is of such a size as to give a cylinder of parallel light which approximately fills the face of the prism, P. After refraction the light passes through the objective, Oi, and is brought to a focus upon the (double) slit in the objective, Sj. This slit will admit any two desired regions of the spectrum.
(3) Spacing, Eeversing, and Projecting Devices: Imme- diately behind this slit are situated a series of spacing and reversing, total reflection prisms and three small projection lenses. In the drawing, red and green beams are designated. In order to permit reversing the two beams (no matter from what region they are admitted) must be brought within 8 mm. of each other. This is accomplished by introducing the two small reflection prisms, Mi and Ma. The arrangement of these is clear from the drawing. The red beam, upon reflection from Ma, passes through its projection lens, 34, to the face of the large total reflection prism, TRi, thence doAvnward to the plaster surface immediately below (not shown in the upper drawing but in the auxiliary drawing; the plaster sur- face is marked " stimulus patch " ). The red beam does not change its position, consequently in order to reverse the right-left position of the two beams, the green must be made to appear now to the right of the red, now to the left. This is accomplished very simply as follows: Immediately behind the slit, Sa, one finds the small total reflection prism M3. The green beam is shown entering this. It is reflected and made to pass through its projection lens, 32, then to reflection prism 38 and to reflection prism TRo, and downward to the plaster surface. Prism M3 is mounted upon a small revolving table. When the table is rotated in the direction of the arrow nearest the slit, the green beam is reflected in the opposite direction, through lens 33 to prism 37 to prism TR3, downward to plaster surface.
(4) Regulation of Intensity: One of the most important things in making tests upon the color responses of animals is the means of controlling the intensity of the light. Rotating sectors are extremely accurate and results from them are easy to duplicate in other laboratories. The principal difficulty with them is that the angular opening cannot be altered while the sector is rotating. At least this is true with all but the most expensive sectors, such as the Brodhun. Since two or three sectors are required, the cost of the latter is prohibitive. The Nichol prism is always exceedingly expensive in large sizes, and furthermore, when it is set to give the maximum transmission not more than 50% of the light is admitted. Another disadvantage arises from the fact that it is not possible to APPARATUS IN VISION 73 use the prisms except with parallel light (at least the angle of divergence must be quite small ). We have finally come to the method of using the smoked wedge. These are shown at Wi, Wa, and W3. Each wedge is paired and mounted in such a way that the movement of a single screw will cause both wedges to move synchronously. The beam is not deviated by the introduction of the wedges (they are purchased connected to a strip of clear optical glass as shown in the diagram). The wedges are calibrated in pairs from the point of greatest transmission to the point of least transmission, by the Bureau of Standards, which furnishes a curve of wave-lengths on the one hand plotted against percentage transmission on the other. The wedge enables us to get over one difficulty which we have experienced in our color work, viz., that of getting higher intensities of light than that given by the standard intensity, as measured by the selenium (5) Introduction of White Light: It often happens in monochromatic light work that one desires to introduce a certain amount of white light into monochromatic light or to substi- tute a white light for a monochromatic light. Both ends can be met by the very simple device sketched in the lower part of Fig. 3. A 4-inch strip of fine French plate glass, PG (upper drawing), is introduced at any chosen point in the pathway of the three beams. This plate is inclined at an angle of 45° to the incident beams. The monochromatic lights pass on through the plate glass with relatively little reflection to the plaster of Paris surface. Immediately under- neath the plate glass one places a Nernst filament properly housed and supplied with a slit for controlling int€nsitJ^ By means of a small lens an image of the Nernst can be projected vertically up- ward to the plate-glass surface, PG. The plate-glass surface reflects the beam (about 10%, which is ample), on to the plaster surface. On any of the plaster surfaces one can use at will ( 1 ) a pure mono- chromatic light, (2) a white light, (3) a light obtained by super- posing a monochromatic light of any desired intensity (stated in terms of energy) and a white light of any desired intensity (stated in photometric terms ).
Use of apparatus. — The description so far has been concerned with obtaining suitable stimuli which appear upon the plaster surfaces. In any given test as we have stated only two surfaces are illuminated. As Fig. 3 is drawn the green light falls upon the extreme left surface and the red upon the central one. A suitable control box (Fig. 1) is placed in front of the plaster surfaces. This control box can be shifted to right or left synchronously with the reversal of the right-left relation of the two lights described above. On p. 220 ff. we have described some tests made with this apparatus.
The selenium cell. — One difficulty in the way of using 74 APPARATUS AND METHODS monochromatic light hitherto has been the lack of any con- venient means of obtaining uniformity of stimulation. Ap- parently the only wholly satisfactory method of affording reproducibility of conditions in the various laboratories is to state (1) the wave-length, (2) the energy carried by the monochromatic band, and (3) the dimensions of the stimulus patch. With these constants given no difficulty need be experienced in repeating the work of another in- vestigator. On account of the small energy carried by such bands as are used in behavior work and the nature of the apparatus, great difficulty is experienced in adapting the ordinary methods of determining energy to our needs. The selenium cell, on account of its very great sensibility, seems at present to afford the best solution of our difficul- ties, however complicated and cumbersome its use may be. In order to use the cell its " sensibility " curve must first be obtained. When a cell connected with a suitable source of current is exposed to light its resistance changes. It varies with length of the wave of the incident light. If now a galvanometer is inserted these variations in resistance show as deflections of the needle of the galvanometer. It is obvious that if we allow a given monochromatic band (e.g., red) whose energy is known to fall upon the cell, a given deflection of the galvanometer will appear. This is then noted. The wave-length of the bundle of light next al- lowed to fall upon the cell is shifted, but its energy is kept constant. This again produces a deflection which is noted. This routine is repeated from one end of the spectrum to the other. When completed we are enabled to plot the sensibility curve of the cell in terms of galvanometric de- flections on the one hand, and the wave-length of the incident beam on the other (the energy carried by each bundle of monochromatic rays being constant). The selenium cell has to be thus calibrated in some physical laboratory where a suitable radiation meter is at hand. It may be used though when no stable base is at hand. The drawing and description given below (Fig. 4) give the facts necessary for installation and use of the system.
USE OF SELENIUM CELL 75 Dark Room Ps.|[ Fig. 4. Ground Plan of Selenium Cell Apparatus