Authors Note  I did not perform these four experiments.  This paper was completed as a summary of one experiment from each of four related articles involving human learning and memory.  The purpose was to document self-teaching performed by students in the class.  Every idea within each experiment is NOT documented as if I was quoting the source.  They are referenced only once each (in the introduction) and then further used as if the data was our own.  However the introduction, conclusion, and tying together of experiments is all my own work.  References other then those four experiments used our properly referenced.

Abstract
Three experiments have been carried out in order to investigate the underlying cognitive components of the typicality effect of face recognition.  Experiments 1 and 2 provide evidence for the roles of face familiarity and memorability in recognition.  Experiment 3 suggests that there is two encoding processes that occur when a subject is presented with a face: distinctive and relational processing.

Educational psychologists welcome any relevant research that will build upon their advancing knowledge of human learning processes. Through such information they may endeavor to increase teaching effectiveness and student progress.

Neural Psychologists, Neurologists and Psychopathologists can use the information to understand better exactly what cognitive processes have been disrupted with lesions of certain cortical association areas. For example, understanding the facial recognition process would advance their understanding of prosopagnosia, a disorganization of the sensory world that prevents the recognition of familiar faces (Gleitman, 1995).

The ability to recognize faces is also of interest beyond the field of psychology. In politics and in the entertainment world a persons' ability to be easily recognized may be key to their success. In police work and in law a persons' ability to recognize faces is extremely important. Any understanding of the reasons for inaccuracy in eyewitness identification is very important.

Distinctiveness of an events attributes are essential in distinguishing instances of the same theme. This effect of typicality has been reported in a variety of memory tasks, including those concerning scenes, scripts, behavioral information, and real-world objects (Mäntylä, 1997). A considerable amount of scientific research has demonstrated that, atypical faces are better recognized then typical ones (Vokey & Read, 1992). Vokey and Read (1992) state that this typicality effect, the idea that atypical faces are more easily recognized then typical ones, leads to a lower rate of false recognition responses to atypical faces, and that the time required to identify such faces is less then that required for faces identified as typical.

Developmental psychologists have realized the role of typicality in face recognition for more than twenty years (Craig & Kermis, 1994). Fagan (1977) found that 5-month-old infants recognized photographs of faces after two weeks of their presentation. He also discovered that these infants had difficulty recognizing the initial faces after being shown a similar face. This is an early example of the typicality effects of face recognition.

The idea that the typicality effect is the factor behind face recognition, has long been supported and is not in question. However, the critical cognitive mechanisms that underlie typicality are still being studies today. This review considers three such attempts to determine the nature of the underlying mechanisms. Experiment one evaluates the familiarity hypothesis of typicality by investigating the depth and elaboration of processing faces (Bartlet, Hurry, & Thorley, 1984). The second experiment suggests that the effect of typicality cannot be explained by familiarity alone. That in fact, typicality processing is the result of two competing components, familiarity and memorability (Vokey & Read, 1992). Experiment three investigates the encoding strategies through which information that is processed by typicality is obtained (Mäntylä, 1997). Experiment 1 Typicality as Familiarity

Unusual (atypical) faces receive deeper, more elaborate, processing then typical ones (Light, Kayra-Stuart, & Hollander, 1979). This lengthy and elaborate processing task increases the probability that distinct facial features will be scanned, and thereby increases the perceived recognition of the face (Bartlett, Hurry, & Thorley, 1984).

In Experiment 1 the familiarity hypothesis was tested. The familiarity hypothesis is a three-part model, according to which (1) new faces have nearly no perceived familiarity, but, new faces judged typical are perceived familiar more often via false recognition, (2) faces previously presented are perceived as more familiar then novel faces, and (3) when both typical and atypical faces are previously presented the atypical faces are more likely to be perceived as familiar (Bartlett, Hurry, & Thorley, 1984).

To evaluate the familiarity hypothesis of typicality encoding, typical and atypical faces, the independent variables, were examined by subjects who were later tested on their recognition of these faces, the dependent variable. Bartlett et al. (1984) hypothesized that the incidence of false recognition should be lower with atypical lures (faces not shown in the familiarization procedures) then with typical lures.

Methods
Participants
Ninety-six undergraduates at the university of Texas at Dallas participated in the main experiment as a means of fulfilling course requirements. The populations average age was approximately 30 years. Twelve additional, unidentified, subjects were used as judges in defining certain aspects of the apparatus.
Apparatus
From a 1965 Texas A & M yearbook 96 to-be-remembered faces and eight filler faces were selected. All of the faces were head-and-shoulder photographs of young Caucasian men, each wearing dark suits, white shirts, and ties. None of the men had beards, mustaches, or glasses. All of them had close-cropped hair. Based on a single investigators intuitions, an attempt was made to find and equal number of typical and atypical faces. The pictures were photographed into 35 mm black-and-white transparencies.

Typicality Ratings. A group of five judges rated each transparencies typicality on a 7-point scale (1 = highly typical). To determine the consistency among the judges, the averaged typicality ratings of the first two judges were correlated against the average ratings of the remaining judges, at r = .80 the experimenters considered the judges typicality ratings to be significantly related. The faces were divided into two groups at the median. The 48-item typical face group had a typicality rating of, M = 2.45, SD = .86. The atypical group also with 48 faces, had a typicality rating of, M = 4.79, SD = .86.

Familiarity Ratings. The faces were all presented to a second group of judges (n = 7), who were to rate each face on a 6-point scale concerning its familiarity (1 = sure unfamiliar). For the purposes of this task, the judges were told that a familiar face would be one they believe they have seen before (e.g., on a city street). The faces judged typical before were perceived as being more familiar then the atypical ones.

Presentation and test lists. From the 96 to-be-remembered faces 24 of the typical and 24 of the atypical faces were randomly assigned into each of the two 48-item input lists. To the 48-item input lists three filler faces were added to the beginning and the end of the list. The recognition test list included all 96 faces, with the 48 faces not of the input list acting as the false recognition lures. This list began with eight "warm-up" faces, six of which were previously seen ("warm-up targets") and two of which were new ("warm-up lures").

Design
A multi-variate between-subjects manipulation of familiarization was used to examine the interactions and effects of typicality and familiarization. This experiment used multiple independent variables, all of which acted on the same dependent variable (recognition accuracy). All subjects were presented with within-list independent variables. These variables were the typicality of the face and the item type (originally presented faces vs. false recognition lures).

Recognition memory was assessed under three different test conditions (n = 32, per condition), together these conditions act as a second set of independent variables. The first was the standard condition in which subjects were simply shown the 48-item input list and later shown the full 96-item list and tested on recognition. A second subject group was tested under a prefamiliarization condition, where all 96 faces were quickly shown to the subject prior to the input list. The third condition was the postfamiliarization group, like the prefamiliarization groups, the input list and complete 96-item list was shown prior to testing. However, the complete list was this time shown after the input list. The pre- versus postfamiliarization independent variables were used for the purpose of establishing generality, it was expected that they should eliminate the false-recognition-rate advantage of atypical lures.

A third set of variables, an input task, was used for exploratory purposes only. This set of variables was irrelevant to the purpose of this experiment, and was therefor subsequently dropped from this review. These variables were held constant across all of the test conditions and are therefor were not acting as confounding agents.

Procedure
Subjects in the standard condition group were seated approximately 3 ft away from a white wall and told that a list of faces would be presented upon this wall. No mention of a forthcoming recognition test was made at this time. The input face list was then presented via slide projector at 2 s per face. An inter-stimulus interval of approximately 15 s was used.

Before the recognition test, subjects were informed that they would be presented with a new list of faces and that they were to judge whether the faces were on the previously seen input list. Subjects responses were made by pressing one of two telegraph keys, which were wired to a Lafayette Reaction Time Control apparatus. Subjects also stated their confidence in their choices on a 3-point scale. Each face in the recognition test was shown for five seconds, followed by a one second blank interval. At the end of the inter-stimulus period subjects were prompted for readiness for the next slide.

Procedure for the pre- and postfamiliarization conditions were identical, except for the presentation of the familiarization set before or after the input list. The subjects were simply told to examine each picture thoroughly. Each slide of the pre- and postfamiliarization phase was presented for 2 s, with no interval between slides.

Results and Discussion
Probabilities of recognition judgements to typical and atypical targets and lures were subjected to analysis of variance across all of the conditions. Two sets of these findings supported the familiarity hypothesis as a mechanism for typicality effects in face recognition. First, familiarization had negative effects on recognition. Clearly, there was an increase in false recognition of the newly presented faces that had been previously judged as familiar. The false recognition rates of new but atypical, unfamiliar, faces were not consistent. Secondly, the correct identification of previously seen faces was significantly more likely for previously seen atypical faces then for typical ones.

Though both pre- and postfamiliarization effects matched the qualitative predictions of the author's hypothesis, its effects were not completely consistent. Where as prefamiliarization eliminated any advantage that atypical lures created, postfamiliarization appeared too only reduce the effect. Reaction time data was not significantly informative to the purpose of this experimental review, and was therefor dropped from this discussion.

At theoretical levels, this studies most important implications is that familiarity information is operative in false recognition and is susceptible to effects of face typicality. The data suggests that the perceived familiarity of never before seen faces is increased if they are typical and that perceived familiarity decreases with the extent to which the face is considered atypical. Unfortunately these experimental results do not justify the conclusion that familiarity is the lone method of encoding that accounts for face recognition. Due to the fact that familiarity seems actually to decrease discrimination of faces, one could conclude that if this were the only encoding process used in face recognition we would likely not be able to remember the faces of a significant portion of the population. Apparently, some other sort of encoding process must also be occurring along with familiarity, this idea is investigated in Experiment 2. Experiment 2 Typicality as Familiarity and Memorability

To explain typicality's effects on face recognition strictly in terms of familiarity does not provide the whole story (Vokey & Read, 1992). Light et al. (1979) argue that there are at least two required components: (1) a familiarity component derived from the facial structure similarities, which is responsible for the typicality effects of false recognition, and (2) a specific memory component derived from the distinctiveness of encoding faces, that is responsible for typicality's effects on correct recognition. Vokey and Read (1992) argue that this component is memorability, and that, familiarity and memorability work in opposition of each other in recognition. This concept is termed the two-component approach to typicality.

The experimenters hypothesize that encoding via the independent variables (memorability and familiarity) will combine to produce the typicality effects of face recognition (the dependent variable). They expect that memorability will be associated with hits (the proper recognition of faces), but independent of false recognition. Opposite effects are expected from familiarity encoding, with which false recognition increases with a faces perceived familiarity.

Method
Participants
Twenty undergraduate university students participated in this experiment. They agreed to be subjects in return for partial course credit in an introductory psychology course.
Apparatus
Black and white yearbook photographs of 209 high school students from a distant city were used. For each individual, both their Grade 10 and Grade 12 yearbook photographs were available. The student's Grade 12 photographs had been previously rated for typicality, general familiarity, and tendency to attract false recognition. From this subject pool 176 faces (88 males and 88 females) were chosen from the extremes of rated typicality. The Grade 10 and Grade 12 photographs of each student were then converted into slides.

Procedure
Subjects were assigned into groups of five people each. Each group was then assigned into one of four counterbalancing conditions. This was done to ensure that each face was used equally as often as a previously presented face and as a lure, and equally as a Grade 10 and Grade 12 slide.

In order too guaranteed some sort of encoding and exposure to presented slides, subjects were asked to rate the likability of each of the 88 shown slides (44 male Grade 10 and Grade 12, and 44 female Grade 10 and Grade 12 faces). Each slide was presented one at a time in random order. Faces were projected for 5 s, with a 5 s break after every 11 slides. Once completed the subjects received an unrelated 20 minute filler task. For the recognition task the 88 prefamiliarized faces were mixed with the 88 novel ones. Stimuli faces were presented one slide every 5 s, with a 5 s break every 11 slides. Subjects responded "yes" or "no" to whether or not they had been previously asked to rate the face for likability.

Results and Discussion
The standard typicality effect on recognition found in other studies was reconfirmed. Analysis of variance and correlations coefficients were used. Correlations were only used for statistical verification of the directional relationship that had been qualitatively observed. Analysis found that typicality was significantly and negatively related to recognition of both Grade 10 and Grade 12 faces. That is, faces judged to be atypical were recognized significantly more often then those judged typical. Typicality was also confirmed to be significantly and positively related to false recognition.

Regression analysis was used to capture the typicality patterns of the isolated independent variables (familiarity and memorability) that were affecting the dependent variable (recognition accuracy). It was found that increases in a faces perceived familiarity led to a reduction in the rate of recognition accuracy, due to significantly increased false recognition. Conversely, for every increase in memorability the rate of recognition accuracy significantly increased. A third regression, this time using typicality as a predictor of hits and false recognition was completed. The increment in variance accounted for by this test was significant. However, it did not significantly improve prediction beyond that accounted for by the two- component variables, familiarization and memorability.

It would appear that evidence of a typicality effect, as a cumulation of familiarization and memorability effects, on face recognition is strong. The fact that the effects of the two components were additive, and that typicality measures alone contributed nothing further, suggests that the typicality effect on face recognition may be nothing more then a summed effect of the two components. Experiment 3 Encoding Strategies on Face Recognition

Typicality effects as the primary variable in determining the accuracy of face recognition is now a well-supported conclusion. Also supported is the idea that, typicality is a dual component process, through which the memorability and familiarization of encoded faces directly effects the extent to which faces can be recognized accurately. What has not yet been considered is what strategies are in fact being used initially to encode such information. One framework states that, in order to reach optimal memory performance two approaches to encoding, distinctive and relational processing, must be used. In this case, distinctive processing is the encoding of faces based on their differences from the norm. Relational processing is the encoding of faces by sorting them based on similarities. The author believes that by manipulating which of these two strategies (relational vs. distinctive; the independent variables) is used, by participants, will not significantly alter the accuracy of hits (faces accurately recognized). However, the author expects that using the distinctive strategies will provide for an increased incidence of subjects precisely remembering the period of encoding, and that using relational strategies will be attributed to participants feelings of familiarization with the faces. Whether subjects precisely remember a face or recognized it due to feelings of familiarity is the dependent variable.

Method
Participants
Twenty undergraduates at Stockholm University participated in this experiment for course credit. Ages ranged between 20 and 35 years, with a mean age of 27.4 years. An equal number of men and women participated. It was confirmed that none of the participants had experience with similar experiments. Eighty-two unidentified undergraduates were used as judges for the attractiveness, distinctiveness, and familiarity ratings required for the experimental procedure.
Apparatus
Stimulus photographs were constructed by the experimenter. 350 female and male adults between the ages of 20 and 40 years were photographed against the same neutral background in a head and shoulder view. The subjects were oriented in a variety of directions and with varied facial expressions. Although it is commonplace in such experiments to eliminate unusual facial characteristics, such as beards and eyeglasses, this procedure was not followed, so that the evaluation of the role of distinctive processing in recollection was possible.

Distinctiveness and Attractiveness Ratings. The original 350 photographs were presented to 50 undergraduates in order to judge each item in terms of distinctiveness. The judges were instructed to define distinctiveness on a 4-point scale based on the faces ability to be easily picked out of a crowd (1 = very difficult). Attractiveness was also rated on a 4-point scale (4 = very attractive).

Familiarity Ratings. The 350 photographs were secondly presented to a new group of undergraduate judges (n = 32). This group was instructed to rate each face on its familiarity. Again, a 4-point scale was used for the typicality rating (4 = sure familiar). A familiar face was defined as one the participants believed they had seen around somewhere (e.g., around campus).

Once all ratings were complete, the 350 items were broken into random sets of 50 items, with the restriction that half were of each gender and that the sets were comparable in terms of pre-rated distinctiveness, attractiveness, and familiarity. For this experiment two of these sets of stimuli were used along with twenty-five filler stimuli. Filler stimuli were used to eliminate any primacy and recency effects.

Procedure
Participants were first informed that the studies purpose was to examine similarities and differences among faces, and that it would require two 30-min sessions separated by a 1-week interval. Participants were not informed of the forthcoming recognition test.

This experiment involved a within-subject design where each subject receives both values of the independent variable, in this case the encoding task. Half of the participants were first instructed for the distinctiveness rating task. These subjects were to examine difference among 25 faces by rating the distinctiveness of each presented face. The instructions and procedure for this task were the same as the distinctiveness task defined in the materials section. This group then completed the sorting task, in which they examined a new set of 25-faces. In this task, participants were to categorize subjects into four student-type categories ("sporty type," "intellectual," "party-goer," and "homebody"). Participants were informed that the categories were simply a frame of reference for similarities and that there was no correct answer.

During the study task phase, the participants were seated 2.5 m from a projection screen. Half of the participants first completed the rating task, followed by the sorting task, and vice versa for the remaining subjects. The faces were presented via slide projector for 8 s per slide, followed by a 5 s inter-stimulus. During the inter-stimulus time the subjects rated their slides accordingly. After the first task, there was a 2 min rest period, during which the next tasks instructions were repeated.

After the study phase, the participants were informed of the forthcoming second session. Two participants were not able to return for reasons unknown to the experimenter. The remaining participants were informed that the previously presented photographs would be re-shown along with non-studied items, and that their task was twofold. They were first to respond "yes" or "no" on whether they recognized the face. Secondly, for each recognized item participants were to indicate whether they precisely "remembered" the face from the study phase, or whether they "know" the face because it seems familiar to them. If the participants had trouble identifying why they knew the face they were to indicate that they "guessed" their response. These three responses were the dependent variables and expected to change based on which task (the independent variable) was used for encoding the faces. All participants reported that they had understood the instructions.

Results and Discussion
The two values of the independent variable (distinctive vs. relational study conditions) produce nearly identical levels of overall recognition, 87% for both. However, the conditions had significantly varied effects on participants judgements of the recollective experience. The distinctive study condition produced a higher level of "remember" response then did the relational 4-type sorting task. The opposite pattern existed for the familiarity "know" response, where the relational sorting task produced more responses. Analysis of variance confirmed these observations by revealing a significant difference between the sorting task (the independent variable) and the response type (remember vs. know; the dependent variable). Statistical analysis proved no significant difference between the sorting tasks and "guessed" responses. However, it did show that the "guessed" responses were responsible for a higher rate of false recognition then the other response types.

These findings, that the rating task (distinctive processing) produced significantly higher "remember" responses than the sorting task (relational processing), and vice versa for the "know" responses, supports the hypothesis that distinct stimuli encoding attributes facilitate the variation in phenomenological experiences of conscious remembering.

General Discussion
The idea that typicality effect is the primary mechanism controlling face recognition is a well-supported hypothesis. Both experiment one and two concluded this, along with their primary conclusions. The typicality theory suggests that faces considered atypical are more easily recognized then those that are typical. As well, typical faces appear to be responsible for the incidence of false recognition.

What was not clear, prior to the three experiments presented here, was which cognitive processes actually underlie the observed typicality effect. Experiments two and three suggest that there are two levels of processing that occur. The first level that should be considered is the two component approach to typicality. This approach suggests that typicality is the result of the culmination of two other distinct cognitive processes. These processes have been defined as familiarity and memorability. Both components act upon pre-encoded information separately, then interlink this information to form the typicality effect. Familiarity is a mental analysis of how similar a face is to other faces, so that, the more familiar a face is the more typical it is. This process accounts for the high rate of false recognition of faces considered typical, an effect that is often observed in studies of face recognition. The second component, memorability, acts upon the distinctiveness of faces. A distinct face is an atypical one that is easily recognized on a face recognition test. Rarely are distinct atypical lures falsely recognized. Together, memorability and familiarity combine their effects, accounting for both the accuracy and the false recognition of faces.

Before the face stimuli can be subjected to the effects of memorability, familiarity, and subsequently typicality, it must be encoded. Experiment three suggests that, there is in fact two ways in which facial stimuli are encoded, one that leads to memorability effects and another which leads to familiarization effects. The first encoding strategy is distinctive processing, with which faces are encoded based on their differences from what one considers the typical. Distinctive processing is the encoding strategy that appears to provide information for memorability effects. The second encoding strategy, relational processing, provides information necessary for the familiarity effect. Such information is encoded by mentally sorting faces based on their similarities. This third set of findings, distinctive versus relational processing, is a fairly new idea, that has yet to be sufficiently studied or repeated. Further investigation is required before these two concepts can be as widely accepted as the typicality, familiarity, and memorability effects of face recognition.

Literature Cited

Bartlett, J. C., Hurry, S., & Thorley, W. (1984). Typicality and familiarity of faces. Memory & Cognition, 12 (3), 219-228.
Gleitman, H. (1995). Psychology (4^th ed.). New York: W. W. Norton & Company, Inc.
Craig, G. C., & Kermis, M. D. (1995). Children Today. Englewood Cliffs, NJ: Prentice-Hall, Inc.
Light, L.L., Kayra-Stuart, F., & Hollander, S. (1979). Recognition memory for typical and unusual faces. Journal of Experimental Psychology: Human Learning and Memory, 5, 212-228.
Mäntylä, T. (1997). Recollection of faces: Remembering difference and knowing similarities. Journal of Experimental Psychology: Learning, Memory, and Cognition, 23, 1203-1216.
Vokey, J. R., & Read, J. D. (1992). Familiarity, memorability, and the effect of typicality on the recognition of faces. Memory & Cognition, 20 (3), 291-302

Written By Michael W. Decaire