Task-specific performance effects with different ... - Semantic Scholar
Applied Ergonomics xxx (2013) 1e6
Contents lists available at ScienceDirect
Applied Ergonomics journal homepage: www.elsevier.com/locate/apergo
Task-specific performance effects with different numeric keypad layouts Jenny T. Armand a, Thomas S. Redick b, *, Joan R. Poulsen a a b
Indiana University Purdue University Columbus, United States Department of Psychological Sciences, Purdue University, 703 Third Street, West Lafayette, IN 47907, United States
a r t i c l e i n f o
a b s t r a c t
Article history: Received 23 August 2012 Accepted 13 November 2013
Two commonly used keypad arrangements are the telephone and calculator layouts. The purpose of this study was to determine if entering different types of numeric information was quicker and more accurate with the telephone or the calculator layout on a computer keyboard numeric keypad. Fifty-seven participants saw a 10-digit numeric stimulus to type with a computer number keypad as quickly and as accurately as possible. Stimuli were presented in either a numerical [1,234,567,890] or phone [(123) 4567890] format. The results indicated that participants’ memory of the layout for the arrangement of keys on a telephone was significantly better than the layout of a calculator. In addition, the results showed that participants were more accurate when entering stimuli using the calculator keypad layout. Critically, participants’ response times showed an interaction of stimulus format and keypad layout: participants were specifically slowed when entering numeric stimuli using a telephone keypad layout. Responses made using the middle row of keys were faster and more accurate than responses using the top and bottom row of keys. Implications for keypad design and cell phone usage are discussed. Ó 2013 Elsevier Ltd and The Ergonomics Society. All rights reserved.
Keywords: Numeric keypads Task-specific performance Interference
1. Introduction 1.1. Background Many daily activities require people to enter numeric information using a keypad. For example, telephones, smart phones, calculators, computers, automated-teller-machines (ATMs), and home alarm systems all have keypads that people frequently use to enter strings of digits. Likewise, there is a wide variety of numerical information that is entered into keypad devices, including phone numbers, account numbers, birthdates, and currency values. Two observations about keypads motivated the current research. First, many people are not explicitly aware that the number keypad layout differs across devices such as telephones and calculators (Rinck, 1999). Second, people typically interact with many of the above devices with conflicting numeric keypad layouts on a daily basis, yet intuitively performance does not seem to show drastic consequences of keypad arrangement inconsistency. Additionally, with more technological devices pervading common human experience, research on layouts of keypads is as relevant today as ever. With hundreds of laptops and telephones, dozens of tablets
* Corresponding author. Tel.: þ1 765 494 5132. E-mail address: [email protected] (T.S. Redick).
and other devices on the market, ergonomic keyboard and keypad layout is a concern for buyers and designers. Despite the frequent use of telephones and calculator/computer keypads, individuals exhibit surprisingly poor memory when explicitly asked to reproduce the layout of numbers on these devices (Fig. 1). For example, Rinck (1999, Experiment 1) tested college students on their ability to correctly place the digits on a blank sheet using their memory for the layout of either a telephone or a calculator. Accuracy for the correct placements of the digits 1 to 9 was 78% for the telephone versus 48% for the calculator layout. That is, approximately one-half of the college students tested could not explicitly recall the location of the digits on a calculator keypad, as presented in Fig. 1. In addition, Rinck found that when participants made errors, especially on the calculator layout, they reversed the layout (for example, entered the layout for numbers on a telephone when they were supposed to enter the layout of a calculator). Jones and Martin (2009) also found relatively low recall for the calculator layout in their sample of college students. Only about 25% of the participants in the control group, which was not provided with any strategies or explicit instructions, reproduced the calculator layout in its entirety with all of the digits in the correct location. In contrast, the same participants reproduced the telephone layout with near-perfect accuracy. Rinck (1999) provided evidence that there is interference that occurs when accessing the mental representation of keypad
0003-6870/$ e see front matter Ó 2013 Elsevier Ltd and The Ergonomics Society. All rights reserved. http://dx.doi.org/10.1016/j.apergo.2013.11.008
Please cite this article in press as: Armand, J.T., et al., Task-specific performance effects with different numeric keypad layouts, Applied Ergonomics (2013), http://dx.doi.org/10.1016/j.apergo.2013.11.008
2
J.T. Armand et al. / Applied Ergonomics xxx (2013) 1e6
Fig. 1. From “Memory for everyday objects: Where are the digits on numerical keypads?” by Rinck (1999), Applied Cognitive Psychology, 13, p. 330. Copyright 1999 by John Wiley & Sons, Ltd.
layouts. However, in practice most individuals are able to switch between devices with relative ease. For example, consider the many work-related tasks completed by a pharmacy technician. She may spend much of her time at work dialing phone numbers into a phone to call customers to tell them their prescriptions are ready, insurance companies to clarify benefit information, or doctors’ offices to verify dosage and medical history. However, she likely also spends a lot of time entering numeric information, including customers’ phone numbers and insurance account numbers, into the computer using the keypad (which is arranged like a calculator). Another example of an occupation where the user enters a variety of different numeric information is a ticket agent for an airline. When checking in for a flight at the airport, an airline ticket agent may use the numeric keypad on a desktop keypad (which matches a calculator layout) to enter the passenger’s confirmation number, airline rewards number, driver’s license identification number, phone number, birthdate, and credit card number. With the advent of alternative telecommunication options, voice-over-internet-protocol users typically use computer numeric keypads to dial telephone numbers. Students working in a research lab often enter various types of numeric information into statistical software packages using the numeric keypad on a desktop computer. Therefore, the purpose of the current research is to determine if entering different types of numeric information is quicker and more accurate with the telephone or the calculator layout. Previous research on this topic has been inconsistent and may be outdated. Deininger (1960) conducted research at Bell Laboratories testing different potential layouts for new touchtone telephone devices. In one experiment, employees entered random phone numbers on 16 possible different layouts, including the then-current arrangement of adding machines (calculators) and the present-day arrangement of telephones (not currently in use). Deininger reported that participants were slightly faster using the telephone (4.92 s) versus the calculator (5.08 s) layout, but no accuracy information was reported. Conrad and Hull (1968) tested participants with no experience using a touchtone telephone or adding machine, and examined their performance at entering random 8-digit strings. The group of participants that used the telephone layout was more accurate than the group that used the calculator layout, although the groups did not statistically differ in the speed of correct digit entry. These results correspond with research assessing users’ preferences for potential keypad layouts. The majority of naïve and experienced participants indicate a preference for the telephone versus the calculator layout (Lutz and Chapanis, 1955). This preference is especially strong when participants are asked to imagine scenarios in which they would enter telephone numbers, but somewhat surprisingly, the preference for the telephone layout
versus the calculator layout also applies to other tasks such as entering personal identification numbers and single digits (Straub and Granaas, 1993). It should be noted that much of this research was conducted at a time when computer keypad usage was rare. Indeed, prior to the 1980s, personal computers were not commonly used in homes and businesses, thus limiting people’s exposure to the calculator keypad layout relative to today. Marteniuk et al. (1996) tested college students on their ability to enter four-digit strings, seven-digit strings, and seven-digit telephone numbers using either a calculator or a telephone keypad layout. Marteniuk et al. focused on the effect of the placement of the zero-digit key (above or below the other keys) across keypad layout types. Although Marteniuk et al. concluded, against their predictions, that “no interactions between the task and the keypad arrangement were found” (p. 325), in the article they only reported the separate analyses conducted for the different stimulus formats. Without the benefit of significance testing, visual inspection of the data indicates that participants were faster and more accurate entering telephone numbers compared to seven-digit strings without telephone formatting. However, they reported no significant effects of keypad layout on either accuracy or the total amount of time to enter the string of digits. The three studies just discussed provide contradictory results. The conclusion from two studies (Conrad and Hull, 1968; Deininger, 1960) is that there is a slight advantage for the telephone layout no matter if the user is entering telephone or numeric information. However, Marteniuk et al. (1996) found no significant effects of keypad layout. The participants in Deininger (1960) and Conrad and Hull (1968) are somewhat unique in that they participated in the research with no history or exposure to the telephone layout, due in part to limited availability of technology in the 1960s. In contrast, in today’s society in many developed countries, virtually all young adults have extensive exposure and practice using telephones and calculators. Thus, the participants in Marteniuk et al. (1996) may not have shown an advantage of either type of keypad arrangement because of their experience with both telephone and calculator layouts that the participants in the earlier studies did not have. 1.2. Current research The current study combines elements of both Rinck (1999, Experiment 1) and Marteniuk et al. (1996), but attempted to expand upon their studies in multiple ways. First, in contrast to Rinck (1999), we tested participants’ knowledge of telephone and calculator layouts using a within-subjects design. Additionally, a potentially important aspect of the stimuli that Marteniuk et al. used was that the numeric stimuli were presented as one continuous sequence of seven digits (e.g., 1947294), whereas the phone numbers were presented with a hyphen (e.g., 396e8142). Because of concerns that the phone stimuli were easier for participants in Marteniuk et al. to chunk in short-term memory, in the current research we added commas to the numeric digit strings to provide comparable chunks for participants to utilize. In addition, we tested all participants on their ability to enter all 10 digits on the keypad on each trial, whether as a numerical [1,234,567,890] or phone [(123) 456-7890] presentation. This decision provided two advantages over Marteniuk et al. First, by including all possible digits once in each stimulus, we controlled across trials for effects of key position upon performance (see below). Second, with the widespread use of cell phones, many dialed numbers today in the United States are ten digits, and thus the current stimuli may be more similar to the numbers dialed in daily telephone usage. Finally, in the current research we tested over twice as many participants and administered twice as many trials per condition than Marteniuk et al., because some trends that were not statistically significant in
Please cite this article in press as: Armand, J.T., et al., Task-specific performance effects with different numeric keypad layouts, Applied Ergonomics (2013), http://dx.doi.org/10.1016/j.apergo.2013.11.008
J.T. Armand et al. / Applied Ergonomics xxx (2013) 1e6
their study might have been due to insufficient power (N ¼ 24 participants in Marteniuk et al.). We predicted that, as in Rinck (1999), participants would be less accurate to explicitly recall the spatial layout of calculators compared to telephones. For the computerized task described below, when participants were asked across four different blocks to enter numeric or phone stimuli into either a calculator or telephone keypad, we predicted an overall advantage for entering phone stimuli, based on Marteniuk et al. (1996). However, in contrast to that research, we also predicted that we would observe an interaction of stimulus type and layout type, such that participants would be less accurate and/or slower in conditions of mismatching stimuli and layouts (e.g., numeric stimuli into a telephone keypad). Our interest in potential task-specific interactions of stimulus type and keypad layout motivated our final prediction. As stated previously, the middle row (4, 5, 6) on the keypad is the same for both telephone and calculator layouts. However, the top and bottom rows are mapped in an opposite fashion in the two layouts (Fig. 1). If, for example, users experience interference from the mental representation of the telephone keypad layout while using the calculator layout, then performance should be worse specifically for responses using the top and bottom rows of the keypad. Thus, we examined performance as a function of the row that was used for each of the 10 responses made on a given trial. 2. Method 2.1. Participants Participants for this study were students or community volunteers recruited from the campus of a small Midwestern university. Sixty-two participants completed the study; after removing participants who did not comply with instructions or had extremely low accuracy on the computerized task (see below), the final sample was comprised of 57 participants (40 female, 17 male), with a mean age of 26.9 (12.4) years old. We specify our rule for terminating data collection based on a recent recommendation by Simmons et al. (2011). Specifically, the first-author conducted the study as part of her capstone research project for graduation, supervised by the second- and third-author. The prescribed rule for all students completing capstone research projects during the academic year was to recruit and test a minimum of 60 participants, and data collection had to be completed by a certain date specified in advance. 2.2. Procedure All participants first completed a memory test similar to Rinck (1999). Participants were asked, in order, to provide the digit layout for a telephone, a calculator, and an ATM. Participants were given feedback after completing the memory test. Performance on the memory test was scored as the accuracy of participants’ recall. If participants reproduced the keypad layout in its entirety with no mistakes in the placement of the numbers, their answer was considered correct (see Fig. 1). After entering demographic information, participants then completed the main part of the experiment. Across four blocks, participants took part in a repeated-measures 2 (numeric or telephone number stimulus) 2 (calculator or phone keypad layout) experiment in which they participated in all four conditions: (a) entering phone stimuli [(123) 456-7890] using a telephone layout; (b) entering numeric stimuli (1,234,567,890) using a telephone layout; (c) entering phone stimuli using a calculator layout; and (d) entering numeric stimuli using a calculator layout. Participants were randomly assigned to one of four possible task orders, which
3
were counterbalanced using a Latin square design. The counterbalancing of the order of task conditions was done to prevent any practice effects that might be observed if all participants had used one keyboard format before the other, or entered one stimulus type before the other. The particular task orders were chosen to: (a) have each condition appear as the first, second, third, or fourth condition across participants; and (b) allow the experimenter to switch the keyboard only once for each participant, halfway through the computerized task to minimize disruption. Stimuli were 10-digit sequences that were either presented in the numerical [1,234,567,890] or phone [(123) 456-7890] format. All stimuli were randomly generated before the experiment with the constraints that: (a) every 0e9 digit appeared in each stimulus; (b) no digit appeared more than once within a stimulus; and (c) the digit 0 could not appear as the first number in the sequence. Note that the first two constraints produced stimuli that differed from Martenuik et al. (1996), in that their seven-digit stimuli did not contain all 10 possible responses and digits could repeat within a stimulus. Stimuli were presented in the center of the screen, and formatted so that the 10 digits appeared in the same 10 absolute locations onscreen regardless of the stimulus type. For each of the four stimulus-layout conditions, participants completed a total of 50 trials. The first 10 trials were practice, and were indicated as such to the participants. During the 10 practice trials, participants received performance feedback. These trials were not analyzed, but ensured that participants understood the instructions. Participants then completed 40 trials without feedback that were analyzed. Each trial started with a fixation screen until the participant pressed the spacebar with his or her nondominant hand. The 10-digit stimulus then appeared onscreen, and the participants were instructed to type the numbers into the keypad as quickly and accurately as possible. A cursor appeared under the current digit in the sequence, and advanced to the next digit after the response was keyed. Once the 10-digit number was entered, the next fixation screen appeared. Both accuracy and reaction time were measured. Accuracy for each stimulus-by-layout condition on the computerized task was derived by obtaining the mean of each participant’s mean accuracy, where all 10 digits must be entered in the correct order to be considered correct. Response time (RT) for each stimulus-by-layout condition on the computerized task was derived by obtaining the mean of each participant’s mean time to type the entire 10 digit string, from the moment the stimuli appeared onscreen until the moment the participant pressed the 10th key. As is the common practice in psychological research, we only included correct responses in the analyses of the RT data. 2.3. Materials For the initial memory test, participants used pencil and paper to record the answers. For the computerized task, participants completed a program created in E-Prime 2.0 (Psychology Software Tools, Pittsburgh, PA). Stimuli were presented visually in the center of the computer monitor screen, and participants used the number keypad of the computer keyboard to enter their responses. For the calculator layout, the standard number keypad layout was used. For the telephone layout, the buttons for the top and bottom rows of the keypad were exchanged. 2.4. Analyses Repeated-measures analyses of variance (ANOVAs) were used for each of the dependent variables in the study. An alpha of .05 was used for significance testing, except where noted in follow-up
Please cite this article in press as: Armand, J.T., et al., Task-specific performance effects with different numeric keypad layouts, Applied Ergonomics (2013), http://dx.doi.org/10.1016/j.apergo.2013.11.008
4
J.T. Armand et al. / Applied Ergonomics xxx (2013) 1e6
analyses to control for familywise error. Effect sizes are provided using partial eta-squared ðh2p Þ. 3. Results 3.1. Memory test Results from the explicit memory test (N ¼ 56 due to incomplete data from one participant) were analyzed with a repeatedmeasures ANOVA with the factor Layout (Telephone, Calculator, ATM). The main effect of Layout was significant, F(2, 110) ¼ 20.75, p < .001, h2p ¼ .274. Follow-up paired t-tests, using a Bonferronicorrected a (.05/3 ¼ .016), indicated that participants were more accurate in their recall of the telephone (95%) layout than either the ATM (68%) or calculator (41%) layout (all p’s < .013). 3.2. Computerized task Overall accuracy and RT results from the computerized task were analyzed with a repeated-measures ANOVA with the factors Stimulus (Phone, Numeric) and Layout (Telephone, Calculator). Looking first at accuracy (Fig. 2), the main effect of Layout was significant, F(1, 56) ¼ 5.98, p ¼ .018, h2p ¼ .097, with participants more accurate using the calculator layout (87%) versus the telephone layout (83%). The main effect of Stimulus was not significant, F(1, 56) ¼ .11, p ¼ .746, h2p ¼ .002, and neither was the interaction, F(1, 56) ¼ 1.26, p ¼ .267, h2p ¼ .022. The results for the mean RTs on correct trials are shown in Fig. 3. The significant main effects of Stimulus, F(1, 56) ¼ 39.30, p < .001, h2p ¼ .412, and Layout, F(1, 56) ¼ 39.96, p < .001, h2p ¼ .416, were qualified by a significant interaction, F(1, 56) ¼ 18.44, p < .001, h2p ¼ .248. As can be seen in Fig. 3, although overall participants were faster when using the calculator keypad and entering telephone numbers, the interaction was due to the pronounced slowing that participants showed when entering numeric stimuli using the telephone layout. When using the calculator keypad, participants were on average 191 ms slower to enter numeric versus phone stimuli. In contrast, when using the telephone keypad, participants were on average 537 ms slower to enter numeric versus phone stimuli. The results thus far have focused on performance on each stimulus; that is, examining accuracy and RT for the entire 10-digit number string presented. This represents the accuracy and RTs of the total of the 10 keystrokes per stimulus presented. In the
Fig. 2. Mean accuracy for the computerized task with error bars representing 1 95% confidence intervals for within-subjects designs (Loftus and Masson, 1994). Phone: Telephone stimulus number format; Number: Numerical stimulus number format; TEL: Telephone keypad design; CAL: Calculator keypad design.
Fig. 3. Mean correct response times for the computerized task with error bars representing 1 95% confidence intervals for within-subjects designs (Loftus and Masson, 1994). Phone: Telephone stimulus number format; Number: Numerical stimulus number format; TEL: Telephone keypad design; CAL: Calculator keypad design.
subsequent analyses, the accuracy and RTs for individual keypresses were examined. This aimed to measure the accuracy and speed of each individual keypress within each 10-digit number string. These were examined using a repeated-measures ANOVA, grouped according to its location in the top, middle, or bottom rows (Table 1).1 Looking first at accuracy, the effect of Row was significant, F(2, 112) ¼ 46.57, p < .01, h2p ¼ .454. For mean correct RTs, the effect of Row was significant, F(2, 112) ¼ 39.48, p < .01, h2p ¼ .414. Follow-up paired t-tests, using a Bonferroni-corrected a (.05/ 3 ¼ .016), indicated that responses using the middle row keys were both more accurate and faster than responses using the top or bottom row keys (all p’s < .008).
4. Discussion Our study was motivated by the observation that people seem to switch between tasks with relative ease on devices using numeric keypads, such as telephones and calculators, despite previously documented inaccuracy in their explicit memories for the digit layouts on the keypads. With the ever-increasing emergence of new devices, potentially with differing keypad layouts, it becomes an important issue to understand empirically how easily (or not) everyday people switch keypad layouts. Replicating Rinck (1999), we found that participants (mostly college students) were more accurate in their memory for telephone versus calculator layouts. In addition, in our study we wanted to address whether users type specific types of numeric information more accurately and more quickly using specific types of numeric keypad layouts. We found that participants were more accurate entering phone numbers no matter what the keypad layout. The phone number advantage is consistent with Marteniuk et al. (1996), but the result here cannot be solely attributed to differences in chunking, as the numeric stimuli were also presented in discrete chunks of information. Most importantly, participants’ time to type in the numerical information showed an interaction of stimulus format and keypad layout: participants were specifically slowed when entering numeric stimuli using a telephone keypad layout. Increased errors and slower RTs were observed specifically for responses using the top and bottom row keys of the keypad relative to the middle row keys,
1 The factors of Stimulus and Layout are not reported here, because the results are redundant with the analyses presented above (e.g., the RT for all 10 responses is the sum of the RT for each of the 10 responses).
Please cite this article in press as: Armand, J.T., et al., Task-specific performance effects with different numeric keypad layouts, Applied Ergonomics (2013), http://dx.doi.org/10.1016/j.apergo.2013.11.008
J.T. Armand et al. / Applied Ergonomics xxx (2013) 1e6 Table 1 Mean accuracy and correct response time for individual key responses as a function of keypad row. Variables
Accuracy (%)
Correct RT (ms)
Top row Middle row Bottom row
96.5 (2.4) 98.4 (1.4) 97.2 (2.1)
681.31 (206.07) 629.67 (175.38) 692.82 (208.17)
Note. Numbers in parentheses are standard deviations for each condition.
indicating interference from a competing mental representation of the other keypad layout. Our results suggest that, in contrast to Marteniuk et al. (1996), users show costs from entering specific types of numeric information using certain configurations of keypad layouts. In our study, the effect was observed in terms of RTs. Although the difference between entering phone and numeric stimuli on a telephone keypad was 537 ms, this difference was observed for entering a single sequence of 10 digits. If a user were to enter multiple 10-digit strings consecutively, as in the example given earlier for a pharmacy technician, that half-second difference could quickly add up and impact productivity. An example of a daily activity that epitomizes the specifically slowed combination of entering a numeric stimulus into a telephone keypad is that of a consumer using the calculator tool on a cell phone. The current results indicate that if a user is planning to conduct numerous calculations, he or she should instead keep a pocket calculator nearby to maximize efficiency, or use a smartphone application with a proper calculator layout. We found that participants were faster and more accurate when using the calculator layout, whether entering phone or numeric stimuli. The speed advantage for the calculator layout observed here may have been driven by the use of a computer keyboard keypad as the response device. Participants with previous computer experience likely have already established the motor memories that allow them to locate and type in number information on this device (Rinck and Ellwart, 2004). By switching the top and bottom rows of the keyboard keypad to create the telephone keypad layout, participants likely dealt with more motor memory interference in this condition than vice-versa. Future research using a touchtone or cell phone as the response device, and alternating the buttons as necessary to create a calculator layout, could confirm this idea. Our results also have implications for research studies that use existing keypad layouts for participants to enter numeric information. For example, Salvucci (2005) investigated participants’ ability to multitask using a desktop computer driving simulator. While participants drove the simulator, they also entered “four of their most familiar 10-digit phone numbers” (p. 479) using the computer keyboard numeric keypad, which was the standard calculator layout instead of a telephone layout (D. D. Salvucci, personal communication, 5/31/13). Of note, Salvucci (2005) eliminated 30% of their participants from the final sample, because those 30% were too inaccurate while entering the phone numbers, indicating these participants may have had difficulty dealing with the interference caused by using a calculator keypad design to enter telephone numbers. This example illustrates a potential source of unwanted variance in participants’ data, if a research study uses the keypad as the input device without taking into account differences between the positioning of keys in the telephone and calculator layouts. In addition, the observation of faster RTs for the calculator layout contradicts Conrad and Hull (1968), who strongly advocated for the use of the telephone layout for all numeric data entry. As mentioned previously, the participants in Conrad and Hull (1968)
5
had no experience with either touchtone telephones or calculators, which may account for the differences between their results and ours. An interesting factor for future research is how changing computing technologies affect human performance. As more consumers opt for laptops or tablets as their main computing devices, instead of traditional desktops, it’s likely that they will have less experience with the numeric keypads present on full keyboards. Thus, it may not be useful to advocate a specific numeric keypad layout as being the best or most efficient. Instead, users’ performance will be determined not only by the task and layout, but also by their history of using different keypad layouts. Although our study contributes to the understanding of how keypad layout can affect accuracy and speed of numeric data entry, there are a few limitations to this study. First, it would be ideal to have a larger sample of participants. The sample size was adequate for significance testing, and larger than previous research on this topic (e.g., Marteniuk et al., 1996), but for the sake of increased external validity, a broader age-group and/or sample with expertise using specific layouts would be interesting for future research. In particular, future work could test competing numeric keypad arrangements with users who have extensive practice using specific numeric keypad arrangements (e.g., accountants) to see if the same effects are obtained as in the non-expert subjects tested here. Second, a more seamless transition between keypad layouts would be valuable for testing quick transitions between keypads. This would help to examine the phenomena of task-switching versus motor memory of keypads. Because our RT measure included aspects of both cognitive decision time and motor movement time, future work examining the effect of novel versus already known phone numbers and digit sequences could be informative. Additionally, the QWERTY keyboard layout of numbers should be tested (with the number keys 1-0 all adjacent in one row), because many laptops and keyboard-free devices (e.g., touchscreen keyboards, tablets) have this layout as the only option for number entry.
5. Conclusion In conclusion, we found evidence that despite users’ relatively inaccurate memories for the layout of a calculator, entering numeric information into a calculator keypad layout was superior to using a telephone keypad layout. In addition, although participants were faster to enter phone numbers compared to other numeric stimuli on either type of layout, they were particularly slowed when entering numeric stimuli into a telephone keypad. Responses made using the middle row of keys were faster and more accurate than responses using the top and bottom row of keys, providing evidence that interference from competing mental representations affects user performance during numeric data entry. Future research should address a broader age group, and include participants with keypad expertise. As more electronic devices come onto the market and are used in number entry at various jobs where accuracy and time are critical, it becomes increasingly important to understand how these different layouts impact both accuracy and speed of number entry.
Acknowledgments Correspondence concerning this article should be sent to Thomas S. Redick, Purdue University, 703 Third Street, West Lafayette, IN 47907 (email to [email protected]). While writing this article, TSR was supported by the Office of Naval Research (Award # N00014-12-1-1011).
Please cite this article in press as: Armand, J.T., et al., Task-specific performance effects with different numeric keypad layouts, Applied Ergonomics (2013), http://dx.doi.org/10.1016/j.apergo.2013.11.008
6
J.T. Armand et al. / Applied Ergonomics xxx (2013) 1e6
References Conrad, R.R., Hull, A.J., 1968. The preferred layout for numeral data-entry keysets. Ergonomics 11, 165e173. Deininger, R.L., 1960. Human factors engineering studies of the design and use of push button telephone tests. Bell Syst. Tech. J. 39, 995e1012. Jones, G.V., Martin, M., 2009. Spatial recall improved by retrieval enactment. Psychon. Bull. Rev. 16, 524e528. Loftus, G.R., Masson, M.E.J., 1994. Using confidence intervals in within-subject designs. Psychon. Bull. Rev. 1, 476e490. Lutz, M., Chapanis, A., 1955. Expected locations of digits and letters on ten-button keysets. J. Appl. Psychol. 39, 314e317.
Marteniuk, R.G., Ivens, C.J., Brown, B.E., 1996. Are there specific performance effects for differently configured numeric keypads? Appl. Ergon. 27, 321e325. Rinck, M., 1999. Memory for everyday objects: where are the digits on numerical keypads? Appl. Cogn. Psychol. 13, 329e350. Rinck, M., Ellwart, T., 2004. Repetition priming during the use and recall of numerical keypads. Eur. J. Cogn. Psychol. 16, 841e861. Salvucci, D.D., 2005. A multitasking general executive for compound continuous tasks. Cogn. Sci. 29, 457e492. Simmons, J.P., Nelson, L.D., Simonsohn, U., 2011. False-positive psychology: undisclosed flexibility in data collection and analysis allows presenting anything as significant. Psychol. Sci. 22, 1359e1366. Straub, H.R., Granaas, M.M., 1993. Task-specific preference for numeric keypads. Appl. Ergon. 24, 289e290.
Please cite this article in press as: Armand, J.T., et al., Task-specific performance effects with different numeric keypad layouts, Applied Ergonomics (2013), http://dx.doi.org/10.1016/j.apergo.2013.11.008
Contents lists available at ScienceDirect
Applied Ergonomics journal homepage: www.elsevier.com/locate/apergo
Task-specific performance effects with different numeric keypad layouts Jenny T. Armand a, Thomas S. Redick b, *, Joan R. Poulsen a a b
Indiana University Purdue University Columbus, United States Department of Psychological Sciences, Purdue University, 703 Third Street, West Lafayette, IN 47907, United States
a r t i c l e i n f o
a b s t r a c t
Article history: Received 23 August 2012 Accepted 13 November 2013
Two commonly used keypad arrangements are the telephone and calculator layouts. The purpose of this study was to determine if entering different types of numeric information was quicker and more accurate with the telephone or the calculator layout on a computer keyboard numeric keypad. Fifty-seven participants saw a 10-digit numeric stimulus to type with a computer number keypad as quickly and as accurately as possible. Stimuli were presented in either a numerical [1,234,567,890] or phone [(123) 4567890] format. The results indicated that participants’ memory of the layout for the arrangement of keys on a telephone was significantly better than the layout of a calculator. In addition, the results showed that participants were more accurate when entering stimuli using the calculator keypad layout. Critically, participants’ response times showed an interaction of stimulus format and keypad layout: participants were specifically slowed when entering numeric stimuli using a telephone keypad layout. Responses made using the middle row of keys were faster and more accurate than responses using the top and bottom row of keys. Implications for keypad design and cell phone usage are discussed. Ó 2013 Elsevier Ltd and The Ergonomics Society. All rights reserved.
Keywords: Numeric keypads Task-specific performance Interference
1. Introduction 1.1. Background Many daily activities require people to enter numeric information using a keypad. For example, telephones, smart phones, calculators, computers, automated-teller-machines (ATMs), and home alarm systems all have keypads that people frequently use to enter strings of digits. Likewise, there is a wide variety of numerical information that is entered into keypad devices, including phone numbers, account numbers, birthdates, and currency values. Two observations about keypads motivated the current research. First, many people are not explicitly aware that the number keypad layout differs across devices such as telephones and calculators (Rinck, 1999). Second, people typically interact with many of the above devices with conflicting numeric keypad layouts on a daily basis, yet intuitively performance does not seem to show drastic consequences of keypad arrangement inconsistency. Additionally, with more technological devices pervading common human experience, research on layouts of keypads is as relevant today as ever. With hundreds of laptops and telephones, dozens of tablets
* Corresponding author. Tel.: þ1 765 494 5132. E-mail address: [email protected] (T.S. Redick).
and other devices on the market, ergonomic keyboard and keypad layout is a concern for buyers and designers. Despite the frequent use of telephones and calculator/computer keypads, individuals exhibit surprisingly poor memory when explicitly asked to reproduce the layout of numbers on these devices (Fig. 1). For example, Rinck (1999, Experiment 1) tested college students on their ability to correctly place the digits on a blank sheet using their memory for the layout of either a telephone or a calculator. Accuracy for the correct placements of the digits 1 to 9 was 78% for the telephone versus 48% for the calculator layout. That is, approximately one-half of the college students tested could not explicitly recall the location of the digits on a calculator keypad, as presented in Fig. 1. In addition, Rinck found that when participants made errors, especially on the calculator layout, they reversed the layout (for example, entered the layout for numbers on a telephone when they were supposed to enter the layout of a calculator). Jones and Martin (2009) also found relatively low recall for the calculator layout in their sample of college students. Only about 25% of the participants in the control group, which was not provided with any strategies or explicit instructions, reproduced the calculator layout in its entirety with all of the digits in the correct location. In contrast, the same participants reproduced the telephone layout with near-perfect accuracy. Rinck (1999) provided evidence that there is interference that occurs when accessing the mental representation of keypad
0003-6870/$ e see front matter Ó 2013 Elsevier Ltd and The Ergonomics Society. All rights reserved. http://dx.doi.org/10.1016/j.apergo.2013.11.008
Please cite this article in press as: Armand, J.T., et al., Task-specific performance effects with different numeric keypad layouts, Applied Ergonomics (2013), http://dx.doi.org/10.1016/j.apergo.2013.11.008
2
J.T. Armand et al. / Applied Ergonomics xxx (2013) 1e6
Fig. 1. From “Memory for everyday objects: Where are the digits on numerical keypads?” by Rinck (1999), Applied Cognitive Psychology, 13, p. 330. Copyright 1999 by John Wiley & Sons, Ltd.
layouts. However, in practice most individuals are able to switch between devices with relative ease. For example, consider the many work-related tasks completed by a pharmacy technician. She may spend much of her time at work dialing phone numbers into a phone to call customers to tell them their prescriptions are ready, insurance companies to clarify benefit information, or doctors’ offices to verify dosage and medical history. However, she likely also spends a lot of time entering numeric information, including customers’ phone numbers and insurance account numbers, into the computer using the keypad (which is arranged like a calculator). Another example of an occupation where the user enters a variety of different numeric information is a ticket agent for an airline. When checking in for a flight at the airport, an airline ticket agent may use the numeric keypad on a desktop keypad (which matches a calculator layout) to enter the passenger’s confirmation number, airline rewards number, driver’s license identification number, phone number, birthdate, and credit card number. With the advent of alternative telecommunication options, voice-over-internet-protocol users typically use computer numeric keypads to dial telephone numbers. Students working in a research lab often enter various types of numeric information into statistical software packages using the numeric keypad on a desktop computer. Therefore, the purpose of the current research is to determine if entering different types of numeric information is quicker and more accurate with the telephone or the calculator layout. Previous research on this topic has been inconsistent and may be outdated. Deininger (1960) conducted research at Bell Laboratories testing different potential layouts for new touchtone telephone devices. In one experiment, employees entered random phone numbers on 16 possible different layouts, including the then-current arrangement of adding machines (calculators) and the present-day arrangement of telephones (not currently in use). Deininger reported that participants were slightly faster using the telephone (4.92 s) versus the calculator (5.08 s) layout, but no accuracy information was reported. Conrad and Hull (1968) tested participants with no experience using a touchtone telephone or adding machine, and examined their performance at entering random 8-digit strings. The group of participants that used the telephone layout was more accurate than the group that used the calculator layout, although the groups did not statistically differ in the speed of correct digit entry. These results correspond with research assessing users’ preferences for potential keypad layouts. The majority of naïve and experienced participants indicate a preference for the telephone versus the calculator layout (Lutz and Chapanis, 1955). This preference is especially strong when participants are asked to imagine scenarios in which they would enter telephone numbers, but somewhat surprisingly, the preference for the telephone layout
versus the calculator layout also applies to other tasks such as entering personal identification numbers and single digits (Straub and Granaas, 1993). It should be noted that much of this research was conducted at a time when computer keypad usage was rare. Indeed, prior to the 1980s, personal computers were not commonly used in homes and businesses, thus limiting people’s exposure to the calculator keypad layout relative to today. Marteniuk et al. (1996) tested college students on their ability to enter four-digit strings, seven-digit strings, and seven-digit telephone numbers using either a calculator or a telephone keypad layout. Marteniuk et al. focused on the effect of the placement of the zero-digit key (above or below the other keys) across keypad layout types. Although Marteniuk et al. concluded, against their predictions, that “no interactions between the task and the keypad arrangement were found” (p. 325), in the article they only reported the separate analyses conducted for the different stimulus formats. Without the benefit of significance testing, visual inspection of the data indicates that participants were faster and more accurate entering telephone numbers compared to seven-digit strings without telephone formatting. However, they reported no significant effects of keypad layout on either accuracy or the total amount of time to enter the string of digits. The three studies just discussed provide contradictory results. The conclusion from two studies (Conrad and Hull, 1968; Deininger, 1960) is that there is a slight advantage for the telephone layout no matter if the user is entering telephone or numeric information. However, Marteniuk et al. (1996) found no significant effects of keypad layout. The participants in Deininger (1960) and Conrad and Hull (1968) are somewhat unique in that they participated in the research with no history or exposure to the telephone layout, due in part to limited availability of technology in the 1960s. In contrast, in today’s society in many developed countries, virtually all young adults have extensive exposure and practice using telephones and calculators. Thus, the participants in Marteniuk et al. (1996) may not have shown an advantage of either type of keypad arrangement because of their experience with both telephone and calculator layouts that the participants in the earlier studies did not have. 1.2. Current research The current study combines elements of both Rinck (1999, Experiment 1) and Marteniuk et al. (1996), but attempted to expand upon their studies in multiple ways. First, in contrast to Rinck (1999), we tested participants’ knowledge of telephone and calculator layouts using a within-subjects design. Additionally, a potentially important aspect of the stimuli that Marteniuk et al. used was that the numeric stimuli were presented as one continuous sequence of seven digits (e.g., 1947294), whereas the phone numbers were presented with a hyphen (e.g., 396e8142). Because of concerns that the phone stimuli were easier for participants in Marteniuk et al. to chunk in short-term memory, in the current research we added commas to the numeric digit strings to provide comparable chunks for participants to utilize. In addition, we tested all participants on their ability to enter all 10 digits on the keypad on each trial, whether as a numerical [1,234,567,890] or phone [(123) 456-7890] presentation. This decision provided two advantages over Marteniuk et al. First, by including all possible digits once in each stimulus, we controlled across trials for effects of key position upon performance (see below). Second, with the widespread use of cell phones, many dialed numbers today in the United States are ten digits, and thus the current stimuli may be more similar to the numbers dialed in daily telephone usage. Finally, in the current research we tested over twice as many participants and administered twice as many trials per condition than Marteniuk et al., because some trends that were not statistically significant in
Please cite this article in press as: Armand, J.T., et al., Task-specific performance effects with different numeric keypad layouts, Applied Ergonomics (2013), http://dx.doi.org/10.1016/j.apergo.2013.11.008
J.T. Armand et al. / Applied Ergonomics xxx (2013) 1e6
their study might have been due to insufficient power (N ¼ 24 participants in Marteniuk et al.). We predicted that, as in Rinck (1999), participants would be less accurate to explicitly recall the spatial layout of calculators compared to telephones. For the computerized task described below, when participants were asked across four different blocks to enter numeric or phone stimuli into either a calculator or telephone keypad, we predicted an overall advantage for entering phone stimuli, based on Marteniuk et al. (1996). However, in contrast to that research, we also predicted that we would observe an interaction of stimulus type and layout type, such that participants would be less accurate and/or slower in conditions of mismatching stimuli and layouts (e.g., numeric stimuli into a telephone keypad). Our interest in potential task-specific interactions of stimulus type and keypad layout motivated our final prediction. As stated previously, the middle row (4, 5, 6) on the keypad is the same for both telephone and calculator layouts. However, the top and bottom rows are mapped in an opposite fashion in the two layouts (Fig. 1). If, for example, users experience interference from the mental representation of the telephone keypad layout while using the calculator layout, then performance should be worse specifically for responses using the top and bottom rows of the keypad. Thus, we examined performance as a function of the row that was used for each of the 10 responses made on a given trial. 2. Method 2.1. Participants Participants for this study were students or community volunteers recruited from the campus of a small Midwestern university. Sixty-two participants completed the study; after removing participants who did not comply with instructions or had extremely low accuracy on the computerized task (see below), the final sample was comprised of 57 participants (40 female, 17 male), with a mean age of 26.9 (12.4) years old. We specify our rule for terminating data collection based on a recent recommendation by Simmons et al. (2011). Specifically, the first-author conducted the study as part of her capstone research project for graduation, supervised by the second- and third-author. The prescribed rule for all students completing capstone research projects during the academic year was to recruit and test a minimum of 60 participants, and data collection had to be completed by a certain date specified in advance. 2.2. Procedure All participants first completed a memory test similar to Rinck (1999). Participants were asked, in order, to provide the digit layout for a telephone, a calculator, and an ATM. Participants were given feedback after completing the memory test. Performance on the memory test was scored as the accuracy of participants’ recall. If participants reproduced the keypad layout in its entirety with no mistakes in the placement of the numbers, their answer was considered correct (see Fig. 1). After entering demographic information, participants then completed the main part of the experiment. Across four blocks, participants took part in a repeated-measures 2 (numeric or telephone number stimulus) 2 (calculator or phone keypad layout) experiment in which they participated in all four conditions: (a) entering phone stimuli [(123) 456-7890] using a telephone layout; (b) entering numeric stimuli (1,234,567,890) using a telephone layout; (c) entering phone stimuli using a calculator layout; and (d) entering numeric stimuli using a calculator layout. Participants were randomly assigned to one of four possible task orders, which
3
were counterbalanced using a Latin square design. The counterbalancing of the order of task conditions was done to prevent any practice effects that might be observed if all participants had used one keyboard format before the other, or entered one stimulus type before the other. The particular task orders were chosen to: (a) have each condition appear as the first, second, third, or fourth condition across participants; and (b) allow the experimenter to switch the keyboard only once for each participant, halfway through the computerized task to minimize disruption. Stimuli were 10-digit sequences that were either presented in the numerical [1,234,567,890] or phone [(123) 456-7890] format. All stimuli were randomly generated before the experiment with the constraints that: (a) every 0e9 digit appeared in each stimulus; (b) no digit appeared more than once within a stimulus; and (c) the digit 0 could not appear as the first number in the sequence. Note that the first two constraints produced stimuli that differed from Martenuik et al. (1996), in that their seven-digit stimuli did not contain all 10 possible responses and digits could repeat within a stimulus. Stimuli were presented in the center of the screen, and formatted so that the 10 digits appeared in the same 10 absolute locations onscreen regardless of the stimulus type. For each of the four stimulus-layout conditions, participants completed a total of 50 trials. The first 10 trials were practice, and were indicated as such to the participants. During the 10 practice trials, participants received performance feedback. These trials were not analyzed, but ensured that participants understood the instructions. Participants then completed 40 trials without feedback that were analyzed. Each trial started with a fixation screen until the participant pressed the spacebar with his or her nondominant hand. The 10-digit stimulus then appeared onscreen, and the participants were instructed to type the numbers into the keypad as quickly and accurately as possible. A cursor appeared under the current digit in the sequence, and advanced to the next digit after the response was keyed. Once the 10-digit number was entered, the next fixation screen appeared. Both accuracy and reaction time were measured. Accuracy for each stimulus-by-layout condition on the computerized task was derived by obtaining the mean of each participant’s mean accuracy, where all 10 digits must be entered in the correct order to be considered correct. Response time (RT) for each stimulus-by-layout condition on the computerized task was derived by obtaining the mean of each participant’s mean time to type the entire 10 digit string, from the moment the stimuli appeared onscreen until the moment the participant pressed the 10th key. As is the common practice in psychological research, we only included correct responses in the analyses of the RT data. 2.3. Materials For the initial memory test, participants used pencil and paper to record the answers. For the computerized task, participants completed a program created in E-Prime 2.0 (Psychology Software Tools, Pittsburgh, PA). Stimuli were presented visually in the center of the computer monitor screen, and participants used the number keypad of the computer keyboard to enter their responses. For the calculator layout, the standard number keypad layout was used. For the telephone layout, the buttons for the top and bottom rows of the keypad were exchanged. 2.4. Analyses Repeated-measures analyses of variance (ANOVAs) were used for each of the dependent variables in the study. An alpha of .05 was used for significance testing, except where noted in follow-up
Please cite this article in press as: Armand, J.T., et al., Task-specific performance effects with different numeric keypad layouts, Applied Ergonomics (2013), http://dx.doi.org/10.1016/j.apergo.2013.11.008
4
J.T. Armand et al. / Applied Ergonomics xxx (2013) 1e6
analyses to control for familywise error. Effect sizes are provided using partial eta-squared ðh2p Þ. 3. Results 3.1. Memory test Results from the explicit memory test (N ¼ 56 due to incomplete data from one participant) were analyzed with a repeatedmeasures ANOVA with the factor Layout (Telephone, Calculator, ATM). The main effect of Layout was significant, F(2, 110) ¼ 20.75, p < .001, h2p ¼ .274. Follow-up paired t-tests, using a Bonferronicorrected a (.05/3 ¼ .016), indicated that participants were more accurate in their recall of the telephone (95%) layout than either the ATM (68%) or calculator (41%) layout (all p’s < .013). 3.2. Computerized task Overall accuracy and RT results from the computerized task were analyzed with a repeated-measures ANOVA with the factors Stimulus (Phone, Numeric) and Layout (Telephone, Calculator). Looking first at accuracy (Fig. 2), the main effect of Layout was significant, F(1, 56) ¼ 5.98, p ¼ .018, h2p ¼ .097, with participants more accurate using the calculator layout (87%) versus the telephone layout (83%). The main effect of Stimulus was not significant, F(1, 56) ¼ .11, p ¼ .746, h2p ¼ .002, and neither was the interaction, F(1, 56) ¼ 1.26, p ¼ .267, h2p ¼ .022. The results for the mean RTs on correct trials are shown in Fig. 3. The significant main effects of Stimulus, F(1, 56) ¼ 39.30, p < .001, h2p ¼ .412, and Layout, F(1, 56) ¼ 39.96, p < .001, h2p ¼ .416, were qualified by a significant interaction, F(1, 56) ¼ 18.44, p < .001, h2p ¼ .248. As can be seen in Fig. 3, although overall participants were faster when using the calculator keypad and entering telephone numbers, the interaction was due to the pronounced slowing that participants showed when entering numeric stimuli using the telephone layout. When using the calculator keypad, participants were on average 191 ms slower to enter numeric versus phone stimuli. In contrast, when using the telephone keypad, participants were on average 537 ms slower to enter numeric versus phone stimuli. The results thus far have focused on performance on each stimulus; that is, examining accuracy and RT for the entire 10-digit number string presented. This represents the accuracy and RTs of the total of the 10 keystrokes per stimulus presented. In the
Fig. 2. Mean accuracy for the computerized task with error bars representing 1 95% confidence intervals for within-subjects designs (Loftus and Masson, 1994). Phone: Telephone stimulus number format; Number: Numerical stimulus number format; TEL: Telephone keypad design; CAL: Calculator keypad design.
Fig. 3. Mean correct response times for the computerized task with error bars representing 1 95% confidence intervals for within-subjects designs (Loftus and Masson, 1994). Phone: Telephone stimulus number format; Number: Numerical stimulus number format; TEL: Telephone keypad design; CAL: Calculator keypad design.
subsequent analyses, the accuracy and RTs for individual keypresses were examined. This aimed to measure the accuracy and speed of each individual keypress within each 10-digit number string. These were examined using a repeated-measures ANOVA, grouped according to its location in the top, middle, or bottom rows (Table 1).1 Looking first at accuracy, the effect of Row was significant, F(2, 112) ¼ 46.57, p < .01, h2p ¼ .454. For mean correct RTs, the effect of Row was significant, F(2, 112) ¼ 39.48, p < .01, h2p ¼ .414. Follow-up paired t-tests, using a Bonferroni-corrected a (.05/ 3 ¼ .016), indicated that responses using the middle row keys were both more accurate and faster than responses using the top or bottom row keys (all p’s < .008).
4. Discussion Our study was motivated by the observation that people seem to switch between tasks with relative ease on devices using numeric keypads, such as telephones and calculators, despite previously documented inaccuracy in their explicit memories for the digit layouts on the keypads. With the ever-increasing emergence of new devices, potentially with differing keypad layouts, it becomes an important issue to understand empirically how easily (or not) everyday people switch keypad layouts. Replicating Rinck (1999), we found that participants (mostly college students) were more accurate in their memory for telephone versus calculator layouts. In addition, in our study we wanted to address whether users type specific types of numeric information more accurately and more quickly using specific types of numeric keypad layouts. We found that participants were more accurate entering phone numbers no matter what the keypad layout. The phone number advantage is consistent with Marteniuk et al. (1996), but the result here cannot be solely attributed to differences in chunking, as the numeric stimuli were also presented in discrete chunks of information. Most importantly, participants’ time to type in the numerical information showed an interaction of stimulus format and keypad layout: participants were specifically slowed when entering numeric stimuli using a telephone keypad layout. Increased errors and slower RTs were observed specifically for responses using the top and bottom row keys of the keypad relative to the middle row keys,
1 The factors of Stimulus and Layout are not reported here, because the results are redundant with the analyses presented above (e.g., the RT for all 10 responses is the sum of the RT for each of the 10 responses).
Please cite this article in press as: Armand, J.T., et al., Task-specific performance effects with different numeric keypad layouts, Applied Ergonomics (2013), http://dx.doi.org/10.1016/j.apergo.2013.11.008
J.T. Armand et al. / Applied Ergonomics xxx (2013) 1e6 Table 1 Mean accuracy and correct response time for individual key responses as a function of keypad row. Variables
Accuracy (%)
Correct RT (ms)
Top row Middle row Bottom row
96.5 (2.4) 98.4 (1.4) 97.2 (2.1)
681.31 (206.07) 629.67 (175.38) 692.82 (208.17)
Note. Numbers in parentheses are standard deviations for each condition.
indicating interference from a competing mental representation of the other keypad layout. Our results suggest that, in contrast to Marteniuk et al. (1996), users show costs from entering specific types of numeric information using certain configurations of keypad layouts. In our study, the effect was observed in terms of RTs. Although the difference between entering phone and numeric stimuli on a telephone keypad was 537 ms, this difference was observed for entering a single sequence of 10 digits. If a user were to enter multiple 10-digit strings consecutively, as in the example given earlier for a pharmacy technician, that half-second difference could quickly add up and impact productivity. An example of a daily activity that epitomizes the specifically slowed combination of entering a numeric stimulus into a telephone keypad is that of a consumer using the calculator tool on a cell phone. The current results indicate that if a user is planning to conduct numerous calculations, he or she should instead keep a pocket calculator nearby to maximize efficiency, or use a smartphone application with a proper calculator layout. We found that participants were faster and more accurate when using the calculator layout, whether entering phone or numeric stimuli. The speed advantage for the calculator layout observed here may have been driven by the use of a computer keyboard keypad as the response device. Participants with previous computer experience likely have already established the motor memories that allow them to locate and type in number information on this device (Rinck and Ellwart, 2004). By switching the top and bottom rows of the keyboard keypad to create the telephone keypad layout, participants likely dealt with more motor memory interference in this condition than vice-versa. Future research using a touchtone or cell phone as the response device, and alternating the buttons as necessary to create a calculator layout, could confirm this idea. Our results also have implications for research studies that use existing keypad layouts for participants to enter numeric information. For example, Salvucci (2005) investigated participants’ ability to multitask using a desktop computer driving simulator. While participants drove the simulator, they also entered “four of their most familiar 10-digit phone numbers” (p. 479) using the computer keyboard numeric keypad, which was the standard calculator layout instead of a telephone layout (D. D. Salvucci, personal communication, 5/31/13). Of note, Salvucci (2005) eliminated 30% of their participants from the final sample, because those 30% were too inaccurate while entering the phone numbers, indicating these participants may have had difficulty dealing with the interference caused by using a calculator keypad design to enter telephone numbers. This example illustrates a potential source of unwanted variance in participants’ data, if a research study uses the keypad as the input device without taking into account differences between the positioning of keys in the telephone and calculator layouts. In addition, the observation of faster RTs for the calculator layout contradicts Conrad and Hull (1968), who strongly advocated for the use of the telephone layout for all numeric data entry. As mentioned previously, the participants in Conrad and Hull (1968)
5
had no experience with either touchtone telephones or calculators, which may account for the differences between their results and ours. An interesting factor for future research is how changing computing technologies affect human performance. As more consumers opt for laptops or tablets as their main computing devices, instead of traditional desktops, it’s likely that they will have less experience with the numeric keypads present on full keyboards. Thus, it may not be useful to advocate a specific numeric keypad layout as being the best or most efficient. Instead, users’ performance will be determined not only by the task and layout, but also by their history of using different keypad layouts. Although our study contributes to the understanding of how keypad layout can affect accuracy and speed of numeric data entry, there are a few limitations to this study. First, it would be ideal to have a larger sample of participants. The sample size was adequate for significance testing, and larger than previous research on this topic (e.g., Marteniuk et al., 1996), but for the sake of increased external validity, a broader age-group and/or sample with expertise using specific layouts would be interesting for future research. In particular, future work could test competing numeric keypad arrangements with users who have extensive practice using specific numeric keypad arrangements (e.g., accountants) to see if the same effects are obtained as in the non-expert subjects tested here. Second, a more seamless transition between keypad layouts would be valuable for testing quick transitions between keypads. This would help to examine the phenomena of task-switching versus motor memory of keypads. Because our RT measure included aspects of both cognitive decision time and motor movement time, future work examining the effect of novel versus already known phone numbers and digit sequences could be informative. Additionally, the QWERTY keyboard layout of numbers should be tested (with the number keys 1-0 all adjacent in one row), because many laptops and keyboard-free devices (e.g., touchscreen keyboards, tablets) have this layout as the only option for number entry.
5. Conclusion In conclusion, we found evidence that despite users’ relatively inaccurate memories for the layout of a calculator, entering numeric information into a calculator keypad layout was superior to using a telephone keypad layout. In addition, although participants were faster to enter phone numbers compared to other numeric stimuli on either type of layout, they were particularly slowed when entering numeric stimuli into a telephone keypad. Responses made using the middle row of keys were faster and more accurate than responses using the top and bottom row of keys, providing evidence that interference from competing mental representations affects user performance during numeric data entry. Future research should address a broader age group, and include participants with keypad expertise. As more electronic devices come onto the market and are used in number entry at various jobs where accuracy and time are critical, it becomes increasingly important to understand how these different layouts impact both accuracy and speed of number entry.
Acknowledgments Correspondence concerning this article should be sent to Thomas S. Redick, Purdue University, 703 Third Street, West Lafayette, IN 47907 (email to [email protected]). While writing this article, TSR was supported by the Office of Naval Research (Award # N00014-12-1-1011).
Please cite this article in press as: Armand, J.T., et al., Task-specific performance effects with different numeric keypad layouts, Applied Ergonomics (2013), http://dx.doi.org/10.1016/j.apergo.2013.11.008
6
J.T. Armand et al. / Applied Ergonomics xxx (2013) 1e6
References Conrad, R.R., Hull, A.J., 1968. The preferred layout for numeral data-entry keysets. Ergonomics 11, 165e173. Deininger, R.L., 1960. Human factors engineering studies of the design and use of push button telephone tests. Bell Syst. Tech. J. 39, 995e1012. Jones, G.V., Martin, M., 2009. Spatial recall improved by retrieval enactment. Psychon. Bull. Rev. 16, 524e528. Loftus, G.R., Masson, M.E.J., 1994. Using confidence intervals in within-subject designs. Psychon. Bull. Rev. 1, 476e490. Lutz, M., Chapanis, A., 1955. Expected locations of digits and letters on ten-button keysets. J. Appl. Psychol. 39, 314e317.
Marteniuk, R.G., Ivens, C.J., Brown, B.E., 1996. Are there specific performance effects for differently configured numeric keypads? Appl. Ergon. 27, 321e325. Rinck, M., 1999. Memory for everyday objects: where are the digits on numerical keypads? Appl. Cogn. Psychol. 13, 329e350. Rinck, M., Ellwart, T., 2004. Repetition priming during the use and recall of numerical keypads. Eur. J. Cogn. Psychol. 16, 841e861. Salvucci, D.D., 2005. A multitasking general executive for compound continuous tasks. Cogn. Sci. 29, 457e492. Simmons, J.P., Nelson, L.D., Simonsohn, U., 2011. False-positive psychology: undisclosed flexibility in data collection and analysis allows presenting anything as significant. Psychol. Sci. 22, 1359e1366. Straub, H.R., Granaas, M.M., 1993. Task-specific preference for numeric keypads. Appl. Ergon. 24, 289e290.
Please cite this article in press as: Armand, J.T., et al., Task-specific performance effects with different numeric keypad layouts, Applied Ergonomics (2013), http://dx.doi.org/10.1016/j.apergo.2013.11.008
