Skill Acquisition With Text-Entry Interfaces

Abstract

Operating information technology challenges older users if it requires executive control, which generally declines with age. Especially for novel and occasional tasks, cognitive demands can be high. We demonstrate how interface design can reduce cognitive demands by studying skill acquisition with the destination entry interfaces of two customary route guidance systems. Young, middle-aged, and older adults performed manual destination entry either with a system operated with multiple buttons in a dialogue encompassing spelling and list selection, or with a system operated by a single rotary encoder, in which an intelligent speller constrained destination entry to a single line of action. Each participant performed 100 training trials. A retention test after at least 10 weeks encompassed 20 trials. The same task was performed faster, more accurately, and produced much less age-related performance differences especially at the beginning of training if interface design reduced demand for executive control, perceptual processing, and motor control.

Keywords

applied cognition cognitive ergonomics skill acquisition text entry interface design

Information technology can enrich and ease the life of older adults as it does for younger people (Cohen-Mansfield et al., 2005; Czaja & Lee, 2007; McMellon & Schiffman, 2002; Rogers & Fisk, 2010; Tun & Lachman, 2010; White et al., 1999). Some opportunities to use information technology can be voluntarily chosen such as mobile phones, private computer hardware and software, websites, entertainment technology, and route guidance systems in private vehicles. Others are hard to avoid, for example, information technology at work, route guidance systems in commercial vehicles, ATMs, ticket machines, and necessary health care technology such as blood-pressure monitors and blood-glucose meters. Older users are often less experienced with information technology in general (Rogers, Stronge, & Fisk, 2006) and on average less well equipped to learn using novel technologies because of declining sensory acuity, motor control, and cognitive abilities (Czaja & Lee, 2007). Thus, designing interfaces that are accessible to older users without extended training is a challenge.

Existing guidelines provide general advice for the design of interfaces, training materials, and efficient training procedures (Fisk, Rogers, Charness, Czaja, & Sharit, 2009; Nichols, Rogers, & Fisk, 2006). In addition, they stress the importance of usability testing with samples of older users because commonly unforeseen usability problems show up in testing. Among the less obvious, usability problems are cognitive demands. They are more elusive and less apparent than, for example, bad contrast or visual clutter. Moreover, cognitive demands are rather task specific and addressed in rather general terms in design guidelines. The advice is, for example, to minimize working memory demands and to provide environmental support for choosing the next action.

In this article, we aim to highlight the potential of reducing cognitive demands for improving the accessibility of information technology. We study two differing design solutions for the same task, which clarify a critical element in reducing cognitive demands: to relieve the user from as many choices and decisions as possible. As will be demonstrated, this can be achieved by dialogue design and the design of controls. In addition to reduced executive control demands, reduced perceptual processing and motor control demands yield important benefits as well.

Figuring out what to do next is necessary in every novel task and it is a form of problem solving, which requires controlled cognitive processing as opposed to merely performing learned action sequences in response to stimuli (Bainbridge & Quintanilla, 1989). Older adults are less well equipped for this kind of problem solving that involves the deployment of executive functions, which often decline with age (Craik & Bialystok, 2006; Fisk, Fisher, & Rogers, 1992; Mayr, Kliegl, & Krampe, 1996). In addition to the declining capacity for controlled processing, important factors in cognitive aging are the general slowing of processing (Cerella, 1990; Salthouse, 1996), reduced sensory acuity when interpreting visual displays and acoustic messages (Czaja & Lee, 2007; Jastrzembski & Charness, 2007), and reduced precision in motor control (Vercruyssen, 1997).

The average speed of learning differs between younger and older users as well (Baron & Cerella, 1993; Czaja, 1996; Jastrzembski & Charness, 2007). Hence, guidelines recommend extended training procedures for older users (Fisk et al., 2009; Nichols et al., 2006). For many IT tasks that occur in the home environment and in the public, however, older users are unlikely to receive formal training. Thus, suboptimal design cannot be made up for by training. Instead, interfaces should be easy to use with minimal prior experience. The issue of learnability is even more important for interfaces that are seldom used because only sporadic practice occurs and at every instance of use there is a lack of training. Long intervals without practice are disadvantageous for skill retention. There is some evidence that the IT performance of older adults is more strongly affected by retention intervals than performance of younger adults (e.g., Mykityshyn, Fisk, & Rogers, 2002). This is another reason why ease of use is of eminent importance for interfaces intended to be accessible to the older population.

The task with which we aim to clarify the connection between cognitive demands and ease of use for older adults is text entry. Not free text entry as it is performed in word processing, but text entry in search of a match in a database, for example, when searching in library systems, mail-order catalogues, address books, public transportation schedules, and route guidance systems. When using a route guidance system to find the way to a novel destination, the first step is to specify the destination. Destination entry finishes when the correct match in the database of destinations is established.

We compared two design solutions for destination entry. In one system (System A), characters are selected with buttons on a remote control, the display shows the constructed string and a dynamically adjusted view on database entries. In the other system (System B), characters are selected with a rotary encoder (dial and push button), the constructed string is automatically completed as far as it is disambiguated, and characters without matches in the database are blocked (intelligent speller). These integrated design solutions differ in demands for executive control, perceptual processing, and motor control.

Executive control demands are higher for System A because to complete destination entry efficiently (avoiding entering every character), the user has to switch from letter selection to selection in the database. Switching is possible at any time, thus constant monitoring of the changing view on the database and deciding whether to switch or to continue spelling is required (see the task analysis in the Method section). In contrast, destination entry in System B proceeds in a single line of action without monitoring and decision demands. With System A, repeated switching between spelling and evaluating the database view until the decision to stop spelling is made creates executive control demands similar to those studied in the task switching paradigm. They are quantified as switch costs. Global switch costs, which are thought to capture the cost of maintaining and scheduling two task sets, indicate a marked age deficit (Wasylyshyn, Verhaeghen, & Sliwinski, 2011). Thus, we expected a stronger age effect with System A.

Perceptual processing demands are higher with System A because monitoring the database view and selecting from the database requires visual search in a list. This perceptual demand is absent with System B. Consequently, the age effect with System A compared with System B should be strengthened because of the age-related decline in perceptual processing speed (Jastrzembski & Charness, 2007).

Finally, motor control demands are higher with the multibutton remote control of System A than with the rotary encoder of System B. Hitting buttons is a motor control task with more degrees of freedom than turning a knob. Thus, a rotary encoder alleviates the age-related decline in motor control (McLaughlin, Rogers, & Fisk, 2009; Rogers, Fisk, McLaughlin, & Pak, 2005) and should increase the age effect further. Moreover, the rotary encoder simplifies the choice reaction task in response selection. Thus, we expected less errors and a smaller effect of training with System B in all age groups.

These system differences take effect in combination. Clearly, System B reduces cognitive demands in destination entry compared with System A. Hence, particularly older users should perform better with System B from the beginning of training. Performance and ease of use were quantified by task times, error rates, skill acquisition, and skill retention.

Acquiring skill in text entry and information search involves learning how to operate controls and which information to attend to. Users do not learn mere stimulus response mappings, but procedures of how to achieve a goal in a sequence of evaluating and responding to displayed information. At the beginning of training, declarative knowledge acquired through instructions or observations has to be held in working memory to figure out how to respond at a certain point in the interaction sequence. At this declarative stage (Anderson, 1982), the controlled and conscious processing of feedback and planning of actions resembles problem solving (Bainbridge & Quintanilla, 1989). Later in learning, when efficient procedures for standard operation have been established, this controlled processing becomes necessary again if unexpected feedback occurs as a result of nonstandard system behavior, user error, or system failure. Learning to use information technology involves not only the speedup of standard procedures including skilled operation of controls and selective attention to critical information, but also reduced error frequency and the acquisition of procedures for correcting errors.

The speedup in task performance resulting from a fixed amount of practice is greatest at the beginning of training and diminishes with increased training. This holds true across a variety of simple and complex tasks (Newell & Rosenbloom, 1981). The characteristic diminishing effect of practice on task time can be described quantitatively with a power function (e.g., Ritter & Schooler, 2001). A simple version of the power law of practice relates the task time (T) to the number of practice trials (N) as:

T = B N^{- c}

where c specifies the rate with which practice decreases the task time from B in the beginning to zero in the limit. The asymptote may actually be higher than zero, but is difficult to estimate unless training is studied for an extended period.

Skill acquisition data from simple tasks may have a better fit with an exponential function (Ghisletta, Kennedy, Rodrigue, Lindenberger, & Raz, 2010; Heathcote, Brown, & Mewhort, 2000). For complex tasks, however, power functions capture training effects successfully (Lee & Anderson, 2001). Even if individual task times decrease discontinuously as a result of strategy shifts, task times averaged across individuals follow the power law (Haider & Frensch, 2002).

Our main objective in the present study is to demonstrate with a realistic text entry task and customary information technology that particularly older users benefit from interface design that minimizes the demand for choices and decisions requiring controlled processing both at the level of monitoring and motor actions. This prediction follows directly from the age-related decline of executive functions and motor control. The dialogue design lower in demand for controlled processing was also lower in demand for perceptual processing. Thus, system differences cannot be attributed to a single factor, but demonstrate the potential of interface design with high ecological validity. We collected training data sets for skill acquisition over 100 destination entry trials for two systems (System A and System B) and three age groups (young, middle, old). The reduction of task times with training was quantified by fitting power functions. In addition to task times, we analyzed error frequencies. A couple of weeks after the training, participants were invited to return and perform another 20 destination entry trials as a test of skill retention.

Method

Young, middle-aged, and older participants performed manual destination entry on two customary route guidance systems. At each age level, one group used a system that was operated with a remote control and in which destinations were entered in a two-stage procedure consisting of spelling and list selection (System A). The second group used a system that was operated with a rotary encoder next to the display and in which destination entry was supported by an intelligent speller (System B).

Participants

The younger age group consisted of 16 participants (7 male, 9 female) who were students between 20 and 27 years of age (M = 22.4, SD = 2.0). The middle-age group consisted of 60 adults (37 male, 23 female) between 31 and 56 years of age (M = 45.4, SD = 6.4). The 16 older participants (15 male, 1 female) were between 66 and 74 years of age (M = 68.8, SD = 2.6) and were recruited from the audience of a series of public lectures at Chemnitz University of Technology. All participants held valid driving licenses and received monetary compensation. They were informed about the purpose and the procedure of the study and were told that they were free to quit at any time.

At each age level, participants were divided in two groups with approximately equal mean age; one group was trained with System A, the other with System B. The mean age in the system groups is shown in Table 1. Skill retention was studied after a delay of at least 54 days. Eleven of the younger participants, 55 of the middle-aged participants and all 16 of the older participants returned for the skill-retention test (see Table 1). The task time data of the middle-age group without the data on skill retention and error rates have been reported in a previous article (Jahn, Krems, & Gelau, 2009).

Table 1.

Age of Participants and Delay of the Retention Test in the Age and System Groups.

	Initial training (100 trials)			Retention test (20 trials)
	N	Age M (SD)	Range	N	Days¹ M (SD)	Range
Younger adults
System A	8	22.8 (1.7)	20-26	5	150.4 (34.0)	84-180
System B	8	21.9 (2.2)	20-27	6	102.5 (22.8)	83-174
Middle-aged adults
System A	29	46.3 (6.2)	31-55	25	103.0 (21.0)	71-155
System B	31	44.6 (6.4)	35-56	30	100.3 (15.6)	79-136
Older adults
System A	7	68.5 (2.8)	66-74	7	85.6 (11.4)	68-104
System B	9	69.1 (2.3)	66-73	9	73.3 (12.3)	54-96

Interval between initial training and retention test in days.

Human-Machine Interfaces

We used two customary route guidance systems that both were installed in the same vehicle (a BMW 750iL) and differed with regard to HMI design. Both systems had similar sized color displays, but were operated differently. System A (Blaupunkt TravelPilot DX-N) is typical of add-on systems and was operated by keys on a remote control. System B (BMW Carin 520) was factory installed in the console and was mainly operated by a rotary encoder. We mounted System A’s display flexibly, such that it could be placed in front of System B’s display to experimentally control display position. The location of the systems’ displays was approximately 50 cm to the right and 30 cm down from the normal forward line of sight. They each subtended approximately 5° of visual angle horizontally.

With System A, destination entry required alphanumeric spelling of city names and street names by selecting letters on the display. The design of the main destination entry screen of System A is schematically shown in Figure 1. Two rows of letters are shown on the top. The user moves a cursor to the intended letter by means of four arrow keys on the remote control (up, down, left, and right) and then selects the highlighted letter with the enter key. The enter key was surrounded by the arrow keys on the remote control. The constructed string was displayed in a middle area of the screen. Below were four dynamically adjusted lines, which showed consecutive database entries starting with the letters entered so far. To proceed to the selection of an entry, this window on the database was enlarged by holding the enter key for 2s. The enlarged window contained seven lines covering the whole display. A cursor highlighting one line was moved with the up and down arrow keys. The intended entry had to be selected with the enter key. The switch from spelling to selecting among database entries could be initiated at any time. However, when too few letters were entered before a switch to the database, extended scrolling was necessary to reach the intended entry. Thus, we instructed participants to switch as soon as they noticed the intended entry among the four lines displayed on the spelling screen. Spelling and database selection had to be completed twice for each destination entry, once for the city name and then for the street name. This was true for System B as well. Thus, the top-level methods in NGOMSL task analyses of the entry procedures (Kieras, 1996) are the same for System A and System B:

Method for goal: enter destination.

Step 1. Accomplish goal: enter string <CITY>.

Step 2. Accomplish goal: enter string <STREET>.

Step 3. Accomplish goal: start route guidance.

Step 4. Return with goal accomplished.

The NGOMSL model for System A’s entry procedure contains a selection rule set for deciding when to switch from spelling to database selection that users encounter inevitably and repeatedly:

Method for goal: enter string <STRING>.

Step 1. Retain that string is <STRING>.

Step 2. Accomplish goal: proceed with entering.

Step 3. Decide: If spelling screen is displayed, Then go to 2.

Step 4. Press ok button and forget that string is <STRING>.

Step 5: Return with goal accomplished.

Selection rule set for goal: proceed with entering.

If <ENTERED CHARS> is short,

Then accomplish goal: enter character <NEXT CHAR>.

If an entry in <DATABASE WINDOW> matches <STRING>,

Then accomplish goal: select in database.

If entries in <DATABASE WINDOW> are similar to <STRING>,

Then accomplish goal: select in database.

If <ENTERED CHARS> is not short and entries in <DATABASE WINDOW> are not similar to <STRING>,

Then accomplish goal: enter character <NEXT CHAR>.

Return with goal accomplished.

Figure 1.

Schematic illustration of the route guidance systems’ interface displays for alphanumeric destination entry.

With System B, letters were selected by moving a cursor on a single row with a dial button. The dial button was located next to the bottom left corner of the display and provided haptic feedback on each step that the cursor was moved. Pressing the button selected a letter. As shown schematically in Figure 1, the other information displayed on the destination entry screen of System B was a line with the string entered so far. This string could consist of more letters than were selected because the intelligent speller completed strings as far as they could unambiguously be continued based on the database. Furthermore, only those letters could be selected that would lead to strings that matched entries in the database. The remaining letters were grayed out. On the same screen, when the intended entry was fully displayed, it had to be confirmed by a button press. As with System A, spelling had to be completed twice, once for the city name and then for the street name. The NGOMSL model of System B’s entry procedure contains no selection rule set:

Method for goal: enter string <STRING>.

Step 1. Retain that string is <STRING>.

Step 2. Accomplish goal: enter characters.

Step 3. Verify that <STRING> is displayed.

Step 4. Press dial button and forget that string is <STRING>.

Step 5. Return with goal accomplished.

Method for goal: enter characters.

Step 1. Accomplish goal: enter character <FIRST CHAR>.

Step 2. Read <ENTERED CHARS> from screen.

Step 3. Decide: If <ENTERED CHARS> match <STRING>, Then go to 5.

Step 4. Accomplish goal: enter character <NEXT CHAR> and go to 2.

Step 5. Return with goal accomplished.

In addition to differing controls and procedures for destination entry, a further difference between the two interfaces concerns the methods for error correction. With System A, errors could be corrected by means of an undo key on the remote control. With System B, the cursor had to be moved to a delete field, which had to be selected.

Destination Entry Tasks

Destinations consisted of a city name and a street name, for example Berlin, Scheinerweg. In a series of tests, we selected 100 destinations that worked with both systems and grouped them into 25 sets of four destinations each. The destinations were selected such that the sets of four tasks required an approximately equal and constant number of entry steps for both systems. Cursor movements were not counted as steps. On average, a single destination entry required 8 entry steps with System A and 10 entry steps with System B. Four additional destinations were selected for the instructional trials.

The sets of four destinations were the basis for constructing trial sequences for individual participants. The first set was the same for all participants. The order of the following 24 sets was counterbalanced. This constituted the 100 training trials in the first session. The retention test consisted of 5 sets (20 trials). The first set of retention trials was the same for all participants, the following four sets were the same four sets that a participant had performed at the end of the first session. Each destination was printed on a white card, city name and street name on separate lines.

Procedure

The participant sat in the driver’s seat of the vehicle; the experimenter sat next to the participant in the front passenger seat. At the beginning of the initial training session, the experimenter explained the display and the controls, and demonstrated destination entry twice. During the second destination entry, the experimenter demonstrated and explained the procedure for error correction. Then, the participant was instructed to enter two destinations and to try error correction with the second destination entry. The experimenter put the cards with the destinations to be entered in front of the gearshift as it was done later through the whole experiment.

Each participant completed 100 training trials during the initial training. The first block of trials consisted of 20 trials, the remaining blocks consisted of 16 trials. Before each block, the participant read aloud the destinations, and at the end of each trial, the participant read aloud the displayed distance to the destination. After each block, participants could take a rest. Complete sessions took several hours per participant. When participants returned for the skill-retention test, they received no further instruction. The retention test consisted of 20 trials.

Results

The total task time was defined as the interval from the first to the last button press of a destination entry trial. Total task times and error frequencies were manually coded from video recordings. Six trials were discarded as outliers (more than 4 SDs above the mean for the respective combination of system and age group). Mean total task times were computed for sets of four trials requiring approximately equal numbers of entry steps. We fitted power functions to those means of total task times to quantify the process of skill acquisition. The fitted power functions were of the two-parameter form T = BN^- ^c , for which the asymptote is assumed to be zero; the variability of the data was too high to estimate the asymptote as a third parameter. The best fitting power functions are shown in Table 2 with root-mean-square-error (RMSE) as indicator of goodness of fit (e.g., Loftus, 2002). Error frequencies were coded for each trial and included the error categories spelling error, wrong key, wrong menu selected, and wrong list item selected.

Table 2.

Power Functions Fitted to Total Task Times.

	T = BN^– ^c	RMSE	N
Younger adults
System A	T = 53.49 N^–.071	1.37	8
System B	T = 53.79 N^–.036	1.77	8
Middle-aged adults
System A	T = 90.19 N^–.110	2.12	29
System B	T = 63.13 N^–.057	1.36	31
Older adults
System A	T = 143.42 N^–.134	4.77	7
System B	T = 91.93 N^–.081	2.67	9

RMSE = root-mean-square-error.

Total Task Times

Mean total task times computed for sets of four trials and fitted power functions are shown in Figure 2 for System A and System B separately for the younger, middle, and older age groups. Older adults had a strong advantage with System B throughout training. The middle-age group had an advantage with System B, too, and took less time than older adults with both systems. Younger adults had the lowest task times and showed a reversed system effect. Unexpectedly, younger adults were faster with System A than with System B after few training trials because they hit performance limits of System B. System B produced longer delays between entries because the speller took time to compute displays in response to entered letters; whereas with System A, younger adults could reduce task time further with practice. Therefore, the learning parameter c in Table 2 that reflects the rate of speedup is technically limited for younger adults using System B. For the older and the middle-age groups, however, the higher learning parameter c with System A validly reflects that there was more to learn in using System A. The parameter B indicating the task time at the beginning of training differed between systems by 51, 27, and 0 for the older, the middle, and the younger age group, respectively, which demonstrates that particularly the older participants were much faster with System B at the beginning of training.

Figure 2.

Mean total task times for sets of four destination entry trials with fitted power functions by system and separately for the three age groups: younger, middle-aged, and older adults.

To test system differences and age effects after some practice, we computed mean task times omitting the 20 trials in the first training block. These means for the trials 21 to 100 are shown in the middle column of Table 3. Because the middle-age groups were larger and had a wider age range than the young and old groups, we did not compare the age groups in a factorial design. Instead, we computed a sequential multiple regression entering age as a continuous variable, followed by system and the age by system interaction (top half of Table 4). Both age and system contributed significantly to the finally explained variance of R² = .62. The reversed system effect in the younger age group resulted in a significant system by age interaction.

Table 3.

Mean Total Task Times at the Beginning of Training (Trials 1-20), After Some Practice (Trials 21-100), and at the Retention Test (Trials 101-120).

	Trials 1-20	Trials 21-100		Trials 101-121 (retention test)
Variables	M (SE)	M (SE)	N	M (SE)	N
Younger adults
System A	46.07 (1.46)	40.25 (1.17)	8	46.65 (1.65)	5
System B	50.08 (1.08)	46.54 (1.49)	8	47.25 (1.97)	6
Middle-aged adults
System A	71.21 (2.60)	58.34 (2.14)	29	63.32 (2.45)	25
System B	56.44 (1.26)	50.04 (0.92)	31	53.92 (1.47)	30
Older adults
System A	107.36 (6.25)	83.82 (4.19)	7	98.46 (5.06)	7
System B	78.08 (4.88)	66.42 (2.66)	9	72.18 (4.43)	9

Table 4.

Sequential Multiple Regression Analyses of Mean Task Times and Mean Error Rates After Some Practice (Trials 21 to 100).

Regressor	β	t	p	R² increment	df	F change	p
Task time
Age	.70	9.20	<.001	.49	1, 90	84.72	<.001
Age	.70	9.73	<.001
System	.25	3.42	.001	.06	1, 89	11.69	.001
Age	.71	10.77	<.001
System	.25	3.73	<.001
Age × System	.28	4.20	<.001	.08	1, 88	17.62	<.001
Error Rate
Age	.35	3.53	.001	.12	1, 90	12.49	.001
Age	.35	4.28	<.001
System	.54	6.63	<.001	.29	1, 89	43.89	<.001
Age	.35	4.40	<.001
System	.54	6.70	<.001
Age × System	.14	1.73	.087	.02	1, 88	2.99	.087

The size of the system effect (Cohen’s d) was 1.85 in the older and 0.94 in the middle-age group; the size of the reverse system effect in the younger group was 1.66. The larger system effect in the older than in the middle group was not just the result of age-related general slowing that inflated the system difference in the older group. With System A, older adults were slower than the middle group by a factor of 1.44, whereas with System B the factor was just 1.33.

The retention test took place after at least 10 weeks, but the mean retention interval differed between age groups (Table 1) and in the younger group, a large proportion (37%) of participants in the System A group is missing. Thus, we compared skill retention statistically only within age groups and only for the middle and the older age groups. Nonetheless, the data in Figure 2 indicate that in the younger age group performance in the retention test was similar to the performance at the beginning of training. A further caveat in the comparison of skill retention is apparent in the task time figures: the first four skill-retention trials, which were the same for all participants, might have favored System B.

Lasting effects of training would shorten task times at the retention test compared with task times at the beginning of training. Thus, we computed means of total task times for training trials 1 to 20 (Block 1) and for the trials 101 to 120 (retention test) and compared them separately for the older and middle-age groups in ANOVAs including the within-subjects variable retained training (Block 1 vs. retention test) and the between-subjects variable system (see Table 3).

In the middle-age group, the system by training interaction was significant, F(1, 52) = 6.10, MSE = 17.771, p = .017. The difference between task times in Block 1 and in the retention test was larger for System A than for System B (ds 0.47 and 0.31). The system effect was larger in Block 1 than in the retention test (ds 1.25 and 0.93). The main effects of system, F(1, 52) = 17.42, MSE = 200.676, p < .001, and of retained training, F(1, 52) = 28.19, p < .001, were significant as well.

In the older age group, the main effects of system, F(1, 14) = 16.27, MSE = 373.881, p = .001, and of retained training F(1, 14) = 9.87, MSE = 43.513, p = .007, were significant. There was no significant interaction (F < 1). Nonetheless, the retained training effect was slightly larger for System A than for System B (ds 0.54 and 0.42). The system effect did not decrease from Block 1 to the retention test (ds 1.89 and 1.97).

Error Frequencies

Mean numbers of errors per trial computed for blocks of 20 trials are shown for the younger, middle and older age groups in Figure 3 separately for System A and System B. In all age groups and at all levels of training, error frequencies were higher with System A than with System B. It is apparent that the speedup in task times observed during the initial training for middle-aged and older users of System A was partly the consequence of a reduction in errors with practice. Mean numbers of errors after some training were computed for trials 21 to 100 and are presented in the middle column of Table 5. They show the system effect in all age groups and an effect of age. Errors were more frequent for the older than for the younger and middle-age groups with both systems (see also Figure 3). As for task times, we entered age as a continuous variable in a sequential multiple regression analysis, followed by system and the age by system interaction (bottom half of Table 4). Both age and system contributed significantly to the finally explained variance of R² = .43. The interaction was not statistically significant. The size of the system effect was 1.56, 1.27, and 2.53 in the younger, middle, and older age groups, respectively.

Figure 3.

Mean numbers of errors per trial for blocks of 20 destination entry trials by age group and separately for System A and System B. Trials 101 to 120 constitute the retention test (error bars denote the standard error of the mean).

As with task times, we tested for the effects of retained training comparing Block 1 and the retention test for the middle-aged and the older groups. Mean numbers of errors in trials 1 to 20 (Block 1) and trials 101 to 120 (retention test) are shown in Table 5. For System A, there was no main effect of retained training and no interaction (both Fs < 1). Only the system effect was significant, F(1, 52) = 34.60, MSE = 0.054, p < .001, and of similar size in Block 1 and in the retention test (ds 1.38 and 1.29). The retained training effect sizes were close to zero for System A and System B (ds −0.11 and 0.09).

In the older group, the system effect was larger in Block 1 than in the retention test (ds 2.12 and 1.17). Nonetheless, the system by retained training interaction was not significant, F(1, 14) = 1.20, MSE = 0.038, p = .29. Only the system effect was significant, F(1, 14) = 16.15, MSE = 0.063, p = .001. The main effect of retained training was not significant (F < 1). The retained training effect differed in sign between systems and was larger for System A than for System B (ds 0.47 and −0.20). The small negative training effect for System B may be the result of missing instruction trials before the retention test.

Table 5.

Mean Number of Errors Per Trial at the Beginning of Training (Trials 1-20), After Some Practice (Trials 21-100), and at the Retention Test (Trials 101-120).

	Trials 1-20	Trials 21-100		Trials 101-121 (retention test)
Variables	M (SE)	M (SE)	N	M (SE)	N
Younger adults
System A	0.18 (0.04)	0.20 (0.03)	8	0.36 (0.11)	5
System B	0.09 (0.03)	0.09 (0.02)	8	0.09 (0.03)	6
Middle-aged adults
System A	0.46 (0.05)	0.27 (0.03)	29	0.43 (0.06)	25
System B	0.16 (0.02)	0.10 (0.01)	31	0.15 (0.03)	30
Older adults
System A	0.76 (0.08)	0.47 (0.05)	7	0.65 (0.09)	7
System B	0.32 (0.06)	0.20 (0.03)	9	0.37 (0.08)	9

Discussion

The systems differed in the time required for destination entry, in the rate of errors, in task time variability, and in the course of skill acquisition. Middle-aged and older adults had a clear advantage when they used System B with which destination entry was cognitively less demanding than with System A (see also Jahn et al., 2009). Older adults took more time than middle-aged and young adults to enter destinations with both systems, however, task times were increased disproportionately for System A. Practice effects were also larger for System A and largest in the older sample reflecting that there was more to learn in using System A.

Both disproportionately longer task times and larger practice effects for System A confirmed predictions inferred from its higher demand for controlled processing and the decline of executive functions in cognitive aging, from its higher perceptual processing, and higher motor control demands. Destination entry with System A required more controlled processing for operating controls, for monitoring displayed information, and for deciding between options in the interaction sequence. It required list search and motor control for operating multiple buttons. These cognitive demands slowed operation in particular for older users. Reduced demands for controlled processing after some practice consequently produced larger practice effects in the older sample. In contrast, younger adults were not slowed by the higher demands of System A, although the system effect was still present in error rates. Because of age-related higher cognitive capacities and in part because of experience with similar tasks and controls, they performed almost equally fast with the two systems. They even hit limits caused by idle delays with System B, such that the system difference in task times was reversed.

How much task times with System B were prolonged by system delays even for middle-aged and older users can be estimated from results of a previous study (Jahn et al., 2009), in which both systems were used while driving. Drivers returned to the driving task and averted their gaze from the display during delays, thus, total display glance time is a rough estimate of task time without delays. On average, delays prolonged System B task times by about 12s. Delays with System A were negligible. Hence, the system difference could be even larger with improved system speed. The driving study provided also an opportunity to compare dual task ratings of mental demands to single task ratings collected when destinations were entered at standstill. The size of the system effect in mental demand ratings was 2.0 while driving and 1.3 at standstill. These subjective ratings confirm that the combined cognitive demands were higher with System A, which consequently was harder to operate while driving.

Error frequencies contributed to the higher task times of older adults and to the system effect because task times include the time for error correction. As for task times, we observed strong effects of age and system on error frequencies. The strong system effect for error frequencies obtained for older users corroborates that the demands for controlled processing and motor control of System A challenged particularly older users.

Both middle-aged and older users showed retained training effects for task times at the retention test, however, the system effect was only reduced for the middle-aged group. For the older age group, the system effect was as large as at the beginning of training demonstrating that interfaces with reduced cognitive demands are particularly important for occasionally used information technology. This point is strengthened by the observed error frequencies, which showed hardly any effect of the previous 100 training trials at the retention test in the middle-aged and the older groups (compare Schmidt & Bjork, 1992).

Task times differed much more between age groups for System A than for System B, although exactly the same IT tasks were performed. The system difference resulted from choice reactions and motor control required for manual operation, from differing need for directed inspection of displayed information, and differing demands for choosing among actions. The exact contribution of the respective design features could not be determined because we employed customary route guidance systems. We strove for ecological validity and carefully selected destinations that required approximately equal entry steps with both systems, but we did not vary features independently. Nonetheless, the differing dialogues and controls illustrate how the same IT task can be designed more accessible and less error prone by striving for simple controls and a single line of action.

In part, age effects may have resulted from differing amounts of prior experience in the age groups. Prior skill in operating information technology and using controls similar to the system controls was more advanced in the younger sample according to self-reports. However, with a more representative sample of older users than ours, who were recruited in a public university lecture, the age effect and the system difference might have been even larger.

Our results are in line with theories of cognitive aging that postulate a decline of specific cognitive processes (Craik & Bialystok, 2006; Fisk et al., 1992; Mayr et al., 1996) besides general factors such as general slowing of processing speed (Cerella, 1990; Salthouse, 1996). Specific factor theories are supported by particularly large age effects for complex tasks that demand controlled processing and load working memory (e.g., Mayr et al., 1996; Verhaeghen, Kliegl, & Mayr, 1997). Further evidence for a specific decline of controlled processing in normal aging are disproportionate changes in brain structure and neurotransmitter systems (Raz, 2000). Correlations of age and brain volume are highest for prefrontal areas that support controlled processing and working memory. Furthermore, age-related deterioration of the dopaminergic system affects predominantly prefrontal areas.

Age-related cognitive changes make older users highly susceptible to those system characteristics that challenge novice users’ executive control. Age effects interact strongly with a system’s demand profile and especially in early phases of skill acquisition and after intervals without practice. Therefore, older samples act like a magnifying glass for evaluating interface design and may be employed preferentially for this reason. Our results clearly demonstrate the potential of interface design to improve the accessibility of information technology for older users. In addition to limiting demands for perceptual processing and motor control, minimizing demands for controlled processing by reducing and simplifying choices and decisions is an important goal in such efforts and it is relevant for designing controls and for designing dialogues.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Bundesanstalt für Straßenwesen/Federal Highway Research Institute (BASt) [Project 82.175/2000].

Author Biographies

Georg Jahn is an assistant professor for cognitive psychology at the University of Greifswald in Germany. He obtained his PhD in psychology from the University of Regensburg in 2001.

Josef F. Krems is a full professor and the head of the Department of Cognitive and Industrial Psychology at Chemnitz University of Technology. He received his PhD in psychology from the University of Regensburg in 1984.

References

Anderson

J. R.

(1982). Acquisition of cognitive skill. Psychological Review, 89, 369-406.

Bainbridge

Quintanilla

(Eds.). (1989). Developing skills with information technology. Chichester, UK: Wiley.

Baron

Cerella

(1993). Laboratory tests of the disuse account of cognitive decline. In Cerella

Rybash

J. M.

Hoyer

W. J.

Commons

(Eds.), Adult information processing: Limits on loss (pp. 175-203). San Diego, CA: Academic Press.

Cerella

(Ed.). (1990). Aging and information-processing rate. San Diego, CA: Academic Press.

Cohen-Mansfield

Creedon

M. A.

Malone

T. B.

Kirkpatrick

M. J.

III Dutra

L. A.

Herman

R. P.

(2005). Electronic memory aids for community-dwelling elderly persons: Attitudes, preferences, and potential utilization. Journal of Applied Gerontology, 24, 3-20.

Craik

F. I. M.

Bialystok

(2006). Cognition through the lifespan: Mechanisms of change. Trends in Cognitive Sciences, 10(3), 131-138.

Czaja

S. J.

(1996). Aging and the acquisition of computer skills. In Rogers

W. A.

Fisk

A. D.

Walker

(Eds.), Aging and skilled performance: Advances in theory and applications (pp. 201-220). Hillsdale, NJ: Erlbaum.

Czaja

S. J.

Lee

C. C.

(2007). Information technology and older adults. In Sears

Jacko

J. A.

(Eds.), The human-computer interaction handbook: Fundamentals, evolving technologies, and emerging applications (2nd ed., pp. 777-792). Mahwah, NJ: Erlbaum.

Fisk

A. D.

Fisher

D. L.

Rogers

W. A.

(1992). General slowing alone cannot explain age-related search effects: Reply to Cerella (1991). Journal of Experimental Psychology: General, 121, 73-78.

10.

Fisk

A. D.

Rogers

W. A.

Charness

Czaja

S. J.

Sharit

(2009). Designing for older adults: Principles and creative human factors approaches (2nd ed.). Boca Raton, FL: CRC Press.

11.

Ghisletta

Kennedy

K. M.

Rodrigue

K. M.

Lindenberger

Raz

(2010). Adult age differences and the role of cognitive resources in perceptual-motor skill acquisition: Application of a multilevel negative exponential model. Journals of Gerontology Series B-Psychological Sciences and Social Sciences, 65(2), 163-173.

12.

Haider

Frensch

P. A.

(2002). Why aggregated learning follows the power law of practice when individual learning does not: Comment on Rickard (1997, 1999), Delaney et al. (1998), and Palmeri (1999). Journal of Experimental Psychology: Learning, Memory, and Cognition, 28, 392-406.

13.

Heathcote

Brown

Mewhort

D. J. K.

(2000). The power law repealed: The case for an exponential law of practice. Psychonomic Bulletin & Review, 7, 185-207.

14.

Jahn

Krems

J. F.

Gelau

(2009). Skill acquisition while operating in-vehicle information systems: Interface design determines the level of safety-relevant distractions. Human Factors, 51(2), 136-151.

15.

Jastrzembski

T. S.

Charness

(2007). The model human processor and the older adult: Parameter estimation and validation within a mobile phone task. Journal of Experimental Psychology: Applied, 13, 224-248.

16.

Kieras

D. E.

(1996). A guide to GOMS model usability evaluation using NGOMSL. In Helander

Landauer

Prabhu

(Eds.), The handbook of human-computer interaction (2nd ed., pp. 733-766). Amsterdam, Netherlands: North-Holland.

17.

Lee

F. J.

Anderson

J. R.

(2001). Does learning a complex task have to be complex? A study in learning decomposition. Cognitive Psychology, 42, 267-316.

18.

Loftus

G. R.

(2002). Analysis, interpretation, and visual presentation of experimental data. In Pashler

Wixted

(Eds.), Stevens’ handbook of experimental psychology: Methodology in experimental psychology (3rd ed., Vol. 4, pp. 339-390). New York, NY: Wiley.

19.

Mayr

Kliegl

Krampe

R. T.

(1996). Sequential and coordinative processing dynamics in figural transformations across the life span. Cognition, 59, 61-90.

20.

McLaughlin

A. C.

Rogers

W. A.

Fisk

A. D.

(2009). Using direct and indirect input devices: Attention demands and age-related differences. ACM Transactions on Computer-Human Interaction, 16(1), 1-15.

21.

McMellon

C. A.

Schiffman

L. G.

(2002). Cybersenior empowerment: How some older individuals are taking control of their lives. Journal of Applied Gerontology, 21(2), 157-175.

22.

Mykityshyn

A. L.

Fisk

A. D.

Rogers

W. A.

(2002). Learning to use a home medical device: Mediating age-related differences with training. Human Factors, 44, 354-364.

23.

Newell

Rosenbloom

P. S.

(1981). Mechanisms of skill acquisition and the law of practice. In Anderson

J. R.

(Ed.), Cognitive skills and their acquisition (pp. 1-55). Hillsdale, NJ: Erlbaum.

24.

Nichols

T. A.

Rogers

W. A.

Fisk

A. D.

(2006). Design for aging. In Salvendy

(Ed.), Handbook of human factors and ergonomics (3rd ed., pp. 1418-1445). Hoboken, NJ: Wiley.

25.

Raz

(2000). Aging of the brain and its impact on cognitive performance: Integration of structural and functional findings. In Craik

F. I. M.

Salthouse

T. A.

(Eds.), The handbook of aging and cognition (2nd ed., pp. 1-90). Mahwah, NJ: Erlbaum.

26.

Ritter

F. E.

Schooler

L. J.

(2001). The learning curve. In Smelser

N. J.

Baltes

P. B.

(Eds.), International encyclopedia of the social and behavioral sciences (pp. 8602-8605). Amsterdam: Pergamon.

27.

Rogers

W. A.

Fisk

A. D.

(2010). Toward a psychological science of advanced technology design for older adults. Journals of Gerontology Series B-Psychological Sciences and Social Sciences, 65, 645-653.

28.

Rogers

W. A.

Fisk

A. D.

McLaughlin

A. C.

Pak

(2005). Touch a screen or turn a knob: Choosing the best device for the job. Human Factors, 47, 271-288.

29.

Rogers

W. A.

Stronge

A. J.

Fisk

A. D.

(2006). Technology and aging. In Nickerson

R. S.

(Ed.), Reviews of human factors and ergonomics (Vol. 1, pp. 130-171). Santa Monica, CA: Human Factors and Ergonomics Society.

30.

Salthouse

T. A.

(1996). The processing-speed theory of adult age differences in cognition. Psychological Review, 103, 403-428.

31.

Schmidt

R. A.

Bjork

R. A.

(1992). New conceptualizations of practice: Common principles in three paradigms suggest new concept for training. Psychological Science, 3, 207-217.

32.

Tun

P. A.

Lachman

M. E.

(2010). The association between computer use and cognition across adulthood: Use it so you won’t lose it? Psychology and Aging, 25, 560-568.

33.

Vercruyssen

(1997). Movement control and speed of behavior. In Fisk

A. D.

Rogers

W. A.

(Eds.), Handbook of human factors and the older adult (pp. 55-86). San Diego, CA: Academic Press.

34.

Verhaeghen

Kliegl

Mayr

(1997). Sequential and coordinative complexity in time-accuracy functions for mental arithmetic. Psychology and Aging, 12, 555-564.

35.

Wasylyshyn

Verhaeghen

Sliwinski

M. J.

(2011). Aging and task switching: A meta-analysis. Psychology and Aging, 26(1), 15-20.

36.

White

McConnell

Clipp

Bynum

Teague

Navas

Craven

Halbrecht

(1999). Surfing the net in later life: A review of the literature and pilot study of computer use and quality of life. Journal of Applied Gerontology, 18, 358-378.