Sunteți pe pagina 1din 22

ITUsefulnessand Easeof Use

dent to perceived usefulness, as opposed to


Perceived Usefulness, a parallel, direct determinantof system usage.
Perceived Ease of Implicationsare drawn for future research on
user acceptance.
Use, and User Keywords: User acceptance, end user
Acceptance of computing,user measurement
Information ACMCategories: H.1.2, K.6.1, K.6.2, K.6.3

Technology
Introduction
By: Fred D. Davis Information technologyoffersthe potentialforsub-
Computer and Information Systems stantially improving white collar performance
Graduate School of Business (Curley, 1984; Edelman, 1981; Sharda, et al.,
1988). But performance gains are often ob-
Administration structed by users' unwillingnessto accept and
University of Michigan use available systems (Bowen, 1986; Young,
Ann Arbor, Michigan 48109 1984). Because of the persistence and impor-
tance of this problem, explaining user accep-
tance has been a long-standingissue in MIS
research (Swanson, 1974; Lucas, 1975; Schultz
and Slevin, 1975; Robey, 1979; Ginzberg,1981;
Swanson, 1987). Althoughnumerousindividual,
organizational,and technologicalvariableshave
Abstract been investigated(Benbasat and Dexter, 1986;
Validmeasurement scales for predicting user Franz and Robey, 1986; Markus and Bjorn-
acceptance of computers are in short supply. Anderson, 1987; Robey and Farrow,1982), re-
Most subjective measures used in practice are search has been constrained by the shortage
unvalidated, and their relationship to system of high-qualitymeasures for key determinants
of user acceptance. Past research indicatesthat
usage is unknown. The present research de-
velops and validates new scales for two spe- many measures do not correlate highly with
cific variables, perceived usefulness and per- system use (DeSanctis, 1983; Ginzberg, 1981;
ceived ease of use, which are hypothesized to Schewe, 1976; Srinivasan, 1985), and the size
be fundamental determinants of user accep- of the usage correlationvaries greatlyfromone
tance. Definitionsfor these two variables were study to the next depending on the particular
used to develop scale items thatwere pretested measures used (Baroudi,et al., 1986; Barkiand
for content validityand then tested for reliability Huff,1985; Robey, 1979; Swanson, 1982, 1987).
and construct validityin two studies involving The developmentof improvedmeasures for key
a total of 152 users and four applicationpro- theoreticalconstructs is a research priorityfor
grams. Themeasures were refinedand stream- the informationsystems field.
lined, resultingin two six-item scales with reli- Aside fromtheir theoreticalvalue, better meas-
abilities of .98 for usefulness and .94 for ease ures for predictingand explaining system use
of use. The scales exhibited high convergent, would have great practicalvalue, both for ven-
discriminant, and factorialvalidity.Perceiveduse- dors who would like to assess user demand for
fulness was significantlycorrelatedwithbothself- new design ideas, and for informationsystems
reported current usage (r=.63, Study 1) and managers withinuser organizationswho would
self-predictedfutureusage (r= .85, Study2). Per- like to evaluate these vendor offerings.
ceived ease of use was also significantlycorre-
lated with currentusage (r=.45, Study 1) and Unvalidatedmeasures are routinelyused in prac-
futureusage (r=.59, Study 2). In both studies, tice today throughout the entire spectrum of
usefulness had a significantlygreater correla- design, selection, implementationand evaluation
tion with usage behavior than did ease of use. activities.For example: designers withinvendor
Regression analyses suggest that perceived organizationssuch as IBM(Gould,et al., 1983),
ease of use may actually be a causal antece- Xerox(Brewley,et al., 1983), and DigitalEquip-

MIS Quarterly/September1989 319

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulnessand Ease of Use

ment Corporation(Good, et al., 1986) measure believe that the systems is too hardto use and
user perceptions to guide the development of that the performancebenefits of usage are out-
new informationtechnologies and products;in- weighed by the effort of using the application.
dustry publications often report user surveys That is, in additionto usefulness, usage is theo-
(e.g., Greenberg,1984; Rushinekand Rushinek, rizedto be influencedby perceived ease of use.
1986); several methodologies for software se-
lection call for subjective user inputs (e.g., Perceived usefulness is defined here as "the
Goslar, 1986; Kleinand Beck, 1987); and con- degree to which a person believes that using
temporarydesign principles emphasize meas- a particularsystem would enhance his or her
uringuser reactionsthroughoutthe entiredesign job performance."This follows from the defini-
process (Andersonand Olson 1985; Gould and tion of the word useful: "capableof being used
Lewis, 1985; Johansen and Baker,1984; Mantei advantageously."Withinan organizationalcon-
and Teorey, 1988; Norman,1983; Shneiderman, text, people are generally reinforcedfor good
1987). Despite the widespread use of subjec- performance by raises, promotions, bonuses,
tive measures in practice, littleattentionis paid and other rewards(Pfeffer,1982; Schein, 1980;
to the qualityof the measures used or how well Vroom,1964). A system high in perceived use-
they correlate with usage behavior. Given the fulness, in turn,is one for which a user believes
low usage correlations often observed in re- in the existence of a positive use-performance
search studies, those who base importantbusi- relationship.
ness decisions on unvalidatedmeasures may Perceived ease of use, in contrast,refersto "the
be gettingmisinformedabout a system's accept-
degree to which a person believes that using
abilityto users. a particularsystem would be free of effort."This
The purpose of this research is to pursue better follows from the definitionof "ease": "freedom
measures for predictingand explaininguse. The from difficultyor great effort."Effortis a finite
investigation focuses on two theoretical con- resource that a person may allocate to the vari-
structs, perceived usefulness and perceived ous activitiesfor which he or she is responsible
ease of use, which are theorized to be funda- (Radner and Rothschild, 1975). All else being
mental determinantsof system use. Definitions equal, we claim, an applicationperceived to be
forthese constructsare formulatedand the theo- easier to use than another is more likelyto be
reticalrationalefor their hypothesized influence accepted by users.
on system use is reviewed.New, multi-item meas-
urementscales for perceivedusefulness and per-
ceived ease of use are developed, pretested,
and then validatedin two separate empiricalstud- Theoretical Foundations
ies. Correlationand regression analyses exam- The theoreticalimportanceof perceived useful-
ine the empiricalrelationshipbetween the new ness and perceivedease of use as determinants
measures and self-reportedindicantsof system of user behavioris indicatedby several diverse
use. The discussion concludes by drawingim- lines of research. The impactof perceived use-
plicationsfor future research. fulness on system utilizationwas suggested by
the workof Schultzand Slevin (1975) and Robey
(1979). Schultz and Slevin (1975) conductedan
exploratoryfactor analysis of 67 questionnaire
Perceived Usefulness and items, which yielded seven dimensions. Of
these, the "performance" dimension,interpreted
Perceived Ease of Use by the authors as the perceived "effect of the
Whatcauses people to accept or rejectinforma- model on the manager'sjob performance,"was
tiontechnology?Amongthe manyvariablesthat most highlycorrelatedwithself-predicteduse of
may influencesystem use, previousresearchsug- a decision model (r=.61). Usingthe Schultzand
gests two determinantsthat are especially im- Slevinquestionnaire,Robey (1979) findsthe per-
portant.First,people tend to use or not use an formance dimensionto be most correlatedwith
applicationto the extent they believe it willhelp two objectivemeasures of system usage (r=.79
them performtheir job better. We refer to this and .76). Buildingon Vertinsky,et al.'s (1975)
firstvariableas perceived usefulness. Second, expectancy model, Robey (1979) theorizes that:
even if potentialusers believe that a given ap- "A system that does not help people perform
plicationis useful, they may, at the same time, their jobs is not likelyto be received favorably

320 MIS Quarterly/September1989

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulness and Ease of Use

in spite of careful implementationefforts" (p. to learn a computerlanguage. The self efficacy


537). Althoughthe perceived use-performance paradigmdoes not offer a general measure ap-
contingency, as presented in Robey's (1979) plicable to our purposes since efficacy beliefs
model, parallelsour definitionof perceived use- are theorized to be situationally-specific,with
fulness, the use of Schultz and Slevin's (1975) measures tailored to the domain under study
performance factor to operationalize perform- (Bandura, 1982). Self efficacy research does,
ance expectancies is problematicfor several rea- however, provideone of several theoreticalper-
sons: the instrumentis empiricallyderived via pectives suggesting that perceived ease of use
exploratoryfactoranalysis;a somewhat low ratio and perceived usefulness functionas basic de-
of sample size to items is used (2:1); four of terminantsof user behavior.
thirteenitems have loadings below .5, and sev-
eral of the items clearly fall outside the defini-
tion of expected performance improvements
(e.g., "Myjob will be more satisfying,""Others Cost-benefitparadigm
will be more aware of what I am doing,"etc.). The cost-benefitparadigmfrombehavioraldeci-
An alternativeexpectancy-theoreticmodel, de- sion theory (Beach and Mitchell,1978; Johnson
and Payne, 1985; Payne, 1982) is also relevant
rived from Vroom (1964), was introducedand
to perceived usefulness and ease of use. This
tested by DeSanctis (1983). The use-perform-
research explains people's choice among vari-
ance expectancy was not analyzed separately
ous decision-makingstrategies (such as linear
from performance-rewardinstrumentalitiesand
rewardvalences. Instead,a matrix-orientedmeas- compensatory,conjunctive,disjunctiveand elmi-
urementprocedurewas used to producean over- nation-by-aspects)in terms of a cognitivetrade-
off betweenthe effortrequiredto employthe strat-
all index of "motivationalforce" that combined
these three constructs. "Force"had small but egy and the quality (accuracy) of the resulting
decision. This approach has been effective for
significant correlations with usage of a DSS
withina business simulationexperiment(corre- explainingwhy decision makersaltertheirchoice
lations rangedfrom.04 to .26). The contrastbe- strategies in response to changes in task com-
tween DeSanctis's correlationsand the ones ob- plexity.Althoughthe cost-benefit approach has
served by Robey underscorethe importanceof mainly concerned itself with unaided decision
measurement in predictingand explaininguse. making, recent work has begun to apply the
same form of analysis to the effectiveness of
informationdisplay formats (Jarvenpaa, 1989;
Kleinmuntzand Schkade, 1988).
Self-efficacytheory
The importanceof perceived ease of use is sup- Cost-benefitresearch has primarilyused objec-
tive measures of accuracyand effortin research
ported by Bandura's(1982) extensive research
on self-efficacy, defined as "judgmentsof how studies, downplayingthe distinctionbetween ob-
well one can execute courses of action required jective and subjective accuracy and effort. In-
to deal withprospectivesituations"(p. 122). Self- creased emphasison subjectiveconstructsis war-
efficacy is similarto perceived ease of use as ranted, however, since (1) a decision maker's
definedabove. Self-efficacybeliefs are theorized choice of strategy is theorized to be based on
to functionas proximaldeterminantsof behav- subjectiveas opposed to objectiveaccuracyand
ior. Bandura'stheory distinguishes self-efficacy effort(Beach and Mitchell,1978), and (2) other
research suggests that subjectivemeasures are
judgments from outcome judgments, the latter
often in disagreementwiththeirojbective coun-
being concerned with the extent to which a be-
havior,once successfully executed, is believed terparts(Abelson and Levi, 1985; Adelbrattand
to be linkedto valued outcomes. Bandura's"out- Montgomery,1980; Wright,1975). Introducing
come judgment"variableis similarto perceived measures of the decision maker'sown perceived
usefulness. Bandura argues that self-efficacy costs and benefits, independentof the decision
and outcome beliefs have differingantecedents actually made, has been suggested as a way
and that, "Inany given instance, behaviorwould of mitigatingcriticismsthatthe cost/benefit frame-
be best predicted by considering both self- work is tautological(Abelson and Levi, 1985).
The distinctionmade herein between perceived
efficacy and outcome beliefs" (p. 140).
usefulness and perceived ease of use is similar
Hill,et al. (1987) findthat both self-efficacyand to the distinctionbetween subjective decision-
outcome beliefs exert an influenceon decisions makingperformanceand effort.

MIS Quarterly/September 1989 321

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulnessand Easeof Use

Adoptionof innovations two components: attributedinformationquality


and attributedaccess quality.Potentialusers are
Research on the adoption of innovations also
hypothesized to select and use informationre-
suggests a prominentrole for perceived ease ports based on an implicitpsychologicaltrade-
of use. Intheir meta-analysisof the relationship off between informationqualityand associated
between the characteristicsof an innovationand costs of access. Swanson (1987) performedan
its adoption,Tornatzkyand Klein(1982) findthat
exploratoryfactor analysis in order to measure
compatibility,relative advantage, and complex- informationquality and access quality.A five-
ity have the most consistent significantrelation- factorsolutionwas obtained,withone factorcor-
ships across a broad range of innovationtypes. responding to informationquality (Factor #3,
Complexity,defined by Rogers and Shoemaker "value"),and one to access quality(Factor#2,
(1971) as "the degree to which an innovation "accessibility").Inspectingthe items that load on
is perceived as relativelydifficultto understand these factors suggests a close correspondence
and use" (p. 154), parallels perceived ease of to perceived usefulness and ease of use. Items
use quite closely. As Tornatzkyand Klein(1982) such as "important,""relevant,""useful,"and
pointout, however,compatibilityand relativead- "valuable"load stronglyon the value dimension.
vantage have both been dealt with so broadly Thus, value parallelsperceived usefulness. The
and inconsistentlyin the literatureas to be diffi- fact that relevance and usefulness load on the
cult to interpret. same factor agrees with informationscientists,
who emphasize the conceptual similaritybe-
tween the usefulness and relevance notions
(Saracevic,1975). Several of Swanson's "acces-
Evaluationof information
reports sibility"items, such as "convenient,""controlla-
Past research within MIS on the evaluation of ble," "easy," and "unburdensome,"correspond
informationreports echoes the distinctionbe- to perceived ease of use as defined above. Al-
tween usefulness and ease of use made herein. thoughthe studywas more exploratorythan con-
Larckerand Lessig (1980) factor analyzed six firmatory,with no attempts at construct valida-
items used to rate fourinformation reports.Three tion, it does agree withthe conceptualdistinction
items load on each of two distinct factors: (1) between usefulness and ease of use. Self-
perceived importance,whichLarckerand Lessig reportedinformationchannel use correlated.20
define as "the qualitythat causes a particular with the value dimension and .13 with the ac-
informationset to acquire relevance to a deci- cessibilitydimension.
sion maker,"and the extent to which the infor-
mationelements are "a necessary inputfortask
accomplishment," and (2) perceived usable-
ness, which is defined as the degree to which Non-MISstudies
"the informationformat is unambiguous, clear
or readable"(p. 123). These two dimensionsare Outside the MIS domain, a marketingstudy by
similarto perceived usefulness and perceived Hauserand Simmie (1981) concerninguser per-
ease of use as defined above, repsectively,al- ceptions of alternativecommunicationtechnolo-
though Larckerand Lessig refer to the two di- gies similarlyderivedtwo underlyingdimensions:
mensions collectivelyas "perceivedusefulness." ease of use and effectiveness, the latterbeing
Reliabilitiesfor the two dimensions fall in the similarto the perceived usefulness constructde-
range of .64-.77, short of the .80 minimallevel fined above. Bothease of use and effectiveness
recommended for basic research. Correlations were influentialin the formationof user prefer-
with actual use of informationreportswere not ences regardinga set of alternativecommuni-
addressed in their study. cation technologies. The human-computerinter-
action (HCI) research community has heavily
emphasized ease of use in design (Branscomb
and Thomas, 1984; Card, et al., 1983; Gould
and Lewis, 1985). For the most part, however,
Channeldispositionmodel these studies have focused on objective meas-
Swanson (1982, 1987) introducedand tested a ures of ease of use, such as task completion
model of "channeldisposition"forexplainingthe time and errorrates. In many vendor organiza-
choice and use of informationreports.The con- tions, usabilitytesting has become a standard
cept of channel dispositionis defined as having phase in the product development cycle, with

322 MIS Quarterly/September1989

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulnessand Easeof Use

large investmentsin test facilitiesand instrumen- retained, yielding 10 items for each construct.
tation.Althoughobjective ease of use is clearly Next, a field study (Study 1) of 112 users con-
relevantto user performancegiven the system cerning two differentinteractivecomputer sys-
is used, subjectiveease of use is more relevant tems was conducted in orderto assess the reli-
to the users' decision whetheror not to use the ability and construct validity of the resulting
system and may not agree with the objective scales. The scales were further refined and
measures (Carrolland Thomas, 1988). streamlined to six items per construct. A lab
study (Study2) involving40 participantsand two
graphics systems was then conducted. Data
fromthe two studies were then used to assess
Convergenceof findings the relationship between usefulness, ease of
There is a strikingconvergence among the wide use, and self-reportedusage.
range of theoretical perspectives and research
studies discussed above. Although Hill, et al. Psychometriciansemphasize that the validityof
a measurementscale is builtin fromthe outset.
(1987) examined learninga computerlanguage, As Nunnally(1978) pointsout, "Ratherthan test
Larckerand Lessig (1980) and Swanson (1982,
the validityof measures after they have been
1987) dealt with evaluating informationreports, constructed, one should ensure the validityby
and Hauser and Simmie (1981) studied com-
the plan and procedures for construction"(p.
municationtechnologies,all are supportiveof the
258). Carefulselection of the initialscale items
conceptualand empiricaldistinctionbetweenuse-
fulness and ease of use. The accumulatedbody helps to assure the scales willpossess "content
of knowledgeregardingself-efficacy,contingent validity,"defined as "the degree to which the
score or scale being used represents the con-
decision behavior and adoption of innovations
provides theoreticalsupport for perceived use- cept about which generalizations are to be
made" (Bohrnstedt,1970, p. 91). In discussing
fulness and ease of use as key determinants
content validity,psychometriciansoften appeal
of behavior.
to the "domainsampling model," (Bohrnstedt,
From multipledisciplinaryvantage points, per- 1970; Nunnally,1978) which assumes there is
ceived usefulness and perceived ease of use a domainof content correspondingto each vari-
are indicatedas fundamentaland distinctcon- able one is interested in measuring. Candidate
structsthat are influentialin decisions to use in- items representativeof the domain of content
formationtechnology. Althoughcertainlynot the should be selected. Researchers are advised to
only variables of interest in explaininguser be- begin by formulatingconceptual definitionsof
havior (for other variables, see Cheney, et al., what is to be measured and preparingitems to
1986; Davis, et al., 1989; Swanson, 1988), they fit the constructdefinitions(Anastasi, 1986).
do appear likelyto play a centralrole. Improved
measures are needed to gain furtherinsightinto Following these recommendations, candidate
items for perceived usefulness and perceived
the nature of perceived usefulness and per-
ease of use were generated based on theircon-
ceived ease of use, and their roles as determi-
nants of computer use. ceptual definitions,stated above, and then pre-
tested in order to select those items that best
fit the content domains. The Spearman-Brown
Prophecy formula was used to choose the
numberof items to generate for each scale. This
formulaestimates the numberof items needed
Scale Development and to achieve a given reliability based on the
Pretest number of items and reliabilityof comparable
existingscales. Extrapolatingfrompast studies,
A step-by-step process was used to develop the formula suggests that 10 items would be
new multi-itemscales having high reliabilityand needed for each perceptualvariableto achieve
validity.The conceptual definitionsof perceived reliabilityof at least .80 (Davis, 1986). Adding
usefulness and perceived ease of use, stated fouradditionalitems for each constructto allow
above, were used to generate 14 candidate for item elimination,it was decided to generate
itemsforeach constructfrompast literature.Pre- 14 items for each construct.
test interviewswere then conducted to assess
the semantic content of the items. Those items The initialitem pools for perceived usefulness
that best fitthe definitionsof the constructswere and perceived ease of use are given in Tables

MIS Quarterly/September1989 323

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulnessand Easeof Use

1 and 2, respectively. In preparingcandidate other items in order to yield a more pure indi-
items, 37 publishedresearch papers dealingwith cant of the conceptual variable.
user reactions to interactivesystems were re-
viewed in other to identifyvarious facets of the Pretest interviewswere performedto furtheren-
constructs that should be measured (Davis, hance content validityby assessing the corre-
1986). The items are wordedin referenceto "the spondence betweencandidateitemsand the defi-
electronicmail system," which is one of the two nitions of the variables they are intended to
test applicationsinvestigatedin Study 1, reported measure. Itemsthatdon'trepresenta construct's
below. The items withineach pool tend to have contentvery well can be screened out by asking
a lot of overlap in their meaning, which is con- individualsto rankthe degree to whicheach item
sistent with the fact that they are intended as matches the variable's definition,and eliminat-
measures of the same underlying construct. ing items receiving low rankings.In eliminating
Thoughdifferentindividualsmay attributeslightly items, we want to make sure not to reduce the
differentmeaning to particularitem statements, representativeness of the item pools. Our item
the goal of the multi-itemapproachis to reduce pools may have excess coverage of some areas
any extranneous effects of individualitems, al- of meaning (or substrata;see Bohrnstedt,1970)
lowing idiosyncrasies to be cancelled out by withinthe content domain and not enough of
Table 1. InitialScale Items for Perceived Usefulness

1. Myjob would be difficultto performwithoutelectronicmail.


2. Using electronicmail gives me greatercontrolover my work.
3. Using electronicmail improvesmy job performance.
4. The electronicmail system addresses my job-relatedneeds.
5. Using electronicmail saves me time.
6. Electronicmail enables me to accomplishtasks more quickly.
7. Electronicmail supportscriticalaspects of my job.
8. Using electronicmail allows me to accomplishmore workthan wouldotherwise be
possible.
9. Using electronicmail reduces the time I spend on unproductiveactivities.
10. Using electronicmail enhances my effectiveness on the job.
11. Using electronicmail improvesthe qualityof the workI do.
12. Using electronicmail increases my productivity.
13. Using electronicmail makes it easier to do my job.
14. Overall,I findthe electronicmail system useful in my job.

Table 2. InitialScale Items for Perceived Ease of Use

1. I often become confused when I use the electronicmailsystem.


2. I make errorsfrequentlywhen using electronicmail.
3. Interactingwiththe electronicmailsystem is often frustrating.
4. I need to consult the user manualoften when using electronicmail.
5. Interactingwiththe electronicmailsystem requiresa lot of my mentaleffort.
6. I find it easy to recover fromerrorsencounteredwhile using electronicmail.
7. The electronicmail system is rigidand inflexibleto interactwith.
8. I find it easy to get the electronicmailsystem to do what I want it to do.
9. The electronicmail system often behaves in unexpected ways.
10. I find it cumbersome,to use the electronicmail system.
11. My interactionwiththe electronicmail system is easy for me to understand.
12. It is easy for me to rememberhow to performtasks using the electronicmail system.
13. The electronicmail system provideshelpfulguidance in performingtasks.
14. Overall,I findthe electronicmail system easy to use.

324 MIS Quarterly/September1989

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulnessand Easeof Use

others. By asking individualsto rate the similar- over work"was retainedsince, althoughit was
ity of items to one another, we can performa ranked fairly low, it fell in the top 9 and may
clusteranalysis to determinethe structureof the tap an importantaspect of usefulness.
substrata,remove items where excess coverage
is suggested, and add items where inadequate
Lookingnow at perceived ease of use (Table
coverage is indicated. 4), we again find three main clusters. The first
Pretest participantsconsisted of a sample of 15 relates to physical effort, while the second re-
experienced computer users from the Sloan lates to mental effort.Selecting the six highest-
School of Management,MIT,includingfive sec- priorityitems and eliminatingthe seventh pro-
retaries, five graduate students and five mem- vides good coverage of these two clusters. Item
bers of the professionalstaff. In face-to-face in- 11 ("understandable")was reworded to read
terviews,participantswere asked to performtwo "clearand understandable"in an effortto pick
tasks, prioritizationand categorization, which up some of the content of item 1 ("confusing"),
were done separately for usefulness and ease which has been eliminated.The thirdcluster is
of use. For prioritization,they were first given somewhat more difficultto interpretbut appears
a card containingthe definitionof the targetcon- to be tappingperceptionsof how easy a system
structand asked to read it. Next,they were given is to learn. Rememberinghow to performtasks,
13 index cards each having one of the items using the manual, and relyingon system guid-
forthat constructwrittenon it. The 14thor "over- ance are all phenomenaassociated withthe proc-
all" item for each construct was omitted since ess of learningto use a new system (Nickerson,
its wordingwas almost identicalto the label on 1981; Robertsand Moran,1983). Furtherreview
the definitioncard (see Tables 1 and 2). Partici- of the literaturesuggests that ease of use and
pants were asked to rankthe 13 cards accord- ease of learning are strongly related. Roberts
ing to how well the meaning of each statement and Moran(1983) find a correlationof .79 be-
matched the given definitionof ease of use or tween objective measures of ease of use and
usefulness. ease of learning. Whiteside, et al. (1985) find
that ease of use and ease of learning are
For the categorization task, participantswere stronglyrelatedand conclude that they are con-
asked to put the 13 cards intothree to five cate- gruent. Studies of how people learn new sys-
gories so that the statements withina category tems suggest that learning and using are not
were most similarin meaning to each other and separate, disjoint activities, but instead that
dissimilarin meaning from those in other cate- people are motivatedto begin performingactual
gories. This was an adaptationof the "owncate- workdirectlyand try to "learnby doing"as op-
gories" procedure of Sherif and Sherif (1967). posed to going throughuser manuals or online
Categorizationprovidesa simple indicantof simi- tutorials(Carrolland Carrithers,1984; Carroll,
laritythat requiresless time and effortto obtain et al., 1985; Carrolland McKendree,1987).
than other similaritymeasurement procedures
such as paid comparisons. The similaritydata
was cluster analyzed by assigning to the same In this study, therefore, ease of learningis re-
clusteritems thatseven or more subjects placed garded as one substratumof the ease of use
in the same category. The clusters are consid- construct, as opposed to a distinct construct.
ered to be a reflectionof the domain substrata Since items 4 and 13 provide a ratherindirect
for each constructand serve as a basis of as- assessment of ease of learning,they were re-
sessing coverage, or representativeness,of the placed with two items that more directlyget at
item pools. ease of learning:"Learningto operate the elec-
tronic mail system is easy for me," and "I find
The resultingrankand cluster data are summa- it takes a lot of effortto become skillfulat using
rized in Tables 3 (usefulness) and 4 (ease of electronic mail." Items 6, 9 and 2 were elimi-
use). Forperceivedusefulness, noticethat items nated because they did not cluster with other
fall into three main clusters. The firstcluster re- items, and they received low priorityrankings,
lates to job effectiveness, the second to produc- which suggests that they do not fit well within
tivityand time savings, and the thirdto the im- the content domain for ease of use. Together
portance of the system to one's job. If we with the "overall"items for each construct,this
eliminate the lowest-rankeditems (items 1, 4, procedureyielded a 10-itemscale for each con-
5 and 9), we see that the three majorclusters struct to be empiricallytested for reliabilityand
each have at least two items. Item 2, "control constructvalidity.

MIS Quarterly/September1989 325

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulnessand Ease of Use

Table 3. Pretest Results: Perceived Usefulness


Old New
Item # Item Rank Item # Cluster
1 Job DifficultWithout 13 C
2 ControlOver Work 9 2
3 Job Performance 2 6 A
4 Addresses My Needs 12 C
5 Saves Me Time 11 B
6 WorkMoreQuickly 7 3 B
7 Criticalto MyJob 5 4 C
8 AccomplishMoreWork 6 7 B
9 Cut UnproductiveTime 10 B
10 Effectiveness 1 8 A
11 Qualityof Work 3 1 A
12 Increase Productivity 4 5 B
13 Makes Job Easier 8 9 C
14 Useful NA 10 NA

Table 4. Pretest Results: Perceived Ease of Use


Old New
Item # Item Rank Item # Cluster
1 Confusing 7 B
2 ErrorProne 13
3 Frustrating 3 3 B
4 Dependence on Manual 9 (replace) C
5 MentalEffort 5 7 B
6 ErrorRecovery 10
7 Rigid& Inflexible 6 5 A
8 Controllable 1 4 A
9 Unexpected Behavior 11
10 Cumbersome 2 1 A
11 Understandable 4 8 B
12 Ease of Remembering 8 6 C
13 Provides Guidance 12 (replace) C
14 Easy to Use NA 10 NA
NA Ease of Learning NA 2 NA
NA Effortto Become Skillful NA 9 NA

able on IBMsystems and offers both full-screen


Study 1 and command-drivenediting capabilities. The
A field study was conducted to assess the reli- questionnaire asked participants to rate the
ability,convergent validity,discriminantvalidity, extent to which they agree with each statement
and factorialvalidityof the 10-item scales re- by circlinga numberfromone to seven arranged
sultingfromthe pretest. A sample of 120 users horizontallybeneath anchor point descriptions
withinIBMCanada'sTorontoDevelopmentLabo- "StronglyAgree," "Neutral,"and "StronglyDis-
ratorywere given a questionnaireasking them agree." Inorderto ensure subjectfamiliarity with
to rate the usefulness and ease of use of two the systems being rated, instructionsasked the
systems availablethere: PROFS electronicmail participantsto skip over the section pertaining
and the XEDITfile editor. The computingenvi- to a given system if they never use it. Responses
ronmentconsisted of IBMmainframesaccessi- were obtained from 112 participants,for a re-
ble through327X terminals.The PROFS elec- sponse rate of 93%. Of these 112, 109 were
tronic mail system is a simple but limited users of electronic mail and 75 were users of
messaging facility for brief messages. (See XEDIT.Subjects had an average of six months'
Panko, 1988.) The XEDITeditoris widely avail- experiencewiththe two systems studied.Among

326 MIS Quarterly/September1989

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulnessand Easeof Use

the sample, 10 percent were managers, 35 per- differenttraitor withdifferentitems used to meas-
cent were administrativestaff, and 55 percent ure a differenttrait(Campbelland Fiske, 1959).
were professionalstaff (whichincludeda broad For perceived usefulness, 1,800 such compari-
mixof marketanalysts,productdevelopmentana- sons were confirmedwithoutexception. Of the
lysts, programmers,financial analysts and re- 1,800 comparisons for ease of use there were
search scientists). 58 exceptions (3%). This represents an unusu-
ally high level of discriminantvalidity(Campbell
and Fiske, 1959; Silk, 1971) and implies that
the usefulness and ease of use scales possess
Reliabilityand validity a high concentrationof trait variance and are
The perceived usefulness scale attained Cron- not strongly influenced by methodological
bach alpha reliabilityof .97 for both the elec- artifacts.
tronicmailand XEDITsystems, while perceived
ease of use achieved a reliabilityof .86 for elec- Table 5 gives a summaryfrequencytable of the
tronic mail and .93 for XEDIT.When observa- correlationscomprisingthe MTMMmatricesfor
tions were pooled for the two systems, alpha usefulness and ease of use. Fromthis table it
was .97 for usefulness and .91 for ease of use. is possible to see the separation in magnitude
between monotraitand heterotraitcorrelations.
Convergentand discriminantvaliditywere tested The frequencytable also shows that the hetero-
using multitrait-multimethod (MTMM)analysis trait-heteromethod correlationsdo not appearto
(Campbelland Fiske, 1959). The MTMMmatrix be substantiallyelevated above the heterotrait-
containsthe intercorrelationsof items (methods)
monomethodcorrelations.This is an additional
appliedto the two differenttest systems (traits),
electronic mail and XEDIT.Convergentvalidity diagnostic suggested by Campbell and Fiske
refers to whether the items comprisinga scale (1959) to detect the presence of method
behave as if they are measuringa common un- variance.
derlyingconstruct.In orderto demonstratecon- The few exceptions to the convergent and dis-
vergent validity,items that measure the same criminantvaliditythatdid occur, althoughnot ex-
trait should correlate highly with one another
tensive enough to invalidatethe ease of use
(Campbelland Fiske, 1959). That is, the ele- scale, all involved negatively phrased ease of
ments in the monotraittriangles (the submatrix
use items.These "reversed"items tendedto cor-
of intercorrelationsbetween items intended to
relate more with the same item used to meas-
measure the same construct for the same
ure a differenttraitthan they didwithotheritems
system) withinthe MTMMmatrices should be of the same trait, suggesting the presence of
large.Forperceivedusefulness,the 90 monotrait- common method variance. This is ironic,since
heteromethodcorrelationswere all significantat
reversed scales are typicallyused in an effort
the .05 level. For ease of use, 86 out of 90,
corre- to reduce common methodvariance.Silk (1971)
or 95.6%, of the monotrait-heteromethod
lationswere significant.Thus, our data supports similarlyobserved minordepartures from con-
the convergent validityof the two scales. vergent and discriminantvalidityfor reversed
items. The five positively worded ease of use
Discriminantvalidityis concerned with the abil- items had a reliabilityof .92 compared to .83
ity of a measurement item to differentiatebe- for the five negative items. This suggests an im-
tween objects being measured. For instance, provementin the ease of use scale may be pos-
withinthe MTMMmatrix,a perceived usefulness sible with the eliminationor reversal of nega-
item appliedto electronicmail should not corre- tively phrased items. Nevertheless, the MTMM
late too highly with the same item applied to analysis supported the ability of the 10-item
XEDIT.Failureto discriminatemay suggest the scales for each constructto differentiatebetween
presence of "commonmethod variance,"which systems.
means that an item is measuringmethodological
artifactsunrelatedto the target construct(such Factorialvalidityis concerned with whetherthe
as individualdifferences in the style of respond- usefulness and ease of use items form distinct
ingto questions (see Campbell,et al., 1967; Silk, constructs. A principal components analysis
1971) ). The test for discriminantvalidityis that using oblique rotation was performed on the
an item should correlatemore highlywith other twenty usefulness and ease of use items. Data
items intended to measure the same traitthan were pooled across the two systems, for a total
with either the same item used to measure a of 184 observations. The results show that the

MIS Quarterly/September1989 327

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulnessand Ease of Use

Table 5. Summary of Multitrait-Multimethod Analyses


Construct
Perceived Usefulness Perceived Ease of Use
Same Trait/ Different Same Trait/ Different
Diff. Method Trait Diff. Method Trait
Correlation Elec. Same Diff. Elec. Same Diff.
Size Mail XEDIT Meth. Meth. Mail XEDIT Meth. Meth.
-.20 to -.11 1
-.10 to -.01 6 1 5
.00 to .09 3 25 2 1 32
.10 to .19 2 27 2 5 40
.20 to .29 5 25 9 1 11
.30 to .39 7 14 2 2 1
.40 to .49 9 9
.50 to .59 4 3 11
.60 to .69 14 4 3 13
.70 to .79 20 11 3 8
.80to .89 7 26 2
.90 to .99 4
# Correlations 45 45 10 90 45 45 10 90

usefulness and ease of use items load on dis- how to performtasks"), which the pretest indi-
tinctfactors(Table6). The multitrait-multimethod cated was concerned withease of learning,was
analysisand factoranalysisbothsupportthe con- replaced by a reversal of item 9 ("easy to
struct validityof the 10-item scales. become skillful"),which was specifically de-
signed to more directly tap ease of learning.
These items include two from cluster C, one
Scale refinement each fromclusters A and B, and the overallitem.
In applied testing situations, it is importantto (See Table 4.) In order to improverepresenta-
keep scales as brief as possible, particularly tive coverage of the content domain, an addi-
when multiplesystems are going to be evalu- tional A item was added. Of the two remaining
ated. The usefulness and ease of use scales A items (#1, Cumbersome, and #5, Rigid and
were refined and streamlinedbased on results Inflexible),item5 is readilyreversedto form"flex-
from Study 1 and then subjected to a second ible to interactwith."This item was added to
roundof empiricalvalidationin Study2, reported form the sixth item, and the order of items 5
below. Applyingthe Spearman-Brownprophecy and 8 was permutedin order to prevent items
formulato the .97 reliabilityobtained for per- fromthe same cluster (items 4 and 5) fromap-
ceived usefulness indicatesthat a six-itemscale pearing next to one another.
composed of items having comparable reliabil-
In order to select six items to be used for the
ity would yield a scale reliabilityof .94. The five
positive ease of use items had a reliabilityof usefulness scale, an item analysis was per-
.92. Taken together, these findingsfrom Study formed. Corrected item-totalcorrelationswere
1 suggest that six items would be adequate to computed for each item, separately for each
achieve reliabilitylevels above .9 while main- system studied. Average Z-scores of these cor-
relationswere used to rankthe items. Items 3,
taining adequate validitylevels. Based on the
results of the field study, six of the 10 items for 5, 6, 8, 9 and 10 were top-rankeditems. Refer-
each construct were selected to form modified ring to the cluster analysis (Table 3), we see
scales. that this set is well-representativeof the content
domain, includingtwo items fromcluster A, two
For the ease of use scale, the five negatively from cluster B and one from cluster C, as well
worded items were eliminateddue to their ap- as the overall item (#10). The items were per-
parentcommon method variance, leaving items muted to prevent items from the same cluster
2, 4, 6, 8 and 10. Item 6 ("easy to remember fromappearingnext to one another.The result-

328 MIS Quarterly/September1989

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulnessand Easeof Use

Table 6. Factor Analysis of Perceived Usefulness and


Ease of Use Questions: Study 1
Factor 1 Factor 1
Scale Items (Usefulness) (Ease of Use)
Usefulness
1 Qualityof Work .80 .10
2 Control over Work .86 -.03
3 WorkMoreQuickly .79 .17
4 Criticalto MyJob .87 -.11
5 Increase Productivity .87 .10
6 Job Performance .93 -.07
7 AccomplishMoreWork .91 -.02
8 Effectiveness .96 -.03
9 Makes Job Easier .80 .16
10 Useful .74 .23
Ease of Use
1 Cubersome .00 .73
2 Ease of Learning .08 .60
3 Frustrating .02 .65
4 Controllable .13 .74
5 Rigid& Inflexible .09 .54
6 Ease of Remembering .17 .62
7 MentalEffort -.07 .76
8 Understandable .29 .64
9 Effortto Be Skillful -.25 .88
10 Easy to Use .23 .72

ing six-item usefulness and ease of use scales (.69), and overall(.64). Allcorrelationswere sig-
are shown in the Appendix. nificantat the .001 level.
Regression analyses were performedto assess
the joint effects of usefulness and ease of use
Relationshipto use on usage. The effect of usefulness on usage,
Participants were asked to self-report their controllingfor ease of use, was significantat the
degree of currentusage of electronic mail and .001 level for electronic mail (b=.55), XEDIT
XEDITon six-position categorical scales with (b=.69), and pooled (b=.57). In contrast, the
boxes labeled "Don'tuse at all,""Use less than effect of ease of use on usage, controllingfor
once each week," "Use aboutonce each week," usefulness, was non-significantacross the board
"Use several times a week," "Use about once (b=.01 for electronic mail; b=.02 for XEDIT;
each day," and "Use several times each day." and b=.07 pooled). In other words, the signifi-
Usage was significantlycorrelatedwithboth per- cant pairwise correlationbetween ease of use
ceived usefulness and perceived ease of use and usage vanishes when usefulness is con-
for both PROFS mail and XEDIT.PROFS mail trolledfor. The regression coefficients obtained
usage correlated.56 with perceived usefulness for each individualsystem within each study
and .32 with perceived ease of use. XEDIT were not significantly different (F3, 178= 1.95,
usage correlated .68 with usefulness and .48 n.s.). As the relationshipbetween independent
withease of use. Whendata were pooled across variablesin a regression approachperfectlinear
systems, usage correlated .63 with usefulness dependence, multicollinearity can degrade the
and .45 withease of use. The overallusefulness- parameterestimates obtained.Althoughthe cor-
use correlationwas significantlygreaterthan the relations between usefulness and ease of use
ease of use-use correlationas indicated by a are significant, according to tests for multi-
test of dependent correlations (t181=3.69, collinearitythey are not large enough to com-
p<.001) (Cohen and Cohen, 1975). Usefulness promise the accuracy of the estimated regres-
and ease of use were significantlycorrelated sion coefficientssince the standarderrorsof the
with each other for electronicmail (.56), XEDIT estimates are low (.08 for both usefulness and

MIS Quarterly/September1989 329

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
IT Usefulness and Ease of Use

ease of use), and the covariances between the The study involved evaluating two IBM PC-
parameter estimates are negligible (-.004) based graphics systems: Chart-Master(by De-
(Johnston, 1972; Mansfieldand Helms, 1982). cision Resources, Inc.of Westport,CN) and Pen-
Based on partialcorrelationanalyses, the vari- draw (by Pencept, Inc. of Waltham,MA).Chart-
ance in usage explained by ease of use drops Master is a menu-drivenpackage that creates
by 98% when usefulness is controlledfor. The numericalbusiness graphs, such as bar charts,
regressionand partialcorrelationresultssuggest line charts, and pie charts based on parameters
that usefulness mediates the effect of ease of defined by the user. Throughthe keyboardand
use on usage, i.e., that ease of use influences menus, the user inputsthe data for, and defines
usage indirectlythroughits effect on usefulness the desired characteristicsof, the chart to be
(J.A. Davis, 1985). made. The user can specify a wide variety of
options relatingto titlefonts, colors, plotorienta-
tion, cross-hatching pattern, chart format, and
so on. The chart can then be previewedon the
screen, saved, and printed.Chart-Masteris a
successful commercialproductthat typifies the
Study 2 categoryof numericbusiness chartingprograms.
A lab study was performedto evaluate the six-
item usefulness and ease of use scales result- Pendrawis quite differentfromthe typicalbusi-
ing from scale refinementin Study 1. Study 2 ness chartingprogram.Ituses bit-mappedgraph-
was designed to approximateapplied prototype ics and a "directmanipulation"interfacewhere
testing or system selection situations,an impor- users draw desired shapes using a digitizer
tant class of situations where measures of this tablet and an electronic "pen"as a stylus. The
kind are likelyto be used in practice. In proto- digitizertablet supplants the keyboard as the
type testing and system selection contexts, pro- inputmedium. By drawingon a tablet, the user
spective users are typicallygiven a briefhands- manipulatesthe image, which is visible on the
on demonstrationinvolvingless than an hourof screen as it is being created. Pendraw offers
actually interactingwith the candidate system. capabilities typical of PC-based, bit-mapped
Thus, representativeusers are asked to rate the "paint"programs (see Panko, 1988), allowing
future usefulness and ease of use they would the user to performfreehanddrawingand select
expect based on relativelylittleexperience with from among geometric shapes, such as boxes,
the systems being rated. We are especially in- lines, and circles. A varietyof line widths, color
terested in the propertiesof the usefulness and selections and title fonts are available. The
ease of use scales when they are worded in digitizeris also capable of performingcharacter
a prospective sense and are based on limited recognition,converting hand-printercharacters
experience with the target systems. Favorable into various fonts (Ward and Blesser, 1985).
psychometric properties under these circum- Pencept had positionedthe Pendrawproductto
stances would be encouraging relative to their complete with business chartingprograms.The
use as early warning indicants of user accep- manualintroducesPendrawby guidingthe user
tance (Ginzberg,1981). throughthe process of creating a numericbar
chart. Thus, a key marketing issue was the
The lab study involved40 voluntaryparticipants
extent to whichthe new productwouldcompete
who were evening MBAstudents at Boston Uni-
favorablywithestablishedbrands,such as Chart-
versity. They were paid $25 for participatingin Master.
the study. They had an average of five years'
workexperience and were employed full-timein
several industries,includingeducation (10 per- Participantswere given one hour of hands-on
cent), government(10 percent),financial(28 per- experience with Chart-Masterand Pendraw,
cent), health (18 percent),and manufacturing(8 using workbooksthat were designed to follow
percent). They had a range of priorexperience the same instructionalsequence as the user
with computers in general (35 percent none or manuals for the two products,while equalizing
limited;48 percent moderate; and 17 percent the style of writingand eliminatingvalue state-
extensive) and personal computers in particular ments (e.g., "See how easy that was to do?").
(35 percent none or limited;48 percent moder- Half of the participantstried Chart-Masterfirst
ate; and 15 percent extensive) but were unfa- and half tried Pendraw first. After using each
miliarwith the two systems used in the study. package, a questionnairewas completed.

330 MIS Quarterly/September 1989

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulness and Ease of Use

Reliabilityand validity questions were worded as follows: "Assuming


Pendrawwouldbe availableon my job, I predict
Cronbachalpha was .98 for perceived useful- that I willuse it on a regularbasis in the future,"
ness and .94 for perceived ease of use. Con- followed by two seven-point scales, one with
vergent validitywas supported,withonly two of likely-unlikelyend-pointadjectives,the other,re-
72 monotrait-heteromethodcorrelations falling versed in polarity,withimprobable-probable end-
below significance. Ease of use item 4 (flexibil-
point adjectives. Such self-predictions,or "be-
ity),appliedto Chart-Master, was not significantly havioralexpectations,"are among the most ac-
correlatedwith either items 3 (clear and under- curate predictors available for an individual's
standable) or 5 (easy to become skillful).This future behavior (Sheppard, et al., 1988; War-
suggests that, contraryto conventionalwisdom, shaw and Davis, 1985). For Chart-Master,use-
flexibilityis not always associated with ease of fulness was significantly correlated with self-
use. As Goodwin(1987) pointsout, flexibilitycan
predicted usage (r=.71, p<.001), but ease of
actually impair ease of use, particularly for use was not (r=.25, n.s.) (Table 8). Chart-
novice users. With item 4 omitted, Cronbach Masterhad a non-significantcorrelationbetween
alpha for ease of use would increase from .94 ease of use and usefulness (r=.25, n.s.). For
to .95. Despite the two departures to conver-
Pendraw,usage was significantlycorrelatedwith
gent validityrelated to ease of use item 4, no both usefulness (r=.59, p<.001) and ease of
exceptions to the discriminantvaliditycriteriaoc- use (r=.47, p<.001). The ease of use-useful-
curred across a total of 720 comparisons (360 ness correlation was significiant for Pendraw
for each scale).
(r=.38, p<.001). When data were pooled across
Factorialvaliditywas assessed by factor ana- systems, usage correlated.85 (p<.001) withuse-
lyzingthe 12 scale items using principalcompo- fulness and .59 (p<.001) with ease of use (see
nents extraction and oblique rotation.The re- Table 8). Ease of use correlatedwithusefulness
.56 (p<.001). The overall usefulness-use corre-
sultingtwo-factorsolutionis very consistent with
distinct,unidimensionalusefulness and each of lationwas significantlygreater than the ease of
use scales (Table7). Thus, as in Study 1, Study use-use correlation,as indicatedby a test of de-
2 reflects favorablyon the convergent, discrimi- pendentcorrelations(t77= 4.78, p<.001) (Cohen
nant, and factorialvalidityof the usefulness and and Cohen, 1975).
ease of use scales.
Regression analyses (Table 9) indicatethat the
effect of usefulness on usage, controllingfor
ease of use, was significantat the .001 level
Relationshipto use for Chart-Master (b = .69), Pendraw (b = .76) and
Participants were asked to self-predict their overall (b=.75). In contrast, the effect of ease
future use of Chart-Masterand Pendraw. The of use on usage, controllingfor usefulness, was

Table 7. Factor Analysis of Perceived Usefulness


and Ease of Use Items: Study 2
Factor 1 Factor 2
Scale Items (Usefulness) (Ease of Use)
Usefulness
1 WorkMoreQuickly .91 .01
2 Job Performance .98 -.03
3 Increase Productivity .98 -.03
4 Effectiveness .94 .04
5 Makes Job Easier .95 -.01
6 Useful .88 .11
Ease of Use
1 Easy to Learn -.20 .97
2 Controllable .19 .83
3 Clear & Understandable -.04 .89
4 Flexible .13 .63
5 Easy to Become Skillful .07 .91
6 Easy to Use .09 .91

MIS Quarterly/September 1989 331

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulness and Ease of Use

Table 8. Correlations Between Perceived Usefulness,


Perceived Ease of Use, and Self-Reported
System Usage
Correlation
Usefulness Ease of Use Ease of Use
& Usage & Usage & Usefulness
Study 1
ElectronicMail(n- 109) .56*** .32*** .56***
XEDIT(n=75) .68*** .48*** .69***
Pooled (n=184) .63*** .45*** .64***
Study 2
Chart-Master(n = 40) .71*** .25 .25
Pendraw(n = 40) .59*** .47*** .38**
Pooled (n = 80) .85*** .59*** .56***
Davis, et al. (1989) (n= 107)
Wave 1 .65*** .27** .10
Wave 2 .70*** .12 .23**
*** p<.001 ** p<.01 * p<.05

Table 9. Regression Analyses of the Effect of Perceived


Usefulness and Perceived Ease of Use on
Self-Reported Usage
Independent Variables
Usefulness Ease of Use R2
Study 1
ElectronicMail(n = 109) .55*** .01 .31
XEDIT(n= 75) .69*** .02 .46
Pooled (n=184) .57*** .07 .38
Study 2
Chart-Master(n = 40) .69*** .08 .51
Pendraw(n=40) .76*** .17 .71
Pooled (n = 80) .75*** .17* .74
Davis, et al. (1989) (n= 107)
After1 Hour .62*** .20*** .45
After14 Weeks .71** -.06 .49
*** p<.001 ** *
p<.01 p<.05
non-significantfor both Chart-Master(b=.08, throughusefulness. Partialcorrelationanalysis
n.s.) and Pendraw(b=.17, n.s.) when analyzed indicates that the variance in usage explained
separately and borderlinesignificantwhen ob- by ease of use drops by 91% when usefulness
servationswere pooled (b= .17, p<.05). The re- is controlledfor. Consistent with Study 1, these
gression coefficients obtained for Pendrawand regressionand partialcorrelationresultssuggest
Chart-Masterwere not significantlydifferent(F3, that usefulness mediates the effect of ease of
74 = .014, n.s.). Multicollinearityis ruled out since use on usage. The implicationsof this are ad-
the standarderrorsof the estimates are low (.07 dressed in the followingdiscussion.
for both usefulness and ease of use) and the
covariances between the parameter estimates
are negligible (-.004). Discussion
Hence, as in Study 1, the significantpairwise The purposeof this investigationwas to develop
correlations between ease of use and usage and validate new measurement scales for per-
drop dramaticallywhen usefulness is controlled ceived usefulness and perceived ease of use,
for, suggesting that ease of use operates two distinctvariables hypothesizedto be deter-

332 MIS Quarterly/September 1989

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulnessand Easeof Use

minantsof computerusage. This effortwas suc- fulness was correlated.63 withself-reportedcur-


cessful inseveral respects. The new scales were rent use in Study 1 and .85 with self-predicted
found to have strong psychometric properties use in Study 2. Perceived ease of use was cor-
and to exhibitsignificantempiricalrelationships related .45 withuse in Study 1 and .69 in Study
with self-reportedmeasures of usage behavior. 2. The same pattern of correlations is found
Also, several new insightswere generated about when correlationsare calculated separately for
the natureof perceived usefulness and ease of each of the two systems in each study (Table
use, and their roles as determinants of user 8). These correlations,especiallythe usefulness-
acceptance. use link,compare favorablywith other correla-
tions between subjective measures and self-
The new scales were developed, refined, and
reporteduse found in the MIS literature.Swan-
streamlined in a several-step process. Explicit son's (1987) "value"dimension correlated .20
definitionswere stated, followedby a theoretical withuse, while his "accessibility"dimensioncor-
analysis from a variety of perspectives, includ- related .13 with self-reporteduse. Correlations
ing: expectancy theory;self-efficacytheory;be- between "userinformationsatisfaction"and self-
havioraldecision theory;diffusionof innovations;
reporteduse of .39 (Barkiand Huff, 1985) and
marketing;and human-computerinteraction,re- .28 (Baroudi,et al., 1986) have been reported.
gardingwhy usefulness and ease of use are hy- "Realismof expectations"has been foundto be
pothesized as importantdeterminantsof system correlated .22 with objectively measured use
use. Based on the stated definitions,initialscale
items were generated. To enhance content va- (Ginzberg,1981) and .43 withself-reporteduse
(Barkiand Huff,1985). "Motiviational force"was
lidity,these were pretested in a small pilotstudy, correlated.25 withsystem use, objectivelymeas-
and several items were eliminated.The remain- ured (DeSanctis, 1983). Among the usage cor-
ing items, 10 foreach of the two constructs,were relationsreportedin the literature,the .79 corre-
tested for validityand reliabilityin Study 1, a lationbetween "performance"and use reported
field study of 112 users and two systems (the
by Robey (1979) stands out. Recallthat Robey's
PROFS electronic mail system and the XEDIT
file editor).Itemanalysis was performedto elimi- expectancy model was a key underpinningfor
the definitionof perceived usefulness stated in
nate moreitemsand refineothers,furtherstream- this article.
liningand purifyingthe scales. The resultingsix-
item scales were subjected to furtherconstruct
validationin Study 2, a lab study of 40 users One of the most significantfindings is the rela-
and two systems: Chart-Master(a menu-driven tive strength of the usefulness-usage relation-
business chartingprogram)and Pendraw(a bit-
ship compared to the ease of use-usage rela-
mapped paint programwith a digitizertablet as tionship. In both studies, usefulness was
its inputdevice).
significantlymore stronglylinkedto usage than
The new scales exhibitedexcellent psychomet- was ease of use. Examiningthe jointdirecteffect
ric characteristics.Convergentand discriminant of the two variableson use in regression analy-
validity were strongly supported by multitrait- ses, this differencewas even more pronounced:
multimethodanalyses in both validationstudies. the usefulness-usage relationship remained
These two data sets also providedstrong sup- large, while the ease of use-usage relationship
portforfactorialvalidity:the patternof factorload- was diminished substantially(Table 8). Multi-
ings confirmedthat a prioristructureof the two collinearityhas been ruled out as an explana-
instruments,withusefulness items loadinghighly tion for the results using specific tests for the
on one factor, ease of use items loading highly presence of multicollinearity.In hindsight, the
on the other factor, and small cross-factorload- prominence of perceived usefulness makes
ings. Cronbachalphareliability for perceiveduse- sense conceptually:users are driven to adopt
fulness was .97 in Study 1 and .98 in Study 2. an applicationprimarilybecause of the functions
Reliabilityfor ease of use was .91 in Study 1 it performsfor them, and secondarily for how
and .94 in Study 2. These findingsmutuallycon- easy or hard it is to get the system to perform
firmthe psychometricstrengthof the new meas- those functions. For instance, users are often
urementscales. willingto cope with some difficultyof use in a
system that provides criticallyneeded function-
As theorized, both perceived usefulness and ality. Althoughdifficultyof use can discourage
ease of use were significantly
correlatedwithself- adoption of an otherwise useful system, no
reportedindicantsof system use. Perceiveduse- amount of ease of use can compensate for a

MIS Quarterly/September1989 333

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulness and Ease of Use

system that does not performa useful function. tern of results in a two-wave study (Tables 8
The prominenceof usefulness over ease of use and 9). Inthatstudy, MBAstudentsubjects were
has importantimplicationsfordesigners, particu- asked to fillout a questionnaireaftera one-hour
larly in the human factors tradition,who have introductionto a word processing program,and
tended to overemphasize ease of use and over- again 14 weeks later. Usage intentions were
look usefulness (e.g., Branscomband Thomas, measured at both time periods, and self-
1984; Chin, et al., 1988; Shneiderman, 1987). reportedusage was measured at the latertime
Thus, a major conclusion of this study is that period. Intentionswere significantlycorrelated
perceived usefulness is a strong correlate of with usage (.35 and .63 for the two points in
user acceptance and should not be ignored by time, respectively).Unlikethe results of Studies
those attemptingto design or implementsuc- 1 and 2, Davis, et al. (1989) found a significant
cessful systems. directeffect of ease of use on usage, controlling
for usefulness, after the one-hour trainingses-
From a causal perspective, the regression re- sion (Table9), althoughthis evolved into a non-
sults suggest that ease of use may be an ante- significanteffect as of 14 weeks later. In gen-
cedent to usefulness, rather than a parallel, eral, though, Davis, et al. (1989) found useful-
direct determinant of usage. The significant ness to be more influentialthan ease of use in
pairwise correlationbetween ease of use and drivingusage behavior,consistent withthe find-
usage all but vanishes when usefulness is con- ings reportedabove.
trolledfor. This, coupled with a significantease
of use-usefulness correlationis exactly the pat- Furtherresearch willshed more lighton the gen-
tern one would expect if usefulness mediated eralityof these findings.Anotherlimitationis that
between ease of use and usage (e.g., J.A. the usage measures employed were self-
Davis, 1985). That is, the results are consistent reported as opposed to objectively measured.
with an ease of use --> usefulness --> usage Not enough is currentlyknownabout how accu-
chain of causality. These results held both for rately self-reportsreflect actual behavior.Also,
pooled observations and for each individual since usage was reportedon the same ques-
system (Table 8). The causal influenceof ease tionnaireused to measure usefulness and ease
of use on usefulness makes sense conceptu- of use, the possibilityof a halo effect should not
ally, too. All else being equal, the easier a be overlooked. Futureresearch addressing the
system is to interactwith,the less effortneeded relationshipbetween these constructs and ob-
to operate it, and the more effortone can allo- jectivelymeasured use is needed before claims
cate to other activities (Radner and Rothschild, about the behavioral predictiveness can be
1975), contributingto overall job performance. madeconclusively.These limitations notwithstand-
Goodwin(1987) also argues forthis flowof cau- ing, the results represent a promising step
sality, concludingfromher analysis that: "There towardthe establishmentof improvedmeasures
is increasing evidence that the effective func- for two importantvariables.
tionalityof a system depends on its usability"
(p. 229). This intriguinginterpretationis prelimi-
nary and should be subjected to furtherexperi-
mentation. If true, however, it underscores the Research implications
theoreticalimportanceof perceived usefulness. Futureresearch is needed to address how other
variables relate to usefulness, ease of use, and
This investigationhas limitationsthat should be acceptance. Intrinsicmotivation,for example,
pointed out. The generality of the findings re- has received inadequate attentionin MIStheo-
mains to be shown by futureresearch. The fact ries. Whereas perceived usefulness is con-
that similarfindingswere observed, withrespect cerned withperformanceas a consequence use,
to boththe psychometricpropertiesof the meas- intrinsicmotivationis concerned with the rein-
ures and the patternof empiricalassociations, forcementand enjoymentrelatedto the process
across two differentuser populations,two differ- of performinga behaviorper se, irrespectiveof
ent systems, and two differentresearch settings whatever external outcomes are generated by
(lab and field), providessome evidence favoring such behavior (Deci, 1975). Although intrinsic
externalvalidity. motivationhas been studiedin the design of com-
putergames (e.g., Malone, 1981), it is just be-
In addition, a follow-upto this study, reported ginningto be recognized as a potentialmecha-
by Davis, et al. (1989) found a very similarpat- nism underlyinguser acceptance of end-user

334 MIS Quarterly/September 1989

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulnessand Easeof Use

systems (Carrolland Thomas, 1988). Currently, throughconcept screening and prototypetest-


the role of affective attitudes is also an open ing to post-implementation assessment. The fact
issue. While some theorists argue that beliefs that the measures performedwell psychometri-
influence behavior only via their indirectinflu- cally both after brief introductionsto the target
ence on attitudes (e.g., Fishbein and Ajzen, system (Study 2, and Davis, et al., 1989) and
1975), others view beliefs and attitudes as co- aftersubstantialuser experience withthe system
determinantsof behavioralintentions(e.g., Tri- (Study 1, and Davis, et al., 1989) is promising
andis, 1977), and still others view attitudes as concerning their appropriateness at various
antecedents of beliefs (e.g., Weiner, 1986). points in the life cycle. Practitionersgenerally
Counterto Fishbeinand Ajzen's(1975) position, evaluate systems not only to predictacceptabil-
both Davis (1986) and Davis, et al. (1989) found ity but also to diagnose the reasons underlying
that attitudes do not fully mediate the effect of lack of acceptance and to formulateinterven-
perceivedusefulness and perceivedease of use tions to improveuser acceptance. Inthis sense,
on behavior. research on how usefulness and ease of use
can be influencedby various externallycontrol-
It should be emphasized that perceived useful- lable factors, such as the functionaland inter-
ness and ease of use are people's subjective face characteristicsof the system (Benbasatand
appraisalof performanceand effort,respectively, Dexter, 1986; Bewley, et al., 1983; Dickson, et
and do not necessarily reflect objective reality. al., 1986), development methodologies (Alavi,
Inthis study, beliefs are seen as meaningfulvari- 1984), training and education (Nelson and
ables in their own right,which functionas be- Cheney, 1987), and user involvementin design
havioraldeterminants,and are not regardedas (Baroudi,et al. 1986; Franz and Robey, 1986)
surrogatemeasures of objectivephenomena (as is important.The new measures introducedhere
is often done in MIS research, e.g., Ives, et al., can be used by researchers investigatingthese
1983; Srinivasan, 1985). Several MIS studies issues.
have observed discrepanciesbetween perceived
and actual performance(Cats-Bariland Huber, Althoughthere has been a growingpessimism
1987; Dickson, et al., 1986; Gallupe and De- in the field about the abilityto identifymeasures
Sanctis, 1988; Mcintyre,1982; Sharda, et al., that are robustlylinkedto user acceptance, the
1988). Thus, even if an applicationwouldobjec- view taken here is much more optimistic.User
tively improve performance,if users don't per- reactions to computers are complex and multi-
ceive it as useful, they'reunlikelyto use it (Alavi faceted. But if the field continues to systemati-
and Henderson,1981). Conversely,people may cally investigatefundamentalmechanisms driv-
overrate the performancegains a system has ing user behavior, cultivatingbetter and better
to offer and adopt systems that are dysfunc- measuresandcritically examiningalternativetheo-
tional.Giventhatthis study indicatesthat people retical models, sustainable progress is within
act accordingto theirbeliefs about performance, reach.
futureresearchis needed to understandwhy per-
formancebeliefs are often in disagreementwith
objective reality.The possibilityof dysfunctional
impacts generated by informationtechnology Acknowledgements
(e.g., Kottemannand Remus, 1987) emphasizes This research was supportedby grantsfromthe
that user acceptance is not a universalgoal and MITSloan School of Management,IBMCanada
is actuallyundesireablein cases where systems Ltd., and The Universityof MichiganBusiness
fail to providetrue performancegains. School. The author is indebted to the anony-
mous associate editor and reviewers for their
More research is needed to understand how
measures such as those introducedhere per- many helpfulsuggestions.
form in applied design and evaluationsettings.
The growingliteratureon design principles(An-
derson and Olson, 1985; Gould and Lewis, References
1985; Johansen and Baker, 1984; Manteiand Abelson, R.P. and Levi,A. "DecisionMakingand
Teorey, 1988; Shneiderman,1987) calls for the Decision Theory,"in The Handbookof Social
use of subjective measures at various points Psychology, thirdedition, G. Lindsayand E.
throughoutthe developmentand implementation Aronson (eds.), Knopf,New York,NY, 1985,
process, from the earliest needs assessment pp. 231-309.

MIS Quarterly/September1989 335

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulnessand Easeof Use

Adelbratt,T. and Montgomery,H. "Attractiveness MultitmethodMatrix,"Psychological Bulletin


of Decision Rules," Acta Psychologica (45), (56:9), March1959, pp. 81-105.
1980, pp. 177-185. Campbell,D.T., Siegman, C.R. and Rees, M.B.
Alavi, M. "An Analysis of the PrototypingAp- "Direction-of-Wording Effects in the Relation-
proachto Information Systems Development," ships Between Scales," Psychological Bulle-
Communications of the ACM (27:6), June tin (68:5), November 1967, pp. 293-303.
1984, pp. 556-563. Card,S.K., Moran,T.P. and Newell,A. The Psy-
Alavi,M. and Henderson,J.C. "AnEvolutionary chology of Human-Computer Interaction,
Strategyfor Implementinga Decision Support Erlbaum,Hillsdale,NJ, 1984.
System," Management Science (27:11), No- Carroll,J.M.and Carrithers,C. "Training Wheels
vember 1981, pp. 1309-1323. in a User Interface,"Communicationsof the
Anastasi, A. "EvolvingConcepts of Test Valida- ACM (27:8), August 1984, pp. 800-806.
tion," Annual Review of Psychology (37), Carroll, J.M. and McKendree, J. "Interface
1986, pp. 1-15. Design Issues for Advice-GivingExpertSys-
Anderson,N.S. and Olson, J.R. (eds.) Methods tems," Communicationsof the ACM (30:1,
for Designing Software to Fit Human Needs January 1987, pp. 14-31.
and Capabilities:Proceedings of the Work- Carroll,J.M., Mack,R.L.,Lewis,C.H.,Grishkow-
shop on Software Human Factors, National sky, N.L. and Robertson,S.R. "ExploringEx-
Academy Press, Washington,D.C., 1985. ploringa WordProcessor,"Human-Computer
Bandura,A. "Self-EfficacyMechanismin Human Interaction(1), 1985, pp. 283-307.
Agency,"AmericanPsychologist (37:2), Feb- Carroll,J.M. and Thomas, J.C. "Fun,"SIGCHI
ruary1982, pp. 122-147. Bulletin(19:3), January 1988, pp. 21-24.
Barki, H. and Huff, S. "Change, Attitude to Cats-Baril,W.L.and Huber,G.P. "DecisionSup-
Change, and Decision Support System Suc- port Systems for Ill-StructuredProblems:An
and Management(9:5), De-
cess," Information EmpiricalStudy," Decision Sciences (18:3),
cember 1985, pp. 261-268. Summer 1987, pp. 352-372.
Baroudi,J.J., Olson, M.H.and Ives, B. "AnEm- Cheney,P.H., Mann,R.I.and Amoroso,D.L."Or-
piricalStudy of the Impact of User Involve- ganizationalFactors Affectingthe Success of
ment on System Usage and InformationSat- End-User Computing,"Journal of Manage-
isfaction,"Communicationsof the ACM(29:3), ment Information Systems (3:1), Summer
March1986, pp. 232-238. 1986, pp. 65-80.
Beach, L.R. and Mitchell,T.R. "A Contingency Chin, J.P., Diehl, V.A. and Norman, K.L. "De-
Model for the Selection of Decision Strate- velopment of an Instrumentfor Measuring
gies," Academy of ManagementReview (3:3), User Satisfactionof the Human-Computer In-
July 1978, pp. 439-449. terface," CHI'88Human Factors in Comput-
Benbasat, I. and Dexter, A.S. "An Investigation ing Systems, Washington,D.C., May 15-19,
of the Effectiveness of Color and Graphical 1988, ACM,New York,NY, pp. 213-218.
PresentationUnderVaryingTime Constraints, Cohen, J. and Cohen, P. Applied MultipleRe-
MIS Quarterly(10:1), March1986, pp. 59-84. gression/ CorrelationAnalysis for the Behav-
Bewley, W.L.,Roberts,T.L.,Schoit, D. and Ver- ioral Sciences, Erlbaum,Hillsdale,NJ, 1975.
plank, W.L., "HumanFactors Testing in the Curley,K.F. "AreThere any Real Benefits from
Design of Xerox's8010 'Star'OfficeWorksta- Office Automation?"Business Horizons (4),
tion," CHI '83 Human Factors in Computing July-August1984, pp. 37-42.
Systems, Boston, December 12-15, 1983, Davis, F.D. "A Technology Acceptance Model
ACM,New York, NY, pp. 72-77. for EmpiricallyTesting New End-User Infor-
Bohrnstedt,G.W."Reliabilityand ValidityAssess- mation Systems: Theory and Results," doc-
ment in Attitude Measurement,"in Attitude toral dissertation,MITSloan School of Man-
Measurement, G.F. Summers (ed.), Rand- agement, Cambridge,MA, 1986.
McNally,Chicago, IL, 1970, pp. 80-99. Davis, F.D., Bagozzi, R.P. and Warshaw, P.R.
Bowen, W. "The Puny Payoff fromOffice Com- User Acceptance of ComputerTechnology:A
puters,"Fortune,May 26, 1986, pp. 20-24. Comparisonof Two TheoreticalModels,"Man-
Branscomb, L.M. and Thomas, J.C. "Ease of agement Science (35:8), August 1989, pp.
Use: A System Design Challenge,"IBMSys- 982-1003.
tems Journal(23), 1984, pp. 224-235. Davis, J.A. The Logic of Causal Order, Sage,
Campbell, D.T. and Fiske, D.W. "Convergent Beverly Hills,CA, 1985.
and DiscriminantValidationby the Multitrait- Deci, E.L. Intrinsic Motivation, Plenum, New

336 MIS Quarterly/September1989

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulnessand Easeof Use

York,NY, 1975. Management Science (27:1), January 1981,


DeSanctis, G. "ExpectancyTheoryas an Expla- pp. 33-56.
nationof VoluntaryUse of a Decision Support Hill,T., Smith, N.D., and Mann, M.F. "Role of
System," Psychological Reports (52), 1983, EfficacyExpectationsin Predictingthe Deci-
pp. 247-260. sion to Use Advanced Technologies: The
Dickson,G.W., DeSanctis, G. and McBride,D.J. Case of Computers,"Journalof Applied Psy-
"Understanding the Effectivenessof Computer chology, (72:2), May 1987, pp. 307-313.
Graphicsfor Decision Support:A Cumulative Ives,B., Olson,M.H.and Baroudi,J.J. "Themeas-
ExperimentalApproach,"Communicationsof urementof User Information Com-
Satisfaction,"
the ACM (29:1), January 1986, pp. 40-47. munications of the ACM (26:10), October
Edelmann, F. "Managers,ComputerSystems, 1983, pp. 785-793.
and Productivity,"MIS Quarterly(5:3), Sep- Jarvenpaa, S.L. "The Effect of Task Demands
tember 1981, pp. 1-19. and GraphicalFormaton Information Process-
Fishbein, M. and Ajzen, I. "Belief, Attitude,In- ing Strategies,"Management Science (35:3),
tention and Behavior: An Introduction to March1989, pp. 285-303.
Theoryand Research,"Addison-Wesley,Read- Johansen, R. & Baker E., "User Needs Work-
ing, MA 1975. shops: A New Approachto AnticipatingUser
Franz,C.R. and Robey, D. "Organizational Con- Needs for Advanced Office Systems," Office
text, User Involvement,and the Usefulness of Technology and People (2), 1984, pp. 103-
InformationSystems," Decision Sciences 119.
(17:3), Summer 1986, pp. 329-356. Johnson, E.J. and Payne, J.W. "Effortand Ac-
Gallupe,R.B., DeSanctis, G. and Dickson,G.W. curacy in Choice," Management Science
"Computer-BasedSupportfor GroupProblem (31:4), April1985, pp. 395-414.
Finding:An EmpiricalInvestigation," MISQuar- Johnston, J. Econometric Methods, McGraw-
terly (12:2), June 1988, pp. 277-296. Hill,New York,NY, 1972.
Ginzberg,M.J."EarlyDiagnosisof MISImplemen- Klein, G. and Beck, P.O. "A Decision Aid for
tation Failure:Promising Results and Unan- Selecting Among InformationSystems Alter-
swered Questions," Management Science natives,"MISQuarterly(11:2), June 1987, pp.
(27:4), April1981, pp. 459-478. 177-186.
Good, M.,Spine, T.M.,Whiteside,J. and George Kleinmuntz,D.N. and Schkade, D.A. "TheCog-
P. "User-DerivedImpactAnalysis as a Tool nitive Implicationsof InformationDisplays in
for UsabilityEngineering,"CHI'86HumanFac- Computer-SupportedDecision-Making,"Uni-
tors in ComputingSystems, Boston, April13- versity of Texas at Austin, Graduate School
17, 1986, ACM,New York,New Yorkpp. 241- of Business, Departmentof ManagementWork-
246. ing Paper 87/88-4-8, 1988.
Goodwin,N.C."Functionality and Usability,"
Com- Kottemann,J.E. and Remus, W.E. "Evidence
municationsof the ACM (30:3), March1987, and Principlesof Functionaland Dysfunctional
pp. 229-233. DSS," OMEGA(15:2), March1987, pp. 135-
Goslar, M.D. "CapabilityCriteriafor Marketing 143.
Decision Support Systems," Journalof Man- Larcker,D.F. and Lessig, V.P. "PerceivedUse-
agement InformationSystems (3:1), Summer A PsychometricExami-
fulness of Information:
1986, pp. 81-95. nation," Decision Sciences (11:1), January
Gould, J., Conti, J. and Hovanyecz, T. "Com- 1980, pp. 121-134.
posing letterswitha SimulatedListeningType- Lucas, H.C. "Performanceand the Use of an
writer,"Communicationsof the ACM (26:4), InformationSystem," Management Science
April1983, pp. 295-308. (21:8), April1975, pp. 908-919.
Gould,J.D. and Lewis C. "Designingfor Usabil- Malone, T.W. "Towarda Theory of Intrinsically
ity:Key Principlesand WhatDesignersThink," MotivatingInstruction,"CognitiveScience (4),
Communicationsof the ACM (28:3), March 1981, pp. 333-369.
1985, pp. 300-311. Mansfield,E.R. and Helms, B.P. "DetectingMul-
TheAmericanStatistician(36:3),
ticollinearity,"
Greenberg,K. "ExecutivesRate TheirPCs,"PC August 1982, pp. 158-160.
World,September 1984, pp. 286-292. Mantei, M.M. and Teorey, T.J. "Cost/Benefit
Hauser, J.R. and Simmie, P. "ProfitMaximizing Analysis for IncorporatingHuman Factors in
PerceptualPositions:An IntegratedTheoryfor the Software Lifecycle,"Communicationsof
the Selection of ProductFeatures and Price," the ACM (31:4), April1988, pp. 428-439.

MIS Quarterly/September1989 337

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulness and Ease of Use

Markus, M.L. and Bjorn-Anderson,N. "Power ber-December1975, pp. 321-343.


Over Users: It's Exercise by System Profes- Schein, E.H.Organizational Psychology,thirdedi-
sionals," Communicationsof the ACM(30:6), EnglewoodCliffs,NJ, 1980.
tion,Prentice-Hall,
June 1987, pp. 498-504. Schewe, C.D. "The Management Information
Mcintyre, S. "An Experimental Study of the System User:An Exploratory BehavioralAnaly-
Impact of Judgement-Based Marketing sis," Academy of ManagementJournal(19:4),
Models,"Management Science (28:1), Janu- December 1976, pp. 577-590.
ary 1982, pp. 17-23. Schultz, R.L. and Slevin, D.P. "Implementation
Nelson, R.R. and Cheney, P.H. "TrainingEnd and OrganizationalValidity:An EmpiricalIn-
Users: An ExploratoryStudy,"MIS Quarterly vestigation,"in ImplementingOperationsRe-
(11:4), December 1987, pp. 547-559. search/Management Science, R.L. Schultz
Nickerson,R.S. "WhyInteractiveComputerSys- and D.P. Slevin (eds.), American Elsevier,
tems Are Sometimes Not Used by People New York,NY, 1975, pp. 153-182.
Who MightBenefit from Them,"International Sharda,R., Barr,S.H., and McDonnell,J.C. "De-
Journalof Man-MachineStudies (15), 1981, cision Support System Effectiveness: A
pp. 469-483. Reviewand EmpiricalTest,"ManagementSci-
Norman, D.A. "Design Principles for Human- ence (34:2), February1988, pp. 139-159.
Computer Interfaces,"CHI '83 Human Fac- Sheppard,B.H., Hartwick,J. and Warshaw,P.R.
tors in ComputingSystems, Boston, Decem- "The Theory of Reasoned Action: A Meta-
ber 12-15, 1983, ACM,New York,NY, pp. 1- Analysis of Past Research with Recommen-
10. dations for Modifications and Future Re-
Nunnally,J. Psychometric Theory,McGraw-Hill, search," Journal of Consumer Research
New York,NY, 1978. (15:3), December 1988, pp. 325-343.
Panko, R.R. End-User Computing: Manage- Sherif, M. and Sherif, C.W. "The Own Catego-
ment, Applications, and Technology, Wiley, ries Approachin AttitudeResearch,"in Read-
New York,NY, 1988. ings in AttitudeTheoryand Measurement,M.
Payne, J. W. "ContingentDecision Behavior," Fishbein (ed.), Wiley, New York, NY, 1967,
Psychological Bulletin,(92:2), 1982, pp. 382- pp. 190-198.
402. Shneiderman,B. Designing the User Interface,
Pfeffer, J. Organizations and Organization Addison-Wesley,Reading, MA, 1987.
Theory,Pitman,Boston, MA, 1982. Silk, A.J. "Response Set and Measurementof
Radner, R. and Rothschild,M. "Onthe Alloca- Self-Designated OpinionLeadership,"Public
tion of Effort,"Journal of Economic Theory OpinionQuarterly(35), 1971, pp. 383-397.
(10), 1975, pp. 358-376. Srinivasan,A. "AlternativeMeasures of System
Roberts, T.L. and Moran,T.P. "The Evaluation Effectiveness:Associationsand Implications,"
of Text Editors:Methodologyand EmpiricalRe- MISQuarterly(9:3),September1985, pp. 243-
sults," Communicationsof the ACM (26:4), 253.
April1983, pp. 265-283. Swanson, E.B. "ManagementInformationSys-
Robey, D. "UserAttitudesand ManagementIn- tems: Appreciation and Involvement,"
Manage-
formationSystem Use,"Academy of Manage- ment Science (21:2), October 1974, pp. 178-
ment Journal(22:3),September1979, pp. 527- 188.
538. Swanson, E.B. "MeasuringUser Attitudesin MIS
Robey, D. and Farrow,D. "User Involvementin Research: A Review,"OMEGA(10:2), March
InformationSystem Development:A Conflict 1982, pp. 157-165.
Modeland EmpiricalTest," ManagementSci- Swanson, E.B. "Information ChannelDisposition
ence (28:1), January 1982, pp. 73-85. and Use," Decision Sciences (18:1), Winter
Rogers, E.M. and Shoemaker, F.F. Communi- 1987, pp. 131-145.
cation of Innovations:A Cross-CulturalAp- Swanson, E.B. InformationSystem Implemen-
proach, Free Press, New York,NY, 1971. tation:Bridgingthe Gap Between Design and
Rushinek,A. and Rushinek,S.F. "WhatMakes Utilization,Irwin,Homewood, IL, 1988.
Users Happy?"Communicationsof the ACM Tornatzky,L.G.and Klein,K.J."Innovation Char-
(29:7), July 1986, pp. 594-598. acteristicsand InnovationAdoption-Implemen-
Saracevic, T. "Relevance: A Review of and a tation: A Meta-Analysis of Findings," IEEE
Frameworkfor the Thinkingon the Notion in Transactions on Engineering Management
InformationScience," Journal of the Ameri- (EM-29:1),February1982, pp. 28-45.
can Society for InformationScience, Novem- Triandis,H.C. InterpersonalBehavior, Brooks/

338 MIS Quarterly/September1989

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
ITUsefulnessand Easeof Use

Cole, Monterey,CA, 1977. and Iconic Interfaces,"CHI'85 Proceedings,


Vertinsky,I., Barth, R.T. and Mitchell,V.F. "A San Francisco,April14-18, 1985, ACM,New
Study of OR/MS Implementationas a Social York,NY, pp. 185-191.
Change Process," in Implementing Opera- Wright,P. "ConsumerChoice Strategies: Sim-
tions Research/Management Science, R.L. plifyingvs. Optimizing,"Journalof Marketing
Schultz and D.P. Slevin (eds.), American Research (14:1), February 1975, pp. 429-
Elsevier, New York,NY, 1975, pp. 253-272. 433.
Vroom,V.H. Workand Motivation,Wiley, New Young, T.R. "The Lonely Micro,"Datamation
York,NY, 1964. (30:4), April1984, pp. 100-114.
Ward,J.R. and Blesser, B. "InteractiveRecog-
nitionof HandprinterCharactersfor Computer
Input,"IEEEComputerGraphics and Appli-
About the Author
cations, September 1985, pp. 24-37. Fred D. Davis is assistant professorat the Uni-
Warshaw, P.R. and Davis, F.D. "Disentangling versity of MichiganSchool of Business Admini-
BehavioralIntentionand BehavioralExpecta- stration. His doctoral research at the Sloan
tion,"Journalof ExperimentalSocial Psychol- School of Management,MIT,dealt with predict-
ogy (21), May 1985, pp. 213-228. ing and explaininguser acceptance of computer
Weiner,B. "Attribution,Emotion,and Action,"in technology. His current research interests in-
Handbookof Motivationand Cognition,R.M. clude computersupportfordecision making,mo-
Sorrentinoand E.T. Higgins (eds.), Guilford, tivationaldeterminantsof computeracceptance,
New York,NY, 1986, pp. 281-312. intentionsand expectations in human behavior,
Whiteside,J., Jones, S., Levy, P.S. and Wixon, and biased attributionsof the performanceim-
D. "UserPerformanceWithCommand,Menu, pacts of informationtechnology.

MIS Quarterly/September1989 339

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
IT Usefulness and Ease of Use

Appendix
Final MeasurementScales for Perceived Usefulness and
Perceived Ease of Use

Perceived Usefulness
Using CHART-MASTER in my job would enable me to accomplish tasks more quickly.
likely I I I I I I I unlikely
extremely quite slightly neither slightly quite extremely
Using CHART-MASTERwould improve my job performance.
likely 1- I I I I I unlikely
extremely quite slightly neither slightly quite extremely
Using CHART-MASTER in my job would increase my productivity.
likely I I I I I unlikely
extremely quite slightly neither slightly quite extremely
Using CHART-MASTER would enhance my effectiveness on the job.
likely I I I unlikely
extremely quite slightly neither slightly quite extremely
Using CHART-MASTERwould make it easier to do my job.
likely II I I I I I I unlikely
extremely quite slightly neither slightly quite extremely
I would find CHART-MASTER useful in my job.
likely II I I I I I I unlikely
extremely quite slightly neither slightly quite extremely

Perceived Ease of Use


Learning to operate CHART-MASTERwould be easy for me.
likely I I I I I I I I unlikely
extremely quite slightly neither slightly quite extremely
I would find it easy to get CHART-MASTERto do what I want it to do.
likely II I I I I I I unlikely
extremely quite slightly neither slightly quite extremely
My interaction with CHART-MASTERwould be clear and understandable.
likely I I I I I I I I unlikely
extremely quite slightly neither slightly quite extremely
I would find CHART-MASTERto be flexible to interact with.
likely I I I I I I I I unlikely
extremely quite slightly neither slightly quite extremely
It would be easy for me to become skillful at using CHART-MASTER.
likely I I I I I I I I unlikely
extremely quite slightly neither slightly quite extremely
I would find CHART-MASTEReasy to use.
likely I I I I I I I I unlikely
extremely quite slightly neither slightly quite extremely

340 MIS Quarterly/September 1989

This content downloaded from 130.184.237.6 on Thu, 6 Feb 2014 14:35:32 PM


All use subject to JSTOR Terms and Conditions
View publication stats