Recognition Memory for Extremely High Frequency Words

Home , Recognition memory

RECOGNITION MEMORY FOR EXTREMELY HIGH FREQUENCY WORDS

ELIJAH CARL MILLER

Submitted in partial fulfillment of the requirements for the degree of Master of Arts

Department of Psychological Sciences

CASE WESTERN RESERVE UNIVERSITY

January 2020

CASE WESTERN RESERVE UNIVERSITY

SCHOOL OF GRADUATE STUDIES

We hereby approve the thesis of

Elijah C. Miller

Candidate for the degree of Master of Arts.

Committee Chair

Robert L. Greene

Committee Member

Lee A. Tompson

Committee Member

Elizabeth J. Short

Date of Defense

12/03/2019

*We also certify that written approval has been obtained for any proprietary material contained therein 3

Table of Contents

Abstract 5

Introduction 6

Experiment 1 11

Experiment 2 14

Experiment 3 17

General Discussion 21

List of Tables

Table 1 24

Table 2 25

Table 3 26

Table 4 27 5

Recognition Memory for Extremely High Frequency Words

Abstract

ELIJAH C. MILLER

Three experiments are presented in which recognition memory was examined for words that occur at extremely-high frequencies in the English language. In the first experiment, participants received a simple yes-no recognition test. In the second experiment, participants were given a remember-know test of recognition. In the third experiment, a forced-choice test format was used. Across the three studies, memory for control words tended to be more accurate than memory for extremely-high-frequency words. More specifically, participants more often correctly recognized when control words had appeared on a previously presented study list and were also better at correctly identifying if control words had not been presented on the study list. This pattern of results, where one class of stimuli receives more hits and fewer false alarms than another, is called the mirror effect. These results are consistent with the claim that the mirror effect is a regularity of recognition memory.

Recognition Memory for Extremely-High-Frequency Words

The ability to remember what had occurred at a particular time in the past is called episodic memory (Tulving, 1983). A major way of testing episodic memory is recognition. On recognition tests, participants are asked to discriminate between stimuli based on episodic memory. One common way to test recognition memory is the yes-no recognition test. On yes-no tests, stimuli are shown one at a time, and participants are instructed to remember each stimulus that is shown (i.e., a “study list” of words). After seeing all stimuli, participants must then answer yes or no to words they think either appeared or did not appear, respectively, on the study list. Words that had previously appeared on the study list are referred to as “old” words, and words that had not previously appeared on the study list are referred to as “new” words.

Yes-no tests have four possible outcomes (consistent with signal detection theory;

Murdock, 1974): hits, where the participant answers yes and the stimulus did appear on the study list; misses, where the participant answers no and the stimulus did appear on the study list; false alarms, where the participant answers yes and the stimulus did not appear on the study list; and correct rejections, where the participant answers no and the stimulus did not appear on the study list. These outcomes can be interpreted in relation to each other for the purpose of measuring recognition accuracy.

Recognition accuracy may vary as a function of stimulus class; for example, recognition is more accurate for high-frequency words than for low-frequency words and 7 for concrete nouns than for abstract nouns. Glanzer and Adams (1985) noted that people had more hits and fewer false alarms to low-frequency words than for high-frequency words. Glanzer and Adams termed this pattern the mirror effect. The mirror effect occurs when comparing two categories of stimuli whenever the proportion of hits relative to misses directly mirrors the proportion of false alarms relative to correct rejections; that is, the stimulus category that has more hits will have fewer false alarms than the other category.

The mirror effect occurs so frequently that it has been described as one of the fundamental regularities of human memory (Glanzer, Adams, Iversen, & Kim, 1993).

The reason why it is so frequent is unknown. Although it remains unclear why it is so frequent, the mirror effect has been proposed as one of the central effects used to evaluate theories of memory (Ratcliff & McKoon, 2000, p. 575) and “has become a major force behind theorizing about recognition memory” (Ozubko & Joordens, 2011, p. 123). Most data on the mirror effect have been collected using word frequency stimuli, but the effect can occur with other types of stimuli.

Although Glanzer et al. (1993) argued that mirror effects are a regularity of recognition memory, exceptions to this mirror pattern can be found. For example, pseudowords (i.e., pronounceable nonwords, such as “baloof”) on recognition tests tend to have higher rates of both hits and false alarms compared to real words, a pattern known as the “pseudoword effect” (Greene, 2004). This pseudoword effect is an exception to the mirror-effect pattern because people have both more hits and more false alarms to one stimulus class (in this case, pseudowords) than to another (in this case, real 8 words). As a result, participants give far more positive responses to pseudowords than to words.

Studying exceptions to the mirror effect may help to explain why the effect occurs so frequently. It is typically assumed that recognition judgments are largely driven by familiarity and that familiarity is determined by the similarity of a test stimulus to a study list (e.g., Gillund & Shiffrin, 1984; Hintzman, 1988; Murdock, 1974; Ratcliff &

McKoon, 2000). Greene (2004) argued that the relative absence of meaning of pseudowords may make them seem very familiar, as similarity will be determined by letters and phonemes. Ozubko and Joordens (2011) proposed a test for this similarity account of the pseudoword effect. These authors first demonstrated the standard mirror effect by comparing low-frequency with high-frequency words; low-frequency words had more hits but fewer false alarms than high-frequency words. They then argued that extremely-high-frequency words (e.g., AND, THE) are also lacking in meaning. They compared recognition memory for these extremely-high-frequency words with memory for other words and found a pattern analogous to that found with pseudowords; that is, extremely-high-frequency words received more hits and more false alarms than other words. Overall, participants gave many more positive responses to extremely-high-frequency words than to other words. Ozubko and Joordens hypothesized that this pattern occurred because of the similarities between pseudowords and extremely-high-frequency words. Both extremely-high-frequency words and pseudowords follow the orthographic rules of English and therefore resemble each other and many other words in English. However, unlike most words, both 9 extremely-high-frequency words and pseudowords lack a distinct meaning.

Consequently, Ozubko and Joordens hypothesized that exceptions to the mirror effect

(e.g., the pseudoword effect) may occur when stimuli do not have much meaning.

The present study seeks to further clarify the generality of the findings from

Ozubko and Joordens (2011). As has been shown, Ozubko and Joordens’ paper has presented important results for explaining exceptions to the mirror effect. Because the mirror effect is often described as being central for understanding recognition memory, the theoretical contributions made by replicating Ozubko and Joordens would be beneficial. No previous study has replicated Ozubko and Joordens (2011), so the present study proposes to replicate and extend their results.

There are additional reasons that the Ozubko and Joorden (2011) paper should be replicated and extended. In that study, extremely-high frequency words were only compared with a constrained sample of other words (all of them being high-frequency words). It is important to demonstrate that similar findings would be found when extremely-high-frequency words are compared to a different sample of words. The present study will therefore look at low frequency words in comparison to extremely-high-frequency words. Furthermore, the Ozubko and Joordens (2011) study was an online experiment, with participants going through the study list at their own pace. Consequently, it is possible that people had spent different amounts of time on the different types of stimuli; if participants spent almost no time reading the extremely-high-frequency words on the study list, it would not be surprising that their 10 memory was so poor for them. The proposed study was conducted in a laboratory setting with study times controlled for all participants and all stimuli.

The Ozubko and Joordens (2011) experiment was also extended by examining the subjective experiences of the participants. This directly tested the hypothesis that extremely-high-frequency words and pseudowords produce a yes-no test result trend contradictory to the mirror effect because they feel so familiar to participants. Subjective familiarity was therefore included among data collected. Tulving (1985) argued that subjective familiarity can be assessed by asking people to indicate why they gave a positive response on a yes-no test. He used a remember-know procedure in which, after every positive response on a recognition test, a participant must indicate whether that response was based on knowing or remembering. Responses based on knowing are supposed to reflect pure familiarity, whereas responses based on remembering require explicit recall of the encounter with the stimulus. The remember-know test was introduced to the present study to query whether a respondent feels a stimulus is familiar

(i.e. “knowing”) or explicitly recalls encountering it (i.e. “remembering”; Tulving, 1985).

Gardiner and Richardson-Klavehn (2000) claim that there is a real subjective difference between remembering and knowing. Experiment 2 therefore used a remember-know test procedure.

Experiment 3 used a forced-choice test procedure. After studying a list of words as in the first two experiments, participants were tested by being shown pairs of words. They had to pick the member of each pair that had been shown on the study list.

Each test pair contained one extremely-high-frequency word and one control word. If 11 extremely-high-frequency words function like pseudowords, participants would pick them on more than half of the trials. Demonstrating this helped to generalize the results of the first two experiments to a somewhat different procedure. Also, it is commonly assumed that forced-choice tests do not require a response criterion, as participants simply pick the better option on each trial (Greene, 2004, Murdock, 1974). Thus, using a forced choice test extended the results to a simpler measure of recognition.

Experiment 1

Method

Experiment 1 attempted to replicated Ozubko and Joorden (2011). Participants saw a list composed of equal numbers of extremely-high-frequency words and control words. They then received a yes-no recognition test.

Participants. Subjects were 50 Case Western Reserve University undergraduate students enrolled in an introductory psychology course who voluntarily participated in return for course credit.

Materials. Two sets of words were constructed. The extremely-high-frequency words had frequencies between 850 and 28,852 occurrences per million in the Kucera and Francis norms. Control words were equated with respect to number of letters and number of syllables with the other set and had frequencies of 10 and 50 occurrences per million. 12

Procedure. Participants were tested in groups. They saw a study list of 70 words projected on a screen at a rate of one word per second. All words were shown in uppercase letters. The study list contained 30 extremely-high-frequency words, 30 control words, and 10 untested buffer words occupying the first five and last five serial positions. Two different versions of the study list were used so that across subjects all words were equally likely to be studied and non-studied.

Immediately after the 70 study words were presented, the 120 test words were shown to participants on a screen at a five-second rate. Sixty of the test stimuli were words that had been on the study list (i.e., old words), and the remaining 60 stimuli had not been on the study list (i.e., new words). The experimenter provided each subject with a response sheet for the yes-no test. The response sheet contained the following instructions: “WAS THIS STIMULUS ON THE LIST? ANSWER ‘YES’ OR ‘NO’” The response sheet contained columns of numbered yes-no word pairs which participants were instructed to use for providing their responses.

Results

Results from this study, reported as proportions, are presented in Table 1. Shown are the hit rate (i.e., proportion of old items correctly receiving a positive recognition response), false alarm rate (i.e., proportion of new items incorrectly given a positive recognition response), and accuracy score for each class of stimuli. Consistent with the hypothesis of this study, Experiment 1 presented evidence for a difference in memory accuracy between extremely-high-frequency words and control words. A simple way of 13 measuring accuracy of memory for a particular condition is to take the difference between the number of times a subject says yes when it is the right answer (i.e., hits) versus when it is the wrong answer (i.e., false alarms). A higher difference between hits and false alarms should in theory indicate greater accuracy in recognition memory.

Accuracy scores were calculated for extremely-high-frequency words and control words by taking the difference between the number of hits and the number of false alarms for each participant. Then, a paired-samples t-test was carried out. A difference between mean accuracy scores for the control words and extremely-high frequency words was found, t(49) = 11.11, p < .001.

The present study was not intended to simply study differences in accuracy, but instead to investigate the relationship between hits and false alarms. When a mirror effect pattern is found, hit rates and false alarm rates are mirror images: if the first condition has a higher hit rate, then the second condition will have a higher false alarm rate than the first. Ozubko and Joordens (2011) suggested that the use of extremely-high-frequency words might lead to an exception to the mirror effect, with extremely-high-frequency words receiving both more hits and more false alarms than control words; however, our results did not support that prediction. Rather, as is evident from the proportions in Table 1, a typical mirror effect pattern was found. There were significantly more hits to control words compared to extremely-high frequency words, t(49) = 2.25, p = .029. There were significantly more false alarms to extremely-high-frequency words than control words, t(49) = 8.91, p < .001. Therefore, a mirror effect was found here in contrast to the prediction of Ozubko and Joordens (2011) 14 but consistent with the claim that the mirror effect is a regularity of recognition memory

(Glanzer et al., 1993).

Experiment 2

Experiment 2 had two purposes. First, it attempted to replicate the results of

Experiment 1. Second, it extended these results to a remember-know test of recognition memory. A remember-know test is supposed to capture the subjective experience of recognition.

Method

Participants. Subjects were 24 Case Western Reserve University undergraduate students enrolled in a Psychology 101 course who voluntarily participated in return for course credit.

Materials. The sets of words from Experiment 1 were used here as well. Procedure. Participants were again tested in groups. Presentation of the study list followed exactly the procedure of Experiment 1. After the 70 study words were presented, the 120 test words were then shown to participants in a five-second rate. The experimenter provided each subject with a blank response sheet for the remember-know test. The response sheet contained the following instructions:

“WAS THIS WORD ON THE LIST? ANSWER ‘REMEMBER’ ‘KNOW’ OR ‘NO’

REMEMBER: You can remember the experience of seeing it on the list 15

KNOW: You cannot recollect seeing it on the list but you know it was there

NO: It was not on the list” These responses are the three types of responses used in a standard remember-know test.

The response sheet contained columns of numbered remember-know-no sets which participants were instructed to use for providing their responses.

Results

One important purpose of Experiment 2 was to replicate the results of Experiment

1 demonstrating a mirror effect on a list of extremely-high-frequency words and control words. In the context of the remember-know test, hits are defined as correct positive responses (i.e. subject chooses a response of remember or know to an old word). False alarms are incorrect positive responses (i.e. subject chooses a response of remember or know to a new word). Table 2 shows the hit rate, false-alarm rate, and accuracy score

(difference between hits and false alarms) for the two stimulus classes in Experiment 2.

Significance tests showed that the control words received more hits than extremely-high-frequency words, t(23) = 7.10, p < .020, and the extremely-high-frequency words received more false alarms than control words, t(23) =

2.51, p < .001. Thus, the results of Experiment 1 were replicated; a mirror effect was found with extremely-high-frequency words receiving fewer hits but more false alarms than control words. As was the case in Experiment 1, accuracy was significantly higher for control words than for extremely-high-frequency words, t(23) = 8.83, p < .001 16

Experiment 2 was designed not only to replicate Experiment 1 but also to examine participants’ subjective experience of recognizing the words. The remember-know test allowed subjects to specify whether they explicitly remembered seeing a given word, or simply have a vague, implicit sense of familiarity with the word, as though they intuitively know the word was on the memory list. The breakdown of hits and false alarms given “remember” and “know” responses are shown in Table 3. There were significantly more hits given “remember” responses to control words than to extremely-high-frequency words, t(23) = 5.67; p<.01. However, there were more ‘know’ hits to extremely-high-frequency words than to control words, t(23) = 4.23, p<.01. Thus, even when participants gave correct positive responses, the reason for doing so varied between control words and extremely-high-frequency words: Correct positive responses to control words were mostly based on explicit recollection of the original event, while correct positive responses to extremely-high-frequency words were largely based on an overall feeling of familiarity. When participants made false alarms, they indicated a

“know” experience more often for extremely-high-frequency words than for control words, t(23) = 6.03, p<.01. False alarms accompanied by “remember” experiences were rare and did not vary significantly between the two stimulus categories, t(23) = 1.76, p =

.091, though participants gave somewhat more “remember” false alarms to extremely-high-frequency words than to control words.

These results indicate that, in terms of hits, people gave more “remember” responses to control words than to extremely-high-frequency words, but they had more

“know” hits to extremely-high-frequency than to control words. Therefore, even when 17 recognition was accurate, the subjective experience was different for the two stimulus classes. Recognition of extremely-high-frequency words was largely determined on the basis of familiarity, while recognition of control words was usually accompanied by a feeling of conscious recollection of the original experience.

Experiment 3

Experiment 3 extended the results of the first two experiments by using the forced choice recognition memory test. Using this test verified the generalizability of the results of the first two experiments. Participants received a study list of extremely-high-frequency and control words in identical fashion to that done in

Experiments 1 and 2. On the test, participants were shown pairs of words. Each pair of words contained one extremely-high-frequency word and one control word. They were asked to select the member of each pair that had been on the study list.

Glanzer and Bowles (1976) introduced a new way to use forced-choice tests to study recognition memory. In their procedure, some test pairs did not have a single correct answer. Some test pairs contained two old words, so that both choices were equally appropriate. Other test pairs contained two new words, so that neither option was appropriate. Glanzer and Bowles argued that this sort of arrangement was useful for determining the relative familiarity of two options. On the test part of the forced-choice procedure, a word could either be a control word or an extremely-high-frequency word.

A word could also be either old or new. The four possible stimuli types therefore were: 18 old control word, old extremely-high-frequency word, new control word, and new extremely-high-frequency word. From these four possible stimuli types, four possible word pair types were created. The first pair type would pair an old extremely-high frequency word with a new control word; the correct recognition response would be the old extremely-high frequency word. The second pair type would pair a new extremely-high-frequency words with an old control word; here, the correct response would be the old control word.

The other two pair types did not have one correct answer and one incorrect answer. The third pair type had one old extremely-high-frequency word and one old control word; therefore, both options are correct, although participants are only allowed to choose one option. The fourth pair type had one new extremely-high-frequency word and one new control word; here, neither option is correct, although participants were still required to select one option. These latter two pair types are especially useful for demonstrating a mirror effect. Essentially, the mirror effect states that “A stimulus class that seems particularly old when old will seem particularly new when new”. Based on the results of Experiments 1 and 2, control words seem particularly old when old (as shown by more hits) and particularly new when new (as shown by fewer false alarms), relative to extremely-high frequency words. Therefore, on the third pair type (where both options were old), the control words should seem particularly old, and participants should tend to pick control words. On the fourth pair type (where both options were new), the control words should seem particularly new, so participants should tend to pick the extremely-high-frequency words as the old words. 19

Method

Participants. Subjects were 22 Case Western Reserve University undergraduate students enrolled in a Psychology 101 course who voluntarily participated in return for course credit.

Materials. The sets of words from Experiment 1 were used here as well. Procedure. Participants were again tested in groups. They received a study list following the same procedure as the first two experiments. They then received a forced-choice recognition test on which they saw 70 pairs of words. The pairs were shown one pair at a time at a five-second rate. Participants were asked to pick the word that they believed was on the study list from each test pair. All test pairs contained one control word and one extremely-high-frequency word. If extremely-high-frequency words functioned like pseudowords, participants would in theory select them as a response more than half of the time.

The experimenter provided each subject with a blank response sheet for the forced choice test. The response sheet contained the following instructions: “CHOOSE THE STIMULUS THAT YOU ARE MOST CONFIDENT WAS ON THE LIST. CIRCLE YOUR CHOICE.

YOU MUST CIRCLE EXACTLY ONE OPTION ON EACH LINE.” The response sheet contained columns of numbered a-b letter pairs which participants were instructed to use for providing their response. The first ten pairs on the forced-choice test were considered practice and were not scored.

Results 20

The results of Experiment 3 are presented in Table 4. Note that, because half of the test pairs did not contain one correct and one incorrect choice, the proportions presented in Table 4 are not proportions correct. Rather, Table 4 presents the proportion of times when participants choice the extremely-high-frequency word option rather than the control word.

The first pair type presented in Table 4 is when the pair contained one old extremely-high-frequency word and one new control word. The correct option would be the extremely-high-frequency word. Participants made the correct choice .80 of the time, which differs significantly from the chance level of .50, t(21) = 9.92, p < .001.

The second pair type presented in Table 4 is when the pair contained one new extremely-high-frequency word and one old control word. The correct option would be the control word. Participants incorrectly chose the extremely-high-frequency words .24 of the time; that is they made the correct choice of the control word on .76 of these pairs.

This differs significantly from the chance level of .50, t(21) = 8.83, p < .001.

The third pair type presented in Table 4 paired an old extremely-high-frequency word with an old control word. Both options were correct choices, but participants are only allowed to choose one. They chose the extremely-high frequency word on .28 of trials; that is, they chose the control word on .72 of trials. This proportion is significantly different from the chance level of .50, t(21) = 6.49, p < .001. Therefore, when both the extremely-high-frequency word and the control word are old, the control word tends to feel particularly old and is usually chosen. 21

The fourth pair type presented in Table 4 paired a new extremely-high-frequency word with a new control word. Both options were incorrect choices, but participants are required to choose one. They chose the extremely-high frequency word on .78 of trials.

This proportion is significantly different from the chance level of .50, t(21) = 8.22, p <

.001. Therefore, when both the extremely-high-frequency word and the control word are new, the control word tends to feel particularly new and is usually not chosen. Control words feel particularly old when old (as in the third pair type) and feel particularly new when new (as in the fourth pair type). Therefore, these results are consistent with the claim that the comparison of control words with extremely-high-frequency words leads to a mirror effect.

General Discussion

The present study provided evidence across three separate experiments that a mirror effect occurs when extremely-high-frequency words are compared with control words.

Experiment 1 obtained the mirror effect using a yes-no recognition memory test; hit rates were higher for control words than for extremely-high-frequency words and false alarm rates were higher for extremely-high-frequency words than for control words.

Experiment 2 replicated the finding of the mirror effect. Experiment 2 also provided evidence that the mirror effect for extremely-high frequency words may occur because these words are more familiar, generally speaking, compared to control words; participants gave more “know” responses to extremely-high-frequency words and more 22

“remember” responses to control words. Experiment 3 obtained the mirror effect using a forced-choice recognition memory test.

The present study was completed to contribute to literature exploring the cause of the mirror effect as a prominent yet unexplained phenomenon of recognition memory.

Greene (2004) argued that failure to find mirror effect when pseudowords are compared to words was a result of the relative meaninglessness of pseudowords. Ozubko and

Joordens (2011) provided evidence supporting this claim. Ozubko and Joordens also argued that extremely-high-frequency words are also relatively meaningless, so that a comparison of extremely-high-frequency words and meaningful control words should not lead to a mirror effect. Ozubko and Joordens, in their preliminary online study, supported this hypothesis. Specifically, they found that extremely-high-frequency words relative to control words had both higher hits and higher false alarms.

The present study, however, reports results here inconsistent with the finding of

Ozubko and Jordens (2011) that extremely-high-frequency words have both more hits and false alarms than control words. Extremely-high-frequency words were consistently found to display the mirror effect (more false alarms and fewer hits when compared to control words) across the three experiments of the present study. It is not possible, at least from the data of the present study, to determine which factor might explain why the results are inconsistent. Unlike Ozubko and Joordens, encoding times were controlled.

Also, different stimuli were used for Ozubko and Joordens compared to the present study.

If meaninglessness leads to the lack of a mirror effect, then extremely-high-frequency words should have had both higher hits and false alarms because they are in general very 23 low in meaningfulness. Therefore, either Ozubko and Joordens’ (2011) were incorrect to claim that extremely-high-frequency words are relatively meaningless false, or Greene

(2004) and Ozubko and Joordens were wrong that relative meaningless always leads to both higher hits and higher false alarms.

The fact that mirror effect was consistently found for extremely-high-frequency words in the three experiments of the present study supports the claim that the mirror effect is a regularity of recognition memory. Although it is unclear why the mirror effect is so common in recognition memory, Experiment 2 and Experiment 3 of the present study provide evidence supporting the notion that the mirror effect is related to how familiar a class of stimuli tends to feel. Further research is needed to further illuminate why the mirror effect is a regularity of recognition memory, and it may prove particularly helpful for future studies to test how familiarity may contribute to the effect.

Table 1

Results of Experiment 1

Word Type Hit Rate False Alarm Rate Accuracy

Extremely-High-Frequency .68 .39 .29

Control .72 .17 .55

Table 2

Overall Results of Experiment 2

Word Type Hit Rate False Alarm Rate Accuracy

Extremely-High-Frequency .67 .44 .23

Control .74 .24 .50

Table 3

Remember (R) and Know (K) Responses in Experiment 2

Word Type R Hits R False Alarms K Hits K False Alarms Extremely-High-Frequency .28 .11 .39 .33

Control .51 .08 .23 .16

Table 4

Proportion of Pairs on Which the Extremely-High Frequency (EHF) Word Was Chosen in Experiment 3

Pair Type Proportion

EHF Old – Control New .80

EHF New – Control Old .24

EHF Old – Control Old .28

EHF New – Control New .78

APPENDIX 1 ANSWER SHEETS

WAS THIS WORD ON THE LIST? ANSWER ‘YES’ OR ‘NO’ 1. YES NO 41. YES NO 81. YES NO 2. YES NO 42. YES NO 82. YES NO 3. YES NO 43. YES NO 83. YES NO 4. YES NO 44. YES NO 84. YES NO 5. YES NO 45. YES NO 85. YES NO 6. YES NO 46. YES NO 86. YES NO 7. YES NO 47. YES NO 87. YES NO 8. YES NO 48. YES NO 88. YES NO 9. YES NO 49. YES NO 89. YES NO 10. YES NO 50. YES NO 90. YES NO 11. YES NO 51. YES NO 91. YES NO 12. YES NO 52. YES NO 92. YES NO 13. YES NO 53. YES NO 93. YES NO 14. YES NO 54. YES NO 94. YES NO 15. YES NO 55. YES NO 95. YES NO 16. YES NO 56. YES NO 96. YES NO 17. YES NO 57. YES NO 97. YES NO 18. YES NO 58. YES NO 98. YES NO 19. YES NO 59. YES NO 99. YES NO 20. YES NO 60. YES NO 100. YES NO 21. YES NO 61. YES NO 101. YES NO 22. YES NO 62. YES NO 102. YES NO 23. YES NO 63. YES NO 103. YES NO 24. YES NO 64. YES NO 104. YES NO 25. YES NO 65. YES NO 105. YES NO 26. YES NO 66. YES NO 106. YES NO 27. YES NO 67. YES NO 107. YES NO 28. YES NO 68. YES NO 108. YES NO 29. YES NO 69. YES NO 109. YES NO 30. YES NO 70. YES NO 110. YES NO 31. YES NO 71. YES NO 111. YES NO 32. YES NO 72. YES NO 112. YES NO 33. YES NO 73. YES NO 113. YES NO 34. YES NO 74. YES NO 114. YES NO 35. YES NO 75. YES NO 115. YES NO 36. YES NO 76. YES NO 116. YES NO 29

37. YES NO 77. YES NO 117. YES NO 38. YES NO 78. YES NO 118. YES NO 39. YES NO 79. YES NO 119. YES NO 40. YES NO 80. YES NO 120. YES NO

CHOOSE THE WORD THAT YOU ARE MOST CONFIDENT WAS ON THE LIST. CIRCLE YOUR CHOICE. YOU MUST CIRCLE EXACTLY ONE OPTION ON EACH LINE. 1. A B 36. A B 2. A B 37. A B 3. A B 38. A B 4. A B 39. A B 5. A B 40. A B 6. A B 41. A B 7. A B 42. A B 8. A B 43. A B 9. A B 44. A B 10. A B 45. A B 11. A B 46. A B 12. A B 47. A B 13. A B 48. A B 14. A B 49. A B 15. A B 50. A B 16. A B 51. A B 17. A B 52. A B 18. A B 53. A B 19. A B 54. A B 20. A B 55. A B 21. A B 56. A B 22. A B 57. A B 23. A B 58. A B 24. A B 59. A B 25. A B 60. A B 26. A B 61. A B 27. A B 62. A B 32

28. A B 63. A B 29. A B 64. A B 30. A B 65. A B 31. A B 66. A B 32. A B 67. A B 33. A B 68. A B 34. A B 69. A B 35. A B 70. A B

APPENDIX 2 LIST OF STIMULI

Extremely-High Frequency Words ABOUT ALL ALSO AND ANY ARE BECAUSE BEEN BUT COULD DID EACH EVEN FOR FROM HAD HAS HAVE INTO ITS JUST LIKE MADE MAN MANY MAY MORE MOST MUCH MUST NOT NOW ONE ONLY OTHER OUR OUT OVER SHE SHOULD SOME SUCH 34

THAN THAT THEIR THEM THEN THERE THEY THIS THOSE THROUGH WAS WELL WHERE WHO WITH WOULD YOU YOUR

Control Words Angel Anchor Arrow Asleep Atom Ballet Beard Bee Brain Bacon Barn Bell Brick Bubble Canvas Cat Clock Coal Cotton Chicken Corn Disk Dawn Devil Dirt 35

Dollar Drum Ear Egg Elbow Farmer Finger Flower Fork Flame Fog Giant Ghost Hay Ham Honey Horn Hen Inch Infant Ice Iron Jar Jaw Journal Kiss Knight Lamp Lemon Lime Lion Lung Mice Monk Milk Mud Pastor Pistol Pencil Penny Pope Rabbit Rake Razor

REFERENCES

Gardiner, J. M., & Richardson-Klavehn, A. (2000). Remembering and knowing. In E. Tulving & F. I. M. Craik (Eds.), The Oxford handbook of memory. Oxford, UK: Oxford University Press.

Gillund, G., & Shiffrin, R. M. (1984). A retrieval model for both recognition and recall.

Psychological Review, 91, 1-67. Glanzer, M. & Adams, J.K. (1985). The mirror effect in recognition memory. Memory & Cognition, 13, 8-20. Glanzer, M., Adams, J. K., Iverson, G. J., & Kim, K. (1993). The regularities of

recognition memory. Psychological Review, 100, 546-567. Glanzer, M, and Bowles, N. (1976). Analysis of the word frequency effect in recognition memory. Journal of Experimental Psychology: Human Learning and Memory, 2, 21-31. Greene, R. L. (2004). Recognition memory for pseudowords. Journal of Memory and Language, 50, 259-267. Hintzman, D. L. (1988). Judgments of frequency and reconition memory in a multiple- trace recognition memory model. Psychological Review, 95, 528-551. Murdock, B. B. (1974). Human memory: Theory and data. Potomac, MD: Lawrence Erlbaum.

Ozubko, J., & Joordens, S. (2011). The similarities (and familiarities) of pseudowords 37 and extremely high-frequency words: Examining a familiarity-based explanation of the

pseudoword effect. Journal of Experimental Psychology. Learning, Memory, and Cognition, 37, 123-39. Ratcliff, R., & McKoon, G. (2000). Memory models. In E. Tulving and F.I.M. Craik

(Eds.), The Oxford Handbook of Memory (pp. 571-582). Oxford, UK: Oxford University Press.

Tulving, E. (1983). Elements of episodic memory. Oxford, UK: Oxford University Press.

Tulving, E. (1985). Memory and consciousness. Canadian Psychology, 26, 1-12.