Running head: attention efficiency and performance

The Effects of Text Exaggeration and Orthographic Similarity on Correct Identification of Drug Names KRAIG L. SCHELL, KYLE KELLEY & CORY HUNSAKER Angelo State University San Angelo, Texas U.S.A. This study tested the effect of text exaggerations (capital letters, color enhancements) of specific segments of drug names and orthographic similarity differences on correct name recognition. One hundred and three participants saw 80 pairs of drug names over six trial blocks (480 pairs total) where they saw an exaggerated prime word, a black mask for 450 milliseconds, and a target word which was either the same word or a potentially confusable pair word. Participants decided whether the prime and target words were the same. Results indicated that name identification was less accurate at higher levels of orthographic similarity as expected. However, the text exaggerations did not reliably improve performance on the task, contrary to expectations. The practical and theoretical importance of these results is explored, especially regarding current federal mandates concerning exaggerated labeling on drug products. This research was supported by a grant from the Institute for Safe Medication Practices, Huntington Valley, PA. The authors thank Liz Bankhead, Paul Turner and our undergraduate research team for assistance in data collection. Correspondence concerning this article should be addressed to Kraig Schell, Department of Psychology, Angelo State University, Box 10907 ASU Station, San Angelo, TX, 76909. Electronic mail can be sent to The Effects of Text Exaggeration and Orthographic Similarity on Correct Identification of Drug Names Medication errors and patient safety have been salient issues in health care since the Institute of Medicine released their disturbing report on the frequency with which these errors were being made (Kohn, Corrigan & Donaldson, 1999). Since that time, many recommendations and mandates have been put forward in hospitals and pharmacies in an attempt to reduce the incidence of these errors. One of these mandates instated by the Food and Drug Administration required that labeling for a short list of drugs be changed. Specifically, the requirement addressed pairs of drug names that were similar either in orthography or phonology under the assumption that they were vulnerable to confusion. To remedy this, the parts of the names that made them unique were to be exaggerated using capital letters. For example, "hydralazine" and "hydroxyzine" are one of the pairs on the FDA list, and it is clear that they are very similar to one another. So, current labeling for these drugs would change the appearance of the name to be "hydrALAzine" and "hydrOXYzine" to emphasize the unique aspects of each name. Though the idea sounds good in principle, we know of no controlled research that supports the effectiveness of the mandate. Not only could such a line of research evaluate the utility of exaggerated labeling, it could help to identify other interventions to further prevent medication mishaps where the name of the drug is a contributing factor. It is with this purpose in mind that the current study was designed. Orthographic similarity and text exaggeration will be examined to determine whether they affect the ability to detect drug names correctly in a rapid prime-target word matching task. Background Literature: Reading, Attention and Human Factors Of course, the intended effect of text exaggeration is to draw the attention of the pharmacist to the part of the drug name that is unique in hopes of reducing name confusability. However, reading words involves more than just attention, and the assumed effect also implies that the rest of the word (the non-exaggerated portion) will command little attention during performance. Fortunately, the reading literature provides some well-documented premises that bear directly on this investigation, summarized below in two main points. First, it is generally accepted that reading is largely an automatic process taking the form of a skill and requiring little focused attention once well learned (see Just & Carpenter, 1980 for example). The expected effect of the text exaggeration is to make the reading process more effortful, but it is unclear as to whether a few capital letters are salient enough to disrupt the automaticity of reading. If they are not, then the entire drug name is likely to be read automatically and the text exaggeration may actually harm performance rather than help it by altering the expected form of the word. Second, it is largely accepted that there are two reading systems: orthography and phonology (Frost, 1998; Lukatela & Turvey, 1994a, 1994b; Stone & Van Orden, 1993). There is some debate about which system is more dominant under which circumstances (Carreiras, Perea, & Grainger, 1997; Katz & Frost, 2001; Pugh, Rexer, & Katz, 1994), but the mere presence of these two reading systems is enough to make the effects of text exaggeration more complicated. The capital letters alter the word shape – they do not alter the actual orthography of the word (the letter sequences). The shape of the word being read is not generally considered to be very important to reading (Frost, 1998; Perea & Rosa, 2002), so it remains to be seen whether such exaggerations will really have much of an effect. In other words, despite the graphical change, the same letter sequences exist. So, it is likely that text exaggerations must overcome some elemental reading principles in order to have the effect intended by the pharmacy industry, and we are concerned about their ability to Human factors research is also relevant to this issue in a number of ways. First, if text exaggerations are a salient intervention for name recognition, then they should have this effect in the initial stages of processing. Research shows that we are drawn toward unique stimuli in our visual field, sometimes called "singletons" (Duncan & Humphreys, 1989; Yantis & Hillstrom, 1994; but see Egeth & Yantis, 1997 for clarification), so the capital letters amongst "normal" letters must be unique enough to capture attention if they are to work as intended. However, it is uncertain if such text exaggerations can attract attention so powerfully. Second, there are several other types of exaggerations that should be as capable of attracting attention if not more so. Research indicates that color, for example, can capture attention under certain visual search paradigms (Treisman & Gelade, 1980; Yantis, 1998). Furthermore, though there are no universal meanings for colors, red is typically interpreted as "stop" or "wait", at least in many Western cultures (Gage, 1995), which is the message that the text exaggeration intervention is intended to send. So, red letters may be just as useful as capitals, if not more. Third, there is the problem of sensitivity versus response bias. There is no theoretical basis to predict how text exaggerations should affect these two measures. On the one hand, it could be argued that sensitivity should increase and response bias should become more conservative if the text exaggerations provide more salient target information as they are meant to do. On the other hand, one could also say that each exaggeration is essentially a signal, which dramatically increases event probability and should have significant negative effects on response bias and sensitivity (Davies & Parasuraman, 1982; See, Warm, Dember, & Howe, 1997; Wickens & Hollands, 2000). In short, there are reasons in the relevant literature to question whether text exaggerations are the preventative measures that they are meant to be. Calculating Orthographic Similarity: The Trigram Formula A second issue which is relevant to this investigation is orthographic similarity. We define this as a mathematical ratio of trigrams (3-letter sequences) in common between two pair words, expressed in Equation 1: where A = the number of trigrams in Word 1, B = the number of trigrams in Word 2, and C = the number of trigrams in common between the two words. The leading space in each word is counted as a character. For example, Midol has the following trigrams: [_mi, mid, ido, dol] and Haldol has the following trigrams: [_ha, hal, ald, ldo, dol]. The
bolded trigrams are in common, so using the formula above: 2(1) / (4 + 5) = .22. Trigram scores range from 0 (no trigrams in common) to 1 (word is exactly the same). Also, note that shorter words will typically result in higher trigram scores than longer words with the same number of trigrams in common. The trigram formula is not the only approach to orthographic similarity nor it is necessarily ideal, but it provides a good starting point for this line of study and is well-used in the medication error literature (Lambert, 1997; Lambert, Chang & Gupta, 2003; Lambert, Chang & Lin, 2001; Lambert, Lin, Chang & Though it is feasible that word pairs with higher trigram scores might be more easily confused than word pairs with lower trigram scores, this may not necessarily be the case. One reason already mentioned is the potential confound concerning the length of the words in the pairing, but a second (and perhaps more interesting) reason is that some research which has argued that name confusability is one of the primary suspects in medication errors is built on self-report data from medical professionals (Lambert, 1997). Thus, the medication error events studied were subject to self-report bias. Furthermore, although the research points out that, among the events studied, trigram scores tended to be high, the only events selected for study were those where the reporting pharmacist argued a priori that he/she must have confused the correct drug with the one actually dispensed. Thus, the fact that the drugs implicated in these events showed high trigram scores is not surprising – it is likely an artifact of the sampling process. More recent research has shown that, in controlled studies using pharmacists, neighborhood density did significantly affect name recognition, particularly on names presented more frequently during testing (Lambert, 2005). But, an earlier study showed that name similarity facilitated free recall of a list of drug names (Lambert, Chang & Lin, 2003). To further clarify this issue, this study will test two sets of word pairs – a low- trigram and a high-trigram set – to determine under controlled conditions if name confusability is more problematic as trigram score increases. In addition, stimulus presentation frequency will be held constant at the word-pair level. Because we are using pharmacy-naïve participants, we can eliminate the hindsight bias that may be confounding pharmacist reports given that these participants will not be expected to have preconceived notions of "which drug I probably confused the other drug for." Also, we will be able to examine the effects of high orthographic similarity more completely because pharmacy-naïve participants will not possess the rich knowledge structures for these drug products that pharmacists do that may confound a pure perceptual analysis. Hypotheses Because the goal of this study is to empirically test the logic behind exaggerated text and whether that logic holds across levels of orthographic similarity, we propose the following hypotheses: H1: Measures of detection performance (d', raw error rate) will be significantly higher in the capital-letter exaggeration group than in the control group. H2: Measures of detection performance (d', raw error rate) will be significantly higher in the color-enhancement exaggeration group than in the control group. H3: Measures of detection performance (d', raw error rate) will be significantly lower for the high-trigram group than for the low-trigram group in all text exaggeration H4: Measures of detection performance (raw correct decision rate, c) will be more liberal in the high trigram group than in the low trigram group. Participants One hundred and three participants from a southwestern university volunteered to participate in the research. Each participant was paid $6.50 USD for their time, which ultimately totaled 3 hours. The sample was composed of 22 men and 81 women and had a mean age of 22.2 years (SD = 5.6). All participants were enrolled full-time in college at the time of testing and were recruited from a variety of different courses. Materials Stimuli. 80 drug name pairings were selected as stimuli for this study. These name pairings are listed in Appendix A sorted by their trigram scores and divided into "low" and "high" groups. These groupings were constructed using several steps. First, the Food and Drug Administration has mandated that certain drugs be labeled using exaggerated labeling much like that being tested in this study, and these name pairs were included in our name database. Second, a leading U.S. non-profit medication safety organization has published a list of name pairs that have caused confusion in the past based on pharmacist self-reports of errors. Third, a small team of researchers (including the lead author) examined the 2003 version of the Physician's Desk Reference book of prescription drugs for name pairs that appeared similar enough to each other to be potentially confusing. The final database included 128 name pairs, although it should be noted (as is evident from Appendix A) that some individual drug names were used in multiple pairs. The 128 name pairs were filtered further using a trigram formula first used by Lambert and his colleagues in the context of drug name confusability (Lambert, 1997; Lambert, et al, 1999). The trigram formula parses each drug name into successive sets of three letters and compares the trigram sets for the word pairs for identical members. Generally speaking, the more identical trigrams there are between two names, their greater their orthographic similarity and the higher their trigram score will be. To create a complete trigram matrix, all drug names in the database were compared to all other drug names, and then trigram scores were sorted from highest to lowest. This data also can be found in Appendix A. To construct an independent variable based on orthographic similarity, the highest and lowest 40 trigram score pairs were used, eliminating approximately the middle third of the total trigram database. The independent variable for text exaggeration was created by selecting a part of every drug name and emphasizing it graphically using either capital letters (of the same font and size) or color enhancement (the selected word segment printed in red instead of black). Examples of these exaggerations using the two drug names discussed above (hydralazine and hydroxyzine) are shown in Figure 1. Every drug name used in the study could be emphasized using these techniques. Finally, the selected segment of the drug name is the most distinct segment from the potentially confusable pair name; thus, the "ala" in hydralazine would be emphasized to distinguish it from hydroxyzine. Though this is not a foolproof method of selecting which parts of the name to emphasize, it is the method recommended by the FDA and so we employed it in this study. Signal probability for all groups was held constant at 50%; thus, half of the name pairs presented to participants were mismatches and half were matches. This was done to maximize the number of times that each name pair was observed by the participants, so that later analyses can approach the data from the standpoint of individual word pairs to determine the shared characteristics of the most confusable drug names. Equipment. Two Windows-based computers were used for testing, each running Windows 2000 at 1 Ghz with a 17-inch tube monitor (dot pitch .22) at 1024 x 768 resolution. Superlab Pro was used to display the drug name stimuli and RB-834 8-button response pads were used to record participants' decision data. All visual stimuli were Figure 1. Examples of text exaggerations applied to word pairs in this study. presented as JPEG files 3" from the screen bottom and beginning 2.2" from the left edge of the screen. Each stimulus filled a rectangular space 1" high; the width of this space varied with the length of the word displayed. The approximate distance from eye to screen was 32" using a prototypical 66" tall human participant as a reference. Stimuli were presented in black type using Arial font (except for words displayed using the color enhancement exaggeration, in which case part of the word was displayed in red) against a white screen background. Procedure Participants were randomly assigned to an experimental group based on which text exaggeration (control, capital letters, color enhancement) and orthographic similarity (high, low) to which they would be exposed. Stimuli for each experimental condition were saved in a separate experiment file but all files followed the same general procedure as described in this section. Figure 2 shows the screen progression for a given trial in the experiment. The first image in any trial was a series of black plus signs against a white background. This image was displayed for three seconds, after which a prime word appeared containing any necessary text exaggerations. The prime word was displayed for 450 ms, and then immediately replaced with a black bar which covered the entire area where the prime word was located. The black bar was intended to clear the participant's iconic memory registers and encourage short-term memory usage. The final screen in each trial was the target word, which was always presented in standard type with no exaggerations and in the same physical position as the prime word. Participants were instructed to, as quickly as possible, decide whether the target and the prime words were identical. If they were, participants were to press the green button on the response pad; if they were not, the red button on the pad was the correct response. Once a response was made, the plus sign image returned and a new trial began immediately. In an effort to simulate the feedback isolation of a pharmaceutical environment no feedback concerning correct or incorrect decisions was provided at any point during the task. Participants were exposed first to four practice trials using common English words to orient them to the exaggeration type they would see. After these four trials, remaining questions were answered and participants were allowed to relax with eyes closed for one minute. Next, they began the main experiment, which was composed of 480 total word pairs (80 unique word pairs) over six trial blocks. Thus, each unique word Figure 2. Graphic depiction of the screen sequence involved in one experimental trial. pair was seen six times by each participant. However, within trial blocks, the sequencing of the word pairs was randomized to control for practice and order effects. For each trial, response data was recorded in addition to response time data. Also, information concerning which word pair was displayed at which time was also recorded, allowing for future analyses to examine which particular word pairs were most confusable in the study. Finally, between each trial block, participants were asked to relax for three minutes with eyes closed before beginning the next block. Experimenters were on hand to track this relaxation time and initiate the next block of stimuli, but during the experimental blocks, participants were left alone in the primary testing room. Experimenters could be contacted in the event of unique circumstances with two-way radios – this occurred in the case of three participants who were ultimately removed from the database. When the final trial block was finished, participants were asked to complete a post-test (the content of which is not relevant to the research questions in this study) and were released after being informed that their payment for participation would be arriving Mean differences on all primary dependent variables for categorical variables not of specific interest to our hypotheses (gender, age) were conducted to rule out any unintended differential effects. Men and women did not differ significantly on any dependent measure, and there were no significant correlations between participant age and the dependent variables. Thus, all analyses beyond this point will collapse across these categorical factors. Table 1 displays means and standard deviations for raw error rate, raw correct decision rate, raw reaction speeds, d', and c, a measure of response bias considered to possess the strongest psychometric properties (See, et al, 1997). These scores are broken down further by exaggeration type and trigram group. Performance statistics for the high- trigram, capital-exaggeration group were noticeably worse than for any other group. Also, d' was lower for all exaggeration types in the high-trigram groups. Finally, it is also interesting that c was higher in the capital-exaggeration group compared to the other Addressing Hypotheses H1 and H2, Figure 3 shows means and 95% confidence intervals for d' and raw error rate across the three text exaggeration types. Separate lines are plotted for trigram group assignments. No mean differences in performance across text exaggeration types were observed for either performance measure, regardless of Table 1 Descriptive Statistics for Performance Measures Within Text Exaggeration and Trigram Groups Measure Exaggeration Trigram Group Low 97.31 95.20 96.90 95.50 98.28 94.02 Correct Decision Rate 94.29 91.00 95.42 90.36 93.23 85.61 Reaction Time 0.94 0.97 0.91 0.95 0.95 1.17 d-prime 3.82 3.16 3.86 3.20 4.02 2.80 Response Bias (c) 0.19 0.18 0.17 0.21 0.33 0.25 Note: Reaction time in seconds per trial. trigram group assignment (all p's > .05). It should be noted that group differences in d' within the high-trigram condition neared significance (p = .063), but in the opposite direction expected such that the capital-exaggeration group performed more poorly than the other two groups. The maximum calculated power for these comparisons was approximately .65. Therefore, Hypotheses H1 and H2, which suggested that task performance would be different in specific directions between text exaggeration groups, was not supported by these data. Addressing Hypothesis H3, Figure 4 displays means and 95% confidence intervals for d' and raw error rate across trigram groups. With respect to d', the differences between trigram groups are notable and significant for all exaggeration conditions (for control group, F(1, 35) = 9.65, p < .01; for color group, F(1, 31) = 7.58, p < .01; for capitals, F(1, 31) = 17.69, p < .01). For raw error rate, the difference between trigram groups is not as dramatic, but is still reliable across exaggeration conditions (F (1, Figure 3. Means and 95% CIs for d' and raw error rate for different orthographic similarities and within text exaggeration conditions. Exaggeration Type Exaggeration Type Figure 4. Means and 95% CIs for d' and raw error rate between text exaggeration conditions and within levels of orthographic similarity. Exaggeration Type Exaggeration Type 101) = 17.01, p < .01). Within exaggeration groups, significant differences were observed between trigram groups for the capital-letter exaggeration condition (F (1,31) = 16.34, p < .01), but not for the other conditions. On the basis of these findings, we can report that Hypothesis H3 received support from these data. To address Hypothesis H4, raw correct decision rate and response bias (c) were examined between trigram groups across the entire sample and within exaggeration conditions. Participants showed greater liberality in terms of overall false alarm responses in the high trigram group (F (1, 101) = 19.96, p < .01), but no group differences in c. The differences in overall false alarm responses were located in the two exaggerated-text conditions (both p's < .02). The maximum calculated power for comparisons involving response bias was approximately .44. Thus, H4 was only partially supported by these The purpose of this study was to determine the effectiveness of text exaggerations on drug name identification and to assess whether the orthographic similarity of the names moderated accurate identification. We used a prime-target name matching task designed to measure the immediate quality of the memory trace and whether text exaggeration improved that quality. The data showed that the text exaggerations, specifically the capitalization of specific letter segments in the word, did result in higher sensitivity (as measured by d') but only under specific circumstances. For the most part, the text exaggerations did not improve accurate performance beyond a control condition with no exaggerations. In addition, orthographic similarity as measured using trigrams (matched pairs of three-letter segments between two words) was a moderator of accurate name identification such that high-trigram names were identified less accurately than low-trigram names. Since 2001, the Food and Drug Administration has suggested that certain drug names should be exaggerated in some way to differentiate them from a potentially confusable similar name (i.e., glipizide and glyburide, Vioxx and Zyvox). Manufacturers have in many cases assented to these changes but there is little evidence available to indicate that these exaggerations have any value with respect to accurate name identification. The present study proposed that if the exaggerations didn't create an initial memory trace that improved name identification in a focused-attention task, then it probably would not improve performance in a real-world setting like a pharmacy. The results of this study pose serious questions regarding the benefits of cognitive comparisons between exaggerated and unexaggerated versions of the same words. Of course, this is an important question because, although the label on the stock bottle is exaggerated, the drug name on the script form or on the computer screen is not. There are some limitations to the current study that should be noted. First, these participants were not pharmacy-trained individuals. Many had probably never seen the preponderance of these drug names before and perhaps didn't even know how to pronounce some of them correctly. Thought this is a potential limitation it also controlled for potential biases. In other words, trained pharmacy personnel could have introduced two biases into the current investigation: increased dependence on phonological rather than graphicological processing of the names (Frost, 1998; Pugh, et al, 1994) and knowledge of which drug names were likely to appear on the screen due to their training. Although the present simulation provides important insight into how novices respond to the exaggerations, the response of trained professionals remains unclear. Given the findings of this study and the safety sensitive nature of the pharmaceutical industry it would seem prudent to examine exaggeration dynamics in a pharmaceutical setting. A second limitation has to do with the structure of the task. In a real pharmacy, names could be compared several times from the time a script is initiated to the point when it is completed. In our task, the prime words were shown for 450ms and could not be reviewed further. Therefore, it is possible that text exaggerations have their benefit later in the prescription process, not right at the beginning or after only one name comparison. On the other hand, the data do indicate that performance stabilized on the task fairly quickly (probably after 120-150 pairs of names), so further exposure to the names may not have made much difference. Indeed research that shows pharmacists to be less accurate in checking scripts that they filled personally compared to scripts that co- workers filled (eg., Grasha, Reilley, Schell, Tranum & Filburn, 2000). In the same way, once the individual has decided that the names match, further checks may be biased by their initial conclusions. Future research should be done to test this possibility. A third limitation is that drug name pairs of moderate trigram similarity were removed to create the two trigram groups. Thus, it remains unclear whether these drug name pairs would be identified at different sensitivities than the high- and low-trigram pairs. Nevertheless, constructing the experiment such that orthographic similarity was an independent grouping variable did provide the benefit of simple and clear tests of the orthographic similarity hypothesis that confirmed much of the work that Lambert and his colleagues did which suggested that trigram similarities were influential on reported medication errors (Lambert, 1997, 2005; Lambert, et al, 2001; Lambert, et al, 1999). Treating similarity as a continuous measure would have complicated the interpretation of the effects of the factor on name identification, and these effects were important given the reliance of Lambert's work on pharmacist self-reports of errors and the likely sample bias that results from such self-reports. Finally, ceiling effects in the dependent measures are always a concern with this sort of study. Examination of the means in Table 1 show that control group accuracy rates are quite high, even after 480 presentations of the drug names. However, we also believe that a gain of even one percentage point of accuracy is practically important when it comes to medical mistakes, given that may extrapolate into millions of actual The results of this study inform our understanding of whether text exaggerations are actually helpful in preventing name-confusion medication errors. Given the findings of the present study it appears that the beneficial effects of name exaggerations may be questionable. More importantly, this study suggests that exaggerations can be detrimental to name identification, especially at high levels of orthographic similarity. More research is needed to make a final determination as to whether the exaggerations are actually preventative of medication errors where drug name confusion may have played a significant role. Carreiras, M., Perea, M. & Grainger, J. (1997). Effects of orthographic neighborhood in recognition: Cross-task comparisons. Journal of Experimental Psychology:Learning, Memory & Cognition, 23, 857-871. Davies, D.R. & Parasuraman, R. (1982). The psychology of vigilance. London: Duncan, J. & Humphreys, G.W. (1989). Visual search and stimulus similarity. Psychological Review, 96, 433-458. Egeth, H.E. & Yantis, S. (1997). Visual attention: Control, representation, and time Annual Review of Psychology, 48, 269-297. Frost, R. (1998). Toward a strong phonological theory of visual word recognition: True issues and false trails. Psychological Bulletin, 123, 71-99. Gage, J. (1995). Colour and culture. In T. Lamb & J. Bourriau (Eds.), Colour, art and science. (pp. 175-193). New York: Cambridge University Press. Grasha, A.F., Reilley, S., Schell, K.L., Tranum, D. & Filburn, J. (2000). A cognitive systems perspective on human performance in the pharmacy: Implications for accuracy, effectiveness, and job satisfaction. (Technical Report 062100-R). Cincinnati, OH: Cognitive-Systems Performance Laboratory, University of Just, M.A. & Carpenter, P.A. (1980). A theory of reading: From eye fixations to Psychological Review, 87 , 329-354. Katz, L. & Frost, S.J. (2001). Phonology constrains the internal orthographic representation. Reading & Writing: An Interdisciplinary Journal, 14, 297-332. Kohn, L.T., Corrigan, J.M. & Donaldson, M.S. (Eds.) (1999). To err is human - building a safer health care system. Washington, DC: National Academy Press. Lambert, B.L. (2005). Designing safe drug names. Drug Safety, 28, 495-512. Lambert, B.L., Chang, K.Y. & Gupta, P. (2003 ). Effects of frequency and similarity neighborhoods on pharmacists' visual perception of drug names. Social Science & Medicine, 57, 1939-1955. Lambert, B.L., Chang, K.Y. & Lin, S.J. (2001 ). Effect of orthographic and phonological similarity on false recognition of drug names. Social Science & Medicine, 52, Lambert, B.L., Chang, K.Y. & Lin, S.J. (2003 ). Immediate free recall of drug names: effects of similarity and availability. American Journal of Health-System Pharmacy, 60, 156-168. Lambert, B.L., Lin, S.J., Chang, K.Y. & Gandhi, S.K. (1999). Similarity as a risk factor in drug-name confusion errors: the look-alike (orthographic) and sound-alike Medical Care, 37, 1214-1225. Lukatela, G. & Turvey, M.T. (1994a). Visual lexical access is initially phonological: 1. Evidence from associative priming by words, homophones and pseudohomophones. Journal of Experimental Psychology: General, 123, Lukatela, G. & Turvey, M.T. (1994b). Visual lexical access is initially phonological: 2. Evidence from phonological priming by homophones and pseudohomophones. Journal of Experimental Psychology: General, 123, 331-353. Perea, M. & Rosa, E. (2002). Does "whole-word shape" play a role in visual word Perception & Psychophysics, 64 , 785-794. Pugh, K.R., Rexer, K. & Katz, L. (1994). Evidence of flexible coding in visual word Journal of Experimental Psychology: Human Perception & Performance, 20, 807-825. See, J.E., Warm, J.S., Dember, W.N. & Howe, S.R. (1997). Vigilance and signal detection theory: An empirical evaluation of five measures of response bias. Human Factors, 39(1), 14-29. Stone, G.O. & Van Orden, G.C. (1993). Strategic control of processing in word Journal of Experimental Psychology: Human Perception & Performance, 19, 744-774. Treisman, A.M. & Gelade, G. (1980). A feature-integration theory of attention. Cognitive Psychology, 12(1), 97-136. Wickens, C.D. & Hollands, J.G. (2000). Engineering psychology and human performance. (3rd Ed.). Upper Saddle River, NJ: Prentice Hall. Yantis, S. (1998). Control of visual attention. In H. Pashler (Ed.), Attention. (pp. 223- 256). Hove, UK: Psychology Press. Yantis, S. & Hillstrom, A.P. (1994). Stimulus-driven attentional capture: Evidence from equiluminant visual objects. Journal of Experimental Psychology: Human Perception & Performance, 20, 95-107. Drug Name Pairs and Trigram Scores Used In This Study Trigram x 100
prednisolone prednisone Dynacin Dynacirc amitriptylene nortriptylene cyclosporine cycloserine clonazepate lorazepam Cardizem Cardiem Avandamet Avandia daunorubicin doxorubicin ephedrine epinephrine fluoxetine paroxetine chlorpromazine chlorpropamide clonazepam clonazepate nicardipine nifedipine tolazamide tolbutamide acetohexamide acetazolamide clomiphene clomipramine dobutamine dopamine hydralazine hydroxyzine vinblastine vincristine lorazepam clonazepam Adderall Inderal Glucophage Glucovance Claravix Clarinex Indinavir Denavir Atropine Atrovent chlorpromazine promethazine dobutamine tolbutamide Accupril Accuretic levothroid Levoxyl Nifediac nifedipine chlorpromazine chlorzoxazone Lamictal Lamasil fluoxetine fluphenazine sulfadiazine sulfisoxazole bupropion budeprion glipizide glyburide Levaquin Levothroid papaverine paroxetine metformin metronidazole promethazine propoxyphene Alupent Atrovent Celebrex Cerebyx Lovenox Lotrenox bupropion buspirone budeprion buspirone prednisolone testosterone Lamictal Lomotil Methylin Motofen

Source: http://www.angelo.edu/faculty/kschell/downloads/exagg-trigramtech.pdf