General History of the Study of Verbal Learning

     It is the intent of this rather brief description of the SEPO (for serial position) methodology to communicate the very basic ideas and knowledge that support this new psychological assessment instrumentation as well as some inkling of the research findings that are highly supportive. Dr. Stone began to develop some initial insight that later led to more specific formulations back in the 1970s. He wishes here to formally acknowledge the great help that was received from many of his colleagues, and especially from Dr. John Carman, in the form of comments, new insights, and other matters that were the subject of frequent discussions. First of all, let it be understood what is the purpose of the SEPO test methodology. It was developed for the purpose of detection of possession of special knowledge; more particularly, the possible detection of possession of guilty knowledge. When employed in this latter fashion, the SEPO test then functions in a fashion that justifies it being considered as a form of ‘lie detector.’

     The SEPO test is unusually well-grounded in experimental psychology, especially in the field of human learning, mainly that focused upon verbal learning. The original-foundational research upon which the SEPO test rests can very easily be directly traced back to the research conducted by Ebbinghaus, which he described in a historic monograph in 1885, i.e., his experimental study of memory. Ebbinghaus developed a number of methods for the experimental study of memory and retention. His lead was quite quickly followed by psychologists all over the world who improved upon his methods and added new ones. If there is one area of psychological investigation and concern that would seem to be the exclusive prerogative area of interest to psychology, it is the study of learning. The phenomenon of learning (in animals and humans) has been one of the most major focused concerns of experimental and applied psychology, right up to the present time. "The study of learning" has been referred to by many for well over a hundred years as being the most important foundational aspect of the entire field of psychology.

     With practically all memory experiments on humans, the subject or observer is instructed to learn a prescribed ‘lesson’ and he devotes himself attentively to this task, so that memorizing is a good name for the kinds of learning that goes on in the experimental psychology laboratory. Ebbinghaus was the first to create nonsense syllables as a tool in his study of learning of verbal entities. He set the kind of task in which an individual would have to learn lists of serially presented stimuli, such as nonsense syllables. One of the many matters studied by Ebbinghaus and his followers, for many years, was the phenomenon found to be associated with the varying difficulties in learning stimuli (such as nonsense syllables) located in different parts of a serially presented list of stimuli. In 1902, Ebbinghaus described his finding to date on this kind of matter. He found that not all parts of a list of stimuli (numbers, nonsense syllables, words, etc.) are equally easy to master when the list is learned as a whole. He found that the middle of a serially presented list is learned more slowly than is the beginning or the end. This particular finding was labeled by later investigators as the serial position effect. Later researchers more specifically found that when error-rate or correct-rate is graphed according to the serial order of the stimulus items, a rather marked curvilinear line or function is found. It was also found that such a serial position curve tends to become rather symmetrical when the number of stimuli reaches about 17 to 20. The serial position curve is essentially the same for meaningful words as it is for nonsense syllables, though of course, the total number of errors in learning is less in the former case. [Note - the specific research literature references for many of the research findings/results that are described in this quite brief historical introduction are purposely being omitted, simply for the sake of brevity and to hopefully to reduce the complexity of this presented description for persons not having a strong experimental psychology background.] 

     The serial position effect or phenomenon is the nearly invariable tendency for a subject to learn the first few items most rapidly, the last few a little less rapidly, and the items or stimuli just beyond the middle of the list least rapidly. For almost a century, there have been a large number of theoretical attempts to explain this phenomenon, but none has been regarded as being fully or entirely ‘successful’ in the sense of having wide appeal and theoretical generality. A psychologist, who was perhaps one of the most famous regarding his expertise pertaining to verbal learning, has stated, somewhat facetiously, "Suffice it to say that the person who originates a theory that works out to almost everyone’s satisfaction will be in line for an award in psychology equivalent to the Nobel Prize [Underwood, 1966, p. 491]." With this type of comment regarding the phenomenon of the serial position effect, it can be easily believed that it has been the focus of a great deal of research and theorizing in psychology for the past 100 or so years. It represents one of the most predictable psychological behavioral results given the proper experimental arrangement of conditions for serial learning.

     There are two major formats for serial learning. One is basically the simple serial arrangement of the stimuli to be learned. The order of the items or stimuli is held constant from trial to trial, and the items are counted correct only if the subject recalls them correctly in the same order that they were given. In this fashion each stimulus is not only to be remembered as itself but it also serves to ‘cue’ the subject regarding the next serially presented stimulus. Another format is referred to as the method of paired associates. Items are studied in pairs so that later when the first item of the pair is presented the other can be given in response. Each pair of items may be given in an overall serial order which is held constant from trial to trial, and the second item of the pair is counted correctly only if the subject recalls the second item when presented with the first item in the pair. Generally the task that is set for a subject is to learn all of the second items (in the pairs) in the correct serial order. The serial position effect is an expected result with both these kinds of formats or methods of serial learning of stimuli.

     In the previous paragraphs it has been truly difficult to communicate the full impact of the fact that the phenomenon of serial learning has been one of the most studied matters in all of experimental psychology these past 100 or so years. The serial position effect can also be regarded as one of the most predictable human behaviors known when experimental conditions for studying it have been put into place. Even though research focused upon the serial position effect has taken place for over a century, one still quite regularly sees newly reported research on this phenomenon. In fact, in recent years, it has been a factor that is considered in neuropsychological testing for possible brain damage.

The History of More Specific Research Upon Which the SEPO Test is Based

     If a reader of these words has reached this point in this presentation and has understood the just presented section, then a comprehension of the SEPO test for special/guilty knowledge should not be at all difficult. The SEPO test technique is essentially a straight-forward take-off of the serial position effect concept in serial learning. The serial position effect phenomenon is observed when one uses serial anticipation techniques with almost any type of serially associated (and presented) stimuli for memorizing. A serial position effect curve (when error correct response rate percentage is graphed according to the serial order of the stimuli) is almost symmetrical when the number of serially presented stimuli is about 17 to 20 or higher. A simplified explanation for the serial position curve is that for a set of homogeneous stimuli (numbers, nonsense syllables, words, etc.), the particular stimuli in the middle of a serial list are learned with greater error rates than those stimuli near the ends of the lists. A variety of theoretical explanations or theories exist for this extremely reliable behavior fact. A quite old theory by Thorndike (i.e., presented in 1931 and 1932) calls for concepts such as belongingness and isolation. Stimuli at or near the ends of the list (to be memorized) are seen by Thorndike as being ‘isolated’ from most of the other stimuli as they are seen as ‘belonging’ to the first or last stimulus positions (or near to them). Thorndike pointed out that the principle of belongingness often overrides sheer contiguity as a determiner of association. He believed that contiguity was probably a necessary but not a sufficient condition for the establishment of an association. A slight variation of concept was later introduced by the social learning theorists with respect to matters such as the serial position effect. They employed the term concept, anchored to definable position places to explain whey the first and last serially present stimuli are learned the fastest and easiest. Middle of the list stimuli are not so ‘anchored’ or ‘isolated’, therefore they are the last to be learned. Better isolated or better anchored stimuli have position meaning and as a result are more easily learned.

     In a series of experiments accomplished at the University of Berlin and published between 1933 and 1937, Kohler and his associates found some fascinating results which pertained to organizational versus associative interference in learning serially presented stimuli. In her published first study (i.e., in 1933), von Restorff [one of Kohler’s doctoral students] showed a very clear relationship between perceptual laws and the recall of nonsense syllable. She showed that in the recall of nonsense material, part of the difficulty lies in the homogeneity of the material. If lists of serial stimuli are constructed so that one stimulus is of very heterogeneous material, this one will be retained in memory much better than the stimulus items representing material more frequently repeated within the list. The interpretation is that the heterogeneous items stand out like a figure on the ground, exactly as in visual perception. This favorable structuring in perception turns out to be favorable also for memory recall. This effect has carried the label in the field of psychology of learning for the past 65+ years as the von Restorff effect. Rather interesting, the von Restorff effect is not infrequently mentioned by public media advertising consultants - they report that they employ von Restorff effect type thinking when designing both visual and auditory presented advertising messages.

     To further explain what is meant by the von Restorff effect concept, an example will be employed. If you isolate or anchor a stimulus item in the middle of a serial list by choosing to make it heterogeneous with respect to all of the other stimuli in the list, you then disrupt the serial position effect for acquisition. For example, let a list be composed of three-letter nonsense syllables; the list consists of 19 such stimuli. The serial learning of this list will produce the typical serial position effect curve (almost a symmetrical curve with this number of stimuli). However, in our example, we will replace serial stimulus #10 with the word-name, "Bill." Now, if we require an individual to learn this list we will find a quite dramatic change in the usually expected serial position curve. The ‘cognitive’ continuity of all the rest of the stimuli on the list now serves to isolate or anchor that stimulus term which is heterogeneous with respect to the rest of the list (i.e., it simply does not belong using Thorndike’s terminology).

     Let us take this one step further - now if we can embed near the center of a cognitively ‘neutral’ list of all homogeneous terms (stimuli) a term of guilty knowledge significance (heterogeneous with respect to the remainder of the listed stimuli), an innocent individual will not cognitively isolate the term (stimulus) and the usually expected serial position effect curve will be descriptive of his/her learning of the list. However, if the term (stimulus) has cognitive significance (guilty significance) for the individual, that term (stimulus) will be isolated and be consequently learned much faster. The result is that with the term in place (and which has guilty knowledge significance to the individual), the usual serial position curve will be drastically altered. The place in the curve for the stimulus in question will look to be ‘notched.’ In other words, if error rate percentage is being graphed according to the serial list, then the percentage error rate for learning for the guilty knowledge stimulus will be a good deal lower than it would be if the stimulus was homogeneous with respect to the rest of the stimuli comprising the list.

The SEPO Test

     Basically, the idea expressed in the just previous paragraphs is, in very simple terms, the foundation of the SEPO test. Actually, the SEPO test method for the detection of possession of guilty (or special) knowledge is really quite simple. One places a single guilty or special knowledge verbal stimulus (e.g., a word) into a list of similar such words. For an innocent subject, the guilty/special knowledge stimulus is not unusual or specially significant. In fact, for the innocent subject the guilty/special knowledge stimulus is homogeneous with respect to all of the other stimulus words in the list to be learned. Only for the guilty individual (one holding special knowledge) is the is the targeted stimulus word heterogeneous with respect to all the other stimulus words in the list. When the learning (of the list) results are graphed according to correct or error rate (such as %error), the possible presence of a heterogeneous stimulus in the center of the list can be very easily and dramatically discerned. The test for the possession of guilty/special knowledge, even through innocence is being claimed, is rather straight-forward. It simply is to inspect the learning date to see of a von Restorff effect is operating. If so, then the individual being studied is showing a cognitive learning pattern which would suggest that the stimulus term in question (located in the center of the list) was learned in an unusual fashion; i.e.., it was learned faster and with fewer errors than would expect for a homogeneous stimulus located in the same place.

     Fortunately, since the advent of computers (especially so with now the very wide availability of micro computers), development of a testing presentation platform as well as the ‘scoring’ of the learning of a memorization task involving verbal serial learning, can now be relatively easily implemented. Dr. Stone has developed a computer program that can be used on a (now rather antiquated) computer, i.e., the Apple II series; this program has shown itself to be very flexible and is quite ‘user friendly.’ Verbal or numerical stimuli (up to 50 stimuli) are typed into the computer. Format may either be a simple serial arrangement of the stimuli or they may also be made part of a paired associate stimulus arrangement which then does present the paired associates in a serial order. The inter-trial interval (in seconds) is specifiable as is the stimulus presentation time. The duration (in seconds) of the correct answer display can be set. The number of ‘totally correct trials’ can be specified so as to set the learning completion criterion.

     Dr. Stone has become familiar with these flexible specifiable parameters so that they can be set rather easily for the learning tasks at hand. If a subject individual may have some problems that might affect his learning speed and quality, it is possible that the parameters of the learning task might be adjusted in order that such would facilitate his/her learning the tasks at hand. After a subject has learned a list to the specified criterion level, his/her data can be saved to disk for later analysis or it can be analyzed immediately. The software logic was designed to calculate for each stimulus word (or number) the number of errors as well as the error latencies. Also, the program calculates for each stimulus word the number of correct responses as well as correct response latencies. For each trial the number of errors is counted as well as the error latencies. Also, for each trial the number of correct responses is counted as well as correct response latencies. For sets of trials the number of errors or correct responses can be counted. Also, for sets of trials the error and correct response latencies can be recorded. In the Preliminary Data Analysis part of the output portion of the program, it is possible to calculate a number of descriptive statistics regarding an individual’s learning behavior. For example, one part of the data analysis displays the following information (which may also be printed to hard copy): Total number of trials to criterion, total errors, mean errors, total correct, mean correct, mean error latency, and mean correct response latency. It is possible to the use the program to graphically display the above type information for each stimulus in the series, also according to each trial as well as each set of trials.

     Actually, that which has been just described as the new SEPO test method for the detection of possession of guilty or special knowledge is almost the easiest part. A guilty knowledge type test simply cannot be used in many of the kinds of situations where the polygraph up to now has been used (and it is considered by many as having been used most inappropriately!). To attempt to employ a guilty knowledge type test always will require much more careful preparation and preinvestigation than does a more regular screening-type lie detector or polygraph test. Moreover, even where it is appropriately used, a guilty knowledge type test requires ingenuity and psychological sophistication. It is seen that a cognitively based, guilty knowledge test (such as presented and described in this present work) might have as it greatest benefit the protecting of innocent suspects from prolonged police harassment and even false imprisonment.

     An interesting matter was raised by a well-known psychologist critic of the polygraph lie detector methodology, Lyyken (1974, page 728), regarding the guilty knowledge test idea when using a polygraph. He stated that "since the guilty knowledge method does not require that the item be in the form of questions to which the subject must give answers, that is, since he can be asked merely to repeat the alternative or, even to simply sit and listen., the interesting question arises as to whether requiring a criminal suspect to submit to a guilty knowledge test would involve a violation of his Fifth Amendment rights any more than would requiring him to show his face to a witness or to submit to fingerprinting." The SEPO test method proposed and described here in this presentation also does not require that the subject answer questions, it does merely require him to learn a list of serially presented stimuli (e.g., words, numbers).

Research Results Obtained When Using the SEPO TEST

     Let it very clearly be stated here at the beginning of this section: validity research on the SEPO test has produced a number of very successful and encouraging results, however all such results have been obtained in laboratory type testing situations. The SEPO test has yet to be employed in a real-life criminal investigation setting. With this said, the results of research conducted in laboratory settings has consistently produced results that would strongly suggest that utilization of the methodology in a real-life crime investigation would produce valuable crime-solving results. The SEPO test has been used in at least a dozen or so test situations, all designed to examine whether the methodology can, in fact, validly identify the possession, or lack of possession, of guilty or special knowledge. In every single research investigation, the SEPO test performed in a way so as to suggest that it is a very valid test for the purpose it was designed.

     In a staged theft situation, use of the SEPO test correctly identified the actual thieves and correctly identified those who had not acted-out the contrived theft. Identification of individuals who knew a specific targeted foreign language were correctly identified, even though no question, even naming the foreign language, was ever posed. In another investigation, individuals who were then employees of a highly sensitive governmental intelligence agency were correctly identified as being employees of that specific agency without their ever being asked any questions that might have indicated such. Persons who never had any employment association with the agency in question were identified as persons who had had no association with the agency - this was all totally without ever naming the agency in any way. In fact, in all of these completed research investigations of use of the SEPO test, the involved subjects, at the end of the testing, uniformly had no accurate conscious awareness of the purpose of the verbal learning testing in which they had just participated. In the first mentioned study, the ‘thieves’ had no awareness that they had been caught by the SEPO test. The same would also apply to those subjects who were in the foreign language knowledge testing - at the end of the test they showed no awareness that they had displayed a knowledge of a specific foreign language or that they displayed an ignorance of this particular language. When the SEPO test correctly identified the employees of the intelligence agency, those employees were not at all aware that they had been identified as being employed or otherwise associated with that agency. Without going into any details, let it be understood that the SEPO test has been used in a number of different situations, each designed to detect the presence or absence of a specified bit of knowledge bit. As indicated previously, the referred to special knowledge can be also regarded, in certain type situations, as representing guilty knowledge. Research has been conducted that utilized other classes of stimuli, other than just words (nonsense or those with meaning) in paired-associate type learning arrangements. Photographs, when used as the first series (i.e., the stimulus series) along with nonsense syllables in the second series (that to be learned), have been used successfully in SEPO test trials. There seems to be almost no end to the different types of stimulus combinations that can be perhaps made use of when employing SEPO test methodologies. Please note - in the Preliminary SEPO Test Manual there are full and complete descriptions of the validity research investigations that have been only very briefly ‘hinted at’ in the couple above paragraphs. Individuals wishing further additional technical information regarding these multiple validity investigations are advised to contact Dr. Stone. This information, at the present time, is being somewhat controlled in its distribution so as to satisfy proprietary interests in this testing methodology.

     If one has read this presentation up to this point, it then should be clear to the reader that the SEPO test has some very great advantages over the other types of approaches to matter of lie detection and associated matters. In just about all of the other lie detection procedures that have been suggested, the SEPO test is just about singular in that it can be used without the purpose of the testing becoming evident to the individual being tested. Research results have revealed that individuals who have been fully through SEPO testing procedures, almost always go away without a correct understanding of what they were being tested for. In other words, SEPO testing generally can be regarded as being a ‘hidden purpose’ testing. Just like a well-constructed psychometric lie or dissimulation scale, it too can be believed to possess very little or hopefully no face validity, i.e., it does not look the type of test that it actually is or was designed to be.

     Another great advantage of SEPO testing is that no actual questions have to be asked or statements responded to in some type of true/false manner. But most importantly, the test in no way relies upon physiological emotional response which some believe to be associated with attempts at deception or lying. The major problem with use of the polygraph, voice-stress analysis, or eye response is that the users and believers in these forms of lie detection must assume that there is a direct nexus between an act of deceptiveness and heightened emotional response. Unfortunately for those believers/users of the this class of lie detection methodologies, many individuals can deliver blatant bold-faced prevarication without showing changes in emotion and many other individuals become emotionally reactive even when telling the truth. In recent years, the research results, coming from research conducted by objective and unbiased psychologist investigators, generally tends to show that the true validity of the polygraph methodology, when used to detect lies, is just about as good as flipping a coin to detect deception and truthfulness. The same seems to also apply to the voice stress analysis and eye-reactivity approaches to this matter of detection of attempted deception.

     As indicated earlier in this presentation, the SEPO test is not based upon any kind of physiological measurement of emotional reactivity; instead it is based upon a cognitive learning model. The SEPO test has not been developed or conceptualized as a lie detection device but rather as a test for the possession of special knowledge, a subclass of which may be guilty knowledge. 

     There is no doubt that considerable research must be conducted which is directed towards additional study of the methodology suggested in this presentation. However, it can be argued that investigation of the kind and level just hinted at in this presentation can still be found in the international research literature of the last ten years regarding the polygraph approach to the detection of deception. It appears that the pro-polygraph researchers are still very often simply doing research (which most of the time seems to be conducted very badly) that does nothing more than to try to demonstrate that the methodology seems to work in the simple detection of deception. What is required now with our new cognitive approach to the problem (i.e., the SEPO test) is research directed towards refinement of procedure and method. There would seem to be no doubt that the kinds of investigations that have been only very briefly mentioned in this current presentation will need to be done over and over again simply in order to attract a wide interest in the SEPO testing method and technology.

     The overall method and theory behind the SEPO test comes almost exclusively from the experimental psychology field. Operation of the equipment involved can be easily accomplished by technician level persons, however the creation of appropriate and fruitful stimulus lists would seem to require a high professional knowledge of both clinical and experimental psychology, especially in the somewhat narrow field of human verbal learning. Because of this latter matter it is believed at this point that persons involved in offering use of the SEPO test should all be highly trained psychologists and not some briefly trained law-enforcement type technician such as is found in the field of polygraphy. It is the intention of Dr. Stone to encourage and support the applied use of the described methodology when it is carried out by appropriately trained and experienced doctorate-holding psychologists. There is seen a tremendous potential danger to public interest were the SEPO test methodology to ‘fall into the hands’ of low-level trained technicians as has been the case with polygraph.

Reliability Estimations Regarding the SEPO Test

     When a new psychological assessment technique is presented, one of the very first questions pertaining to it involve reliability of measurement concerns. Reliability of measurement pertains to consistency of measurement; reliability information tell how much confidence one can place in a particular measurement. In other words, reliability is somewhat synonymous with the idea of accuracy. A very good question which can be raised relates to the matter of reliability of measurement for this new SEPO test method for testing for possession of special knowledge. Since reliability can be estimated using a number of different paradigms and formats, an attempt will be made here to comment on reliability estimation for the SEPO test methodology. One approach to the estimation of reliability is by consideration of using equivalent forms of the same test and then to note the degree of consistency of the measurement across the two or more forms. With some tests, two (or more) different parallel forms are built when the test is created and developed. This kind of looking at consistency of measurement can also be approached with splitting a single test into two different ‘split-halves’ and then looking at the consistency of measurement across the two halves. An elaborate model of this type is also based upon the idea that the various items of a single test may each be regarded as a separate alternate form of the test and therefore the consistency of measurement is noted ‘between’ the various items. Something analogous to the idea of parallel tests for the same thing are found in the validity study that involved testing for the presence of knowledge of a specific foreign language. In that study, learning of two separate lists of what appear to be paralogs was accomplished. There was noted to be an absolute perfect identification (when using the first list of paralogs) of which two of the involved four subject individuals actually possessed a knowledge of the targeted language and which two did not. When using the second list of paralogs the very same correct identification of which two knew the foreign language was accomplished. In this particular validity investigation situation there was complete consistency noted with respect to the identification (i.e., the ‘score’ idea) on the two alternate forms (i.e., the two lists) of the same test (that were devised to potentially identify the possession of special knowledge of the particular targeted language).

     Another approach to a tentative estimation of reliability associated with the SEPO tests methodology is to note the extremely high consistency of interpretation by the evaluators of the data collected when employing this testing method. In all of the several different validity studies that have been conducted, there existed almost perfect judgmental agreement (across two, and sometimes three ‘judges’) of the serial position curves data which constitute the dependent variable data and which is judged as to whether a classical serial position curve is present or if some marked deviation to a classical serial position curve is present. What is being judged seems (in the already conducted empirical investigations) to be very clear-cut and is subject to only extremely minimal disagreement. In most of the already indicated validity studies there was observed no real disagreements between the evaluating judges’ decisions regarding the presence or absence of classical serial position curves. This high noted consistency between evaluating judges decisions (based solely upon dependent variable data information) suggests a high reliability level for the judgments regarding the data being judged. In a couple of the validity studies, the interjudge reliability was perfect. In none of the validity investigations, conducted so far has there been one in which some form of measured stability over time can be inferred. In other words, no form of test-retest reliability can be commented on at this time.

     Another way to approach the matter of attempting to estimate reliability of measurement for the SEPO test methodology is to note the very high empirical validity of the procedures when used in each of the several already conducted validity studies. Psychologists who are expert in the design and testing of psychometric instruments are fully aware (and this can be very easily mathematically demonstrated to be true) that an inaccurate test cannot be a good or accurate predictor. There is a rule which clearly states how reliability limits validity. For example, the correlation between a test and an independent criterion can never be higher than the square root of the correlation between two forms of the test. One can look at the SEPO test’s collective validity results in this fashion; if there is a very high or successful level of empirical validity being demonstrated, then there must be a correspondingly high reliability of measurement being present for both the criterion measure as well as for the involved predictive test. In all of the mentioned validity studies there was extremely high predictive accuracy shown; this can be interpreted as suggestive or indicative of a very favorable reliability of measurement situation for the new SEPO test methodology. Without high reliability for the SEPO test, it would have been most unlikely that such high predictive accuracy could have been repeatedly seen.

     There would appear to be little or no need to question the matter of reliability of the serial-position curve effect itself as such has been documented for well over a hundred years by psychologists. As noted earlier in this present attempted description of the SEPO test, there is little in all of psychology that is more predictable, under specified conditions, than is the production of the serial-position curve in serial learning. In the not-long-ago Ebbinghaus Centennial Conference (Gorfein & Hoffman, 1987), it was said by one of the major participants (Murdock, 1987) that "the U-shaped serial position curve that is always obtained in serial learning antedated even Ebbinghaus." The use of the term always by a behavioral scientist (and by this one in particular) indicates that the phenomenon being mentioned must be regarded as an extremely reliable one. Therefore, there is seen absolutely no need at all, for the present writer to have to demonstrate, through new research or through listing an exhaustive survey of the very voluminous published research of the serial position curve phenomenon, that this is indeed a very reliable effect. The serial position curve effect can be regarded as one of the most reliable observed matters in all of scientific psychology.

References

Ebbinghaus, H. (1985) Uber das Gedachtnis: Untersuchungen zur experimentelen Psychologie. Leipzig: Duncker.

Ebbinghaus, H. (1902) Grundzuge der Psychologie. Leipzig: Veit.

Gorein, D. S., & Hoffman, R. R. (Eds.) (1987) Memory and learning: The Ebbinghaus centennial conference. Hillsdale, NJ: Earlbaum.

Lykken, D. T. (1974) Psychology and the lie detector industry. American Psychologist, 29, 725-739.

Murdock, Jr., B. B. (1987) Serial-order effect in a distributed-memory model. In D. S. Gorfein and R. R. Hoffman (Eds.), Memory and learning: The Ebbinghaus centennial conference. Hillsdale, NJ: Earlbaum.

Restorff H. von (1933) Analyse von Vorgangen im Spurenfeld. I. Uber die Wirkung von Bereichsbildungen im Spurenfeld. Psychologie Forschung, 21, 56-112.

Thorndike, E. L. (1931) Human learning. New York: Appleton-Century-Crofts.

Thorndike, E. L. (1932) The fundamentals of learning. New York: Teachers College, Columbia University.

Underwood, B. J. (1966) Experimental psychology. (2nd ed.) New York: Appleton-Century-Crofts.
 
 

Dr. Stone's E-mail address is:  lastone2@earthlink.net