ARTICLE IN PRESS

0169-8141/$ - se

doi:10.1016/j.er

�Correspondfax: +1519 973

E-mail addr

International Journal of Industrial Ergonomics 36 (2006) 141–149

www.elsevier.com/locate/ergon

Development and evaluation of the Automotive SeatingDiscomfort Questionnaire (ASDQ)

Dannion R. Smitha, David M. Andrewsb,�, Peter T. Wawrowa

aErgonomics Department, Schukra of North America, 360 Silver Creek Industrial Dr., R.R. #1 Tecumseh, Lakeshore, Ont., Canada N8N 4Y3bDepartment of Kinesiology, 401 Sunset Ave, University of Windsor, Windsor, Ont., Canada N9B 3P4

Received 29 August 2005; accepted 29 September 2005

Available online 28 November 2005

Abstract

The purpose of this study was to develop and evaluate an assessment tool capable of quantifying subjective occupant discomfort in

automotive seating. To date, the majority of questionnaires present in the automotive seating industry have been designed using

questionable developmental methods with suspect statistical rigor. The Automotive Seating Discomfort Questionnaire (ASDQ) was

developed with statistically significant levels of readability, scale reliability, and face validity, using proven methods for questionnaire

development. Methods included key informant interviews, several pilot tests, and an experimental assessment that involved the subjective

evaluation of 3 identical front driver-side seats in 5 different seat positions over 3 sessions. The ASDQ was administered along side an

established automotive seating questionnaire to showcase increases in performance through differences in methodologies, scale usage,

and variable content. The ASDQ was shown to possess significant levels of reliability (po0:05) and internal consistency. A between

questionnaire comparison revealed significantly correlated subject responses (R2 ¼ 0:715), as well as significant differences between

similar questionnaire variables. The choice of measurement scale, increased variable content, establishment of face validity, and

thorough experimental methods resulted in the ASDQ measuring the construct of automotive seating discomfort in a more

comprehensive manner then previously developed industry questionnaires. It was concluded that the ASDQ reliably and repeatedly

measures the construct of automotive seating discomfort, contains face validity, has established a foundation for construct and content

validity development, and provides a comprehensive objective measure of occupant discomfort in automotive seating.

Relevance to industry

This study provides a rigorous questionnaire development process for the automotive seating industry. The resultant questionnaire can

be used in the evaluation of automotive seat designs.

r 2005 Elsevier B.V. All rights reserved.

Keywords: Automotive seating; Discomfort; Questionnaire; Visual analog scale

1. Introduction

An automotive seat represents a work environmentwhich must optimally position the occupant to perform thetask of driving, meet various safety requirements, and beacceptable to the driver’s comfort needs. It is this last pointthat is the most difficult to measure and satisfy, but isregarded as the main criteria by which seats are judged.

e front matter r 2005 Elsevier B.V. All rights reserved.

gon.2005.09.005

ing author. Tel.: +1519 253 3000x2433;

7056.

ess: dandrews@uwindsor.ca (D.M. Andrews).

Thus, automotive seat development is an iterative processresulting in several prototype builds, with each buildfollowed by a subjective evaluation of seat comfort(Kolich, 2004).In the automotive seating industry the goal is to have

seats that are more comfortable than competitive seats.However, comfort is a subjective construct that is difficultto interpret, measure, and specifically define due to itspsychophysical nature (Shen and Parsons, 1997). Thisambiguity is reflected in Random House Webster’s CollegeDictionary definition: ‘‘relief in affliction’’ and ‘‘a state ofease and satisfaction of bodily wants, with freedom from

ARTICLE IN PRESS

Table 1

Seating Discomfort Source Items by Questionnaire Iteration

Initial Intermediate Final

Cushion width Cushion width Cushion width

Cushion length Cushion length Cushion length

Cushion firmness Cushion firmness Cushion firmness

Side cushion support Cushion bolsters Cushion bolsters

Mid cushion support Cushion center Cushion center

Side cushion comfort Cushion contour Cushion contour

Mid cushion comfort Cushion aesthetics Trim

Cushion contour Cushion pressure Trim friction

Cushion aesthetics Trim Trim feel

Cushion pressure Trim friction Backrest height

Trim comfort Trim feel Backrest width

Trim touch Trim aesthetics Backrest firmness

Trim aesthetics Backrest height Backrest bolsters

Trim pressure Backrest width Backrest contour

Backrest height Backrest firmness Lumbar stiffness

Backrest width Backrest bolsters Lumbar prominence

Backrest firmness Backrest middle Lumbar support

Side backrest support Backrest contour Lumbar height

Mid backrest support Backrest aesthetics Lumbar pressure

Side backrest comfort Backrest pressure Overall discomfort

Mid backrest comfort Lumbar stiffness

Backrest contour Lumbar prominence

Backrest aesthetics Lumbar support

Backrest pressure Lumbar height

Lumbar stiffness Lumbar pressure

Lumbar prominence Overall discomfort

Lumbar comfort

Lumbar location

Lumbar pressure

Overall discomfort

In bold: reworded/incorporated into a new source item label.

In italics: deleted item.

Initial: thirty source items identified from a thorough review of literature.

Intermediate: the resultant source items after initial key informant and

pilot testing.

Final: the final 20 source items with significantly associated questions and

wording within the Automotive Seating Discomfort Questionnaire

(ASDQ).

D.R. Smith et al. / International Journal of Industrial Ergonomics 36 (2006) 141–149142

pain and anxiety’’ (Steinmetz, 1997). Measuring gradationsof comfort is a difficult task since once a feeling of relief orsatisfaction has been attained, providing a feeling of morecontentment is intangible.

In contrast, discomfort is a construct that is proposed tolie on the opposite end of a continuum and is thought to beeasier for subjects to identify a degree of affliction.Supporting this assumption, numerous studies have chosento explore occupant discomfort (i.e. Crane et al., 2004; ElFalou et al., 2003; Kolich, 2000; Porter et al., 2003; Reedet al., 1991; Shen and Parsons, 1997). Given this wide bodyof supporting literature, discomfort has taken on numerousdefinitions, including that of Shen and Parsons (1997) as ‘‘ageneric and subjective sensation that arises when humanand physiological homeostasis, psychological well-being,or both, are negatively affected’’ and by Steinmetz (1997)as ‘‘an absence of comfort or ease; hardship or mild pain’’.This hardship or mild pain can be more readily identified,providing seat developers a target to eliminate that shouldtranslate into a more comfortable seat. Thus, the currentchallenge is to determine how physical seat properties andoccupant perceptions contribute to the construct ofdiscomfort.

Items specific to the internal physical components of theseat have been identified through available anthropometricseat component data (i.e. thigh, buttocks), objective designstandards (i.e. seat length, seat width), and recent relatedliterature. This has resulted in 30 internal discomfortsource items related to seating and formed the basis forquestionnaire content (Table 1) (Giacomin and Quattro-colo, 1997; Goonetilleke and Feizhou, 2001; Gyi andPorter, 1999; Kolich, 2004; Kolich, 2000; Reed et al., 1991,1994).

The analysis of recent related literature also identified 3key external factors known to contribute to occupantseating discomfort: seating duration, hand reach, andvibration (Blair et al., 1998; Jung and Choe, 1996; Reedand Massie, 1996). Reed and Massie (1996) showed that82.5% of comfort score variance was accounted for afterbeing seated for 20min. Jung and Choe (1996) showed thatinclusion of proper arm postures influence shoulder andhip position, forcing the subject to inadvertently assumedifferent postures, which ultimately increases the realism ofthe simulation. Vibration has been shown to interact withdiscomfort on a continuum (Dhingra et al., 2003). Thus,when vibration levels are low, discomfort evaluations aredominated by physical seat characteristics and the effectsof vibration are negated (Dhingra et al., 2003). Theseexternal factors must be controlled during questionnaireevaluation, in accordance with the reviewed literature, toincrease experimental trial realism and overall question-naire significance.

Seat evaluation methods include ride-and-drive trials,where subjective evaluations are performed in a vehiclewhile driving, and/or seating buck simulations where thedriving experience is recreated in a static setting (El Falouet al., 2003; Kolich, 2000, 2003; Shen and Parsons, 1997;

Reed and Massie, 1996; Reed et al., 1994). Subjectiveratings are collected mainly by questionnaires and/orverbal interviews. Numerous questionnaires are currentlyavailable to rate automotive seating comfort, howeversupporting evidence states most were created withoutresearch scrutiny and/or proper research design (Craneet al., 2004; Kolich, 2004; Kolich and Taboun, 2004). Todate, the automotive seating industry does not have a goldstandard questionnaire with which to measure the con-struct of seating comfort (Kolich and Taboun, 2004).Although a gold standard does not currently exist in the

industry, the Automobile Seat Comfort Survey (Kolich,2000) is the most established questionnaire in automotiveseating literature. The Automobile Seat Comfort Survey(Kolich, 2000) has been shown to be a reliable tool inproviding numeric ratings of occupant seating comfort.However, scale selection, variable omission, seat selection,and subject size used during the development of thisquestionnaire warrant further consideration. Following

ARTICLE IN PRESSD.R. Smith et al. / International Journal of Industrial Ergonomics 36 (2006) 141–149 143

that discomfort is continuous and should be measuredon a continuum, a 7-point Likert scale, as used in theAutomobile Seat Comfort Survey (Kolich, 2000), isunfavourable due to the intermediate anchors, implyingthat discomfort is a divisible construct and not continuous.However, this questionnaire has been used in Kolich (2000,2003, 2004), and Kolich and Taboun, (2004) all returningsignificant results.

Outside of the automotive industry the Quality of HealthCare in Inflammatory Bowel Disease (QUOTE-IBD)questionnaire developed by van der Eijk et al. (2001)possesses a strong methodological background and statis-tically significant reliability and validity. Beyond theparallel of questionnaire development, the QUOTE-IBDis applicable to seating discomfort due to its use of a visualanalog scale (VAS) to rate pain. VAS is a direct estimationmethod scale that is designed to elicit from a subject adirect quantitative estimate concerning the magnitude ofan attribute (Streiner and Norman, 2003). This isaccomplished by using a line of fixed length, usually10 cm, with anchors located at the extreme ends and nowords describing the intermediate positions (Streiner andNorman, 2003). This provides a continuum on whichsubject responses can be placed. As previously defined, theconstruct of discomfort encompasses both pain andpsychological well-being. Thus, using a VAS, which isknown to provide accurate quantitative estimates of bothattributes, is optimal for evaluating automotive seatingdiscomfort.

The aims of the current study were to develop aquestionnaire that has an acceptable level of readabilityas defined by a Flesch–Kincaid readability score, statisti-cally significant (0:3opo0:8) questions containing keysource items, and face validity as defined in Streiner andNorman (2003). Ideally, the questionnaire should havehigh levels of scale repeatability and contain reliable andinternally consistent sub-scales. It is hypothesized that theestablished questionnaire will provide reliable estimates ofdiscomfort between and within gender, day, and seatposition factors. Ultimately, the developed questionnairewill measure the construct of seating discomfort with alevel of detail that is currently lacking in the automotiveseating industry, utilizing aspects of several establishedquestionnaires in the literature.

2. Methods

2.1. Development of the automotive seating discomfort

questionnaire (ASDQ)

2.1.1. Key informant interviews

Thirty source items identified in the review of literature(Table 1) were tested using a key informant interviewprocess as defined in Streiner and Norman (2003). Theseinterviews consisted of product designers, project managersand engineers, an ergonomic specialist, and sales represen-tatives, all familiar with the automotive seating industry.

This served to omit ambiguous, jargon filled, and/ordouble-barreled questions existing in the questionnaire aswell as to select an aesthetically pleasing layout. Flesch–Kincaid grade level readability tests were performed on alliterations of the ASDQ to ensure an appropriate level ofreadability (between grades 7 and 8).

2.1.2. Pilot testing

Several pilot tests were performed in order to improvestatistical significance and decrease error contained withinthe ASDQ. This process assisted in ensuring that appro-priate levels of questionnaire readability, wording, aes-thetics, and statistical significance of both question variableand resultant data was achieved and maintained in parallelwith questionnaire modification.Three pilot tests were conducted. Pilot test #1 (N ¼ 11)

consisted of volunteers using the ASDQ to evaluate a carseat in situ, as well as providing verbal feedback.Corresponding changes were made to improve questionwording and levels of statistical significance. Pilot test #2,which was conducted with the same subjects (N ¼ 11) andseat/vehicle as pilot test #1, resulted in further revisions(sentence length, wording, and layout changes) to theASDQ. Pilot test #3 (N ¼ 15) used a portion of the subjectpool from the previous pilot tests as well as additionalvolunteers. A test/retest evaluation of a single seat wasconducted with a separation period of 24 h. A factoranalysis was conducted for questionnaire sub-scale identi-fication. Specifically, a principal component analysis wasused to show cumulative variance which established thenumber of sub-scales present as well as identifying both themain sub-scales and those variables associated within eachidentified sub-scale. Sub-scale internal consistency wascalculated using a correlation method with resultantCronbach’s a values establishing significance (a40:7).

2.2. Establishing the ASDQ—experimental trials

2.2.1. Subjects

Eight male and 16 female subjects participated in allexperimental trials (mean age, height, and mass were36.3718.5 yr, 1.6870.09m, and 74.5713.9 kg, respec-tively). Experimental procedures received approval fromthe University of Windsor Research Ethics Board andsubjects provided written consent prior to participation.Subjects were randomly recruited from the Windsor areathrough an employment agency and compensated accord-ingly for their time. Experimental trials took place at theUniversity of Windsor—Human Kinetics Building.

2.2.2. Experimental procedures

Experimental trials were conducted over a 12-dayperiod. Subjects were divided into 2 equal groups. Eachsubject participated on 3 separate days (labelled 1, 2, and 3)in a constant section (morning or afternoon) with twoother subjects (Table 2). Each day was divided into 4

ARTICLE IN PRESS

Table 2

Subject seat position protocol

Group 1—Day 1

Calendar 1 Session Seat Calendar 2 Session Seat

C A B A B C A B A B

Position Null Down-in Down-out Up-in Up-out Position Null Down-in Down-out Up-in Up-out

Morning section

9:00–9:20 1 1 2 3 N/A N/A 9:00–9:20 13 7 8 9 N/A N/A

9:30–9:50 2 2 3 1 N/A N/A 9:30–9:50 14 8 9 7 N/A N/A

10:00–10:20 3 3 1 2 N/A N/A 10:00–10:20 15 9 7 8 N/A N/A

10:30–10:50 4 2 N/A N/A 3 1 10:30–10:50 16 8 N/A N/A 9 7

11:00–11:20 5 1 N/A N/A 2 3 11:00–11:20 17 7 N/A N/A 8 9

11:30–11:50 6 3 N/A N/A 1 2 11:30–11:50 18 9 N/A N/A 7 8

Afternoon section

1:00–1:20 7 4 N/A N/A 5 6 1:00–1:20 19 10 N/A N/A 11 12

1:30–1:50 8 5 N/A N/A 6 4 1:30–1:50 20 11 N/A N/A 12 10

2:00–2:20 9 6 N/A N/A 4 5 2:00–2:20 21 12 N/A N/A 10 11

2:30–2:50 10 5 6 4 N/A N/A 2:30–2:50 22 11 12 10 N/A N/A

3:00–3:20 11 4 5 6 N/A N/A 3:00–3:20 23 10 11 12 N/A N/A

3:30–3:50 12 6 4 5 N/A N/A 3:30–3:50 24 12 10 11 N/A N/A

The above is an example of subject and seat positioning protocol. There were 6 days (12 calendar) of total testing. Each 30-min session was divided into a

20-min condition time, where subjects occupied an assigned seat and watched a movie, a 3–7-min evaluation period, where both the ASDQ and

Automotive Seating Comfort Survey (Kolich, 2000) were administered, and a 3–7-min stretch period, where subjects walked around to prevent muscular

ache and stiffness.

D.R. Smith et al. / International Journal of Industrial Ergonomics 36 (2006) 141–149144

sections with 6 sessions. Each session was 30-min induration.

Three identical front driver-side seats (A–C) from a 2003model sedan, were used throughout the project. Seats Aand B were equipped with identical Schukra brand 4-wayplastic power lumbar support systems (motion capabilitiesup/down and in/out) while seat C was not equipped with alumbar support system. The seats were individuallymounted on identical custom base supports to mimic seatfastening in an automobile. All 3 seats were subjected to anevaluation to verify that the seats being used and thelumbar supports fitted in seats A and B were in factidentical. The report concluded that all 3 seats and the 2lumbar supports had identical physical characteristics.

The seats were situated in a row, 36 cm apart, to simulatea typical automotive seating design. Dividers were erectedand hung between the seats during questionnaire responseto restrict visual contact between subjects. A movie wasplayed during each session, doubling as subject entertain-ment and as a distraction.

Five seat positions were used (Table 2): Down-In (1),Down-Out (2), null (3,6), Up-In (4), and Up-Out (5). Theselabels corresponded with specific lumbar support position-ing. ‘‘Up’’ refers to the lumbar support being located at thevertical maximum of its tracking. ‘‘Down’’ refers to thelumbar support being situated at the vertical base of itstracking. ‘‘In’’ refers to the lumbar support plastic in anon-engaged (no curvature present) position. ‘‘Out’’ refersto the lumbar support plastic in a maximally engaged(maximum curvature) position. Seat A was always in the‘‘In’’ position and seat B was always in the ‘‘Out’’ position.

Seat A was used for seat positions 1 and 4, seat B for seatpositions 2 and 5, and seat C for seat positions 3 and 6. Allsubjects experienced the null seat position twice a daywhere as the other positions were only experienced once.Seat C positions 3 and 6 (seat position 6 is the positionlabel for the second instance each subject was assigned tothe null seat position on a given day) were used to confirmuni-variate reliability between subject responses.Subjects were randomly assigned identifying numbers

that corresponded with predetermined randomized seatingpositions for each session. Subjects were required to sit inthe assigned seat continuously for 20min and wereinstructed to position themselves in an everyday drivingposture that included placing their right leg in anoutstretched position to simulate interaction with theaccelerator pedal. Every 4min, subjects were asked to taketheir right arm and reach outwards to simulate realisticmotions used to interact with an automobile instrumentpanel (i.e. radio, heat controls).The ASDQ and the Automobile Seat Comfort Survey

(Kolich, 2000) were administered after each 20-min seatingsession. This timeframe allowed the seat foam to approachits base-line properties (Reed and Massie, 1996). [Note: theversion of the Automobile Seat Comfort Survey used wasan intermediate version that contained 3 extra variables allshown to have statistically significant reliability in Kolich(2000).]Questionnaire completion took subjects between 3 and

7min. Subjects were instructed to vacate the experimentroom for 3–7min post-evaluation. During this time, eachseat containing a lumbar support was manipulated to

ARTICLE IN PRESS

α = 0.959*α = 0.917*α = 0.921*

α = 0.975*

01234567

1,2,3 4,5,6,7,8,9 7,10,11,12,13,14 14,15,16,17,18,19Trim Cushion Backrest Lumbar

Questions Contained in each Sub-scale

* Cronbach's α > 0.070 is significant.

Tot

al N

umbe

r of

Var

iabl

es

Fig. 1. ASDQ sub-scale variable by question breakdown. Cronbach’s avalues for the 4 identified sub-scales were found to be significant. The total

number of variables and the associated variable question identifiers

contained within each sub-scale are presented.

p = 0.043

0

1

2

3

4

5

1 3 4 5Seat Position

Mea

n A

SD

Q D

isco

mfo

rtS

core

(/1

0)

2

* Significant difference (p< 0.05)

Fig. 2. Plot of means�seat position. A significant difference was found for

subject response scores on the ASDQ between seat positions #4 and #5

using a mixed repeated measure ANOVA. All other seat position

interactions were not found to be significant.

D.R. Smith et al. / International Journal of Industrial Ergonomics 36 (2006) 141–149 145

coincide with the predetermined seat positioning for eachsession as well as permitting subjects the opportunity towalk around and/or stretch before the next session.Subjects re-entered and repeated these procedures untilall sessions were completed for that day. Once all 3 dayswere completed, subjects were verbally debriefed about thepurpose of the study and thanked for their participation.

2.2.3. Statistical analysis

Visual analog scale scores from the ASDQ weremeasured and recorded. Answers from the AutomobileSeat Comfort Survey (Kolich, 2000) were also recordedusing the established Likert scale values from �3 to 3.

Factor analysis and sub-scale calculations were estab-lished using the same methods as pilot test #3. ASDQidentified sub-scales and sub-scale repeatability (Cron-bach’s a value) were compared to pilot #3 data as well asthe double null seat position (seat C).

A 2 (gender)� 3 (day)� 5 (seat position) mixed repeatedmeasures analysis of variance (ANOVA) was performed onASDQ discomfort scores. Tukey HSD post hoc analyseswere performed to identify any significant differences(Statistica v. 5.0, StatsSoft, Tulsa, OK). Alpha was set at0.05 for all comparisons.

The scales used by the ASDQ and the AutomotiveSeating Comfort Survey (Kolich, 2000) were not equiva-lent. In order to directly compare scores between ques-tionnaires, ASDQ VAS scores had to be converted into anabsolute using the Likert scale values between �3 and 3,similar to that described in Kolich (2000).

Variables common to both questionnaires were identi-fied. A paired two-sample t-test for each response score wasused to note significant differences between questionnairescores.

3. Results

3.1. Development of the ASDQ

3.1.1. Key informant interviews

Key informant interview sessions resulted in a decreaseof source items from 30 to 26 as well as numerous wordingchanges (Table 1). The resultant Flesch–Kincaid read-ability score was 7.2, thus a proper level of readability wasobtained. Face validity requirements, as per Streiner andNorman (2003), were achieved.

3.1.2. Pilot testing

A lack of correlational significance between key sourceitems (po0:2) resulted in the deletion of 6 and therewording of 3 items from the ASDQ. Both modificationsresulted in increased questionnaire readability and statis-tical significance. Twenty source items found to contributeto the construct of seating discomfort were ultimatelyidentified (Table 1).

The final ASDQ Flesch–Kincaid value was 7.0. A factoranalysis exposed 3 distinct factors, providing further

support for included source items. Cronbach’s a was above0.8 for the 3 factors labeled as Trim, Cushion, and Lumbarsub-scales.

3.2. Establishing the ASDQ—experimental

Factor analysis identified 4 significant levels in experi-mental data. An analysis of significant componentsidentified these variable sub-scales as Lumbar, Trim,Backrest, and Cushion. Varimax rotated component valueswere used to assign each variable and its associatedquestion to a sub-scale (Fig. 1). Notably, variable labelscushion bolster (q.7) and backrest contour (q.14) weregrouped within two sub-scales. Each variable expressed ahigh rotation value for both stated factors as well as havinga high correlation with all variables contained within thatsub-scale.A significant difference between ASDQ discomfort

ratings for seat positions 4 and 5 was found (p ¼ 0:043)(Fig. 2). All other main effects and interactions were notstatistically significant. This supports that the ASDQmeasures the construct of discomfort without bias and isresponsive to subjective differences in perceived discomfortassociated with physical components of the seat.

ARTICLE IN PRESS

***** *

0.0

0.5

1.0

1.5

2.0

2.5

3.0

A-17

B-LM

BRE-1

1F-1

3G-1

2H-8 J-

5K-4 L-

7

M-6

Equivalent Questions

Res

pons

e S

core Automotive Seating

Comfort Survey-Kolich (2000)

ASDQ - RelativeScore

ASDQ - VAS Score

* Significant difference (p < 0.05) between the Automotive Seating ComfortSurvey and the ASDQ-Relative score using a t-Test: Paired Two Sample for Means.

Fig. 3. Between questionnaire response comparison. ASDQ relative

scoring is defined as ASDQ VAS scores converted into an absolute Likert

scale score between 0 and 3 (as seen in Kolich, 2000) by dividing each score

by 3.333. The resultant values were binned using a conversion protocol

between 0 and 3 (i.e. 0�0.74 ¼ 0, 0.75�1.49 ¼ 1).

D.R. Smith et al. / International Journal of Industrial Ergonomics 36 (2006) 141–149146

A Pearson correlation of 0.71 was found to be significant(p ¼ 0:05) for null seat positions 3 and 6. Thus, subjectresponses to the null seat position were similar throughoutall trials. This supports that individual subjective percep-tions of discomfort remain consistent over time and thatthe VAS contains a significant level of reliability.

3.3. Between-questionnaire comparisons

Of the 13 questions addressed in the Automotive SeatingComfort Survey, 10 were comprised of similar variables/sub-scales found in the ASDQ (Table 3). A high betweenquestionnaires subject response correlation was found(R2 ¼ 0:715). Between questionnaire values for sub-scalereliability addressing labels ‘Cushion’, ‘Backrest’, and‘Entire seat’ were significant due to correlation coefficientsof 0.64, 0.60, and 0.69, respectfully. When compared toconverted ASDQ-relative scores (Fig. 3), AutomotiveSeating Comfort Survey scores were higher for 8 of the10 interactions and possessed an average mean differenceof 3.5% between questionnaire variable response scores.

4. Discussion

The ASDQ (Appendix A) was developed with acceptablelevels of readability, significant questions and wording, facevalidity, high levels of VAS repeatability, and reliable andinternally consistent sub-scales. The ASDQ was notinfluenced by gender, but was sensitive to changes withinthe physical components of the seat. Subject perceptions onthe ASDQ were shown to be consistent over time.

The aim to include key variables contributing to theconstruct of automotive seating discomfort was supported.Face validity was established in the interview and reviewprocess in accordance with Streiner and Norman (2003).Pilot tests #1 and #2 resulted in the inclusion of onlystatistically significant questions and wording in theASDQ. Readability levels were significant throughoutdevelopment due to optimal Flesch–Kincaid scores.

Table 3

Between questionnaire equivalent variable listing

Automotive Seating Comfort Survey ASDQ

A—lumbar support 17—lumbar support

B—lumbar comfort LMBR—lumbar sub-scale

E—back lateral support 11—backrest width

F—back lateral comfort 13—backrest bolsters

G—seat back feel/firmness 12—backrest firmness

H—ischial/buttock comfort 8—cushion center

J—cushion length 5—cushion length

K—thigh comfort 4—cushion width

L—cushion lateral comfort 7—cushion bolsters

M—cushion feel/firmness 6—cushion firmness

A total of 10 similar variables contained within the Automotive Seating

Comfort Survey (Kolich, 2000) and the ASDQ are listed with their

between questionnaire equivalent.

Ultimately, subjects responded that the ASDQ was easyto use and expressed little concern during response times,offering further support for the structure and layout of theASDQ.To ensure trial realism, experimental set-up was con-

trolled for internal and external factors affecting theoccupant. Mandatory physical movements forced subjectsto stay mobile during each session in accordance with Jungand Choe (1996). This considerably increased the realismof the study. The 20min seat duration provided a realistictrial duration and supports the conclusion of Reed andMassie (1996) that 20min is an acceptable duration forstatic automotive seating simulation. The absence of seatvibration from the experimental set-up allowed for thedirect identification of contributing physical seat compo-nents to occupant seating discomfort. Thus, the experi-mental set-up was considered appropriate for performingautomotive seating simulations.A significant difference was found between seat position

4 and 5. The differences between these positions offer anexplanation for the identified main effect. Seat position 4had minimal support where as seat position 5 maintainedmaximum support to the mid-region of the backrest. Formost individuals this would result in the lumbar supportsystem being placed in the thoracic region of the back,creating an opportunity to sense discomfort.No significant gender, day, gender vs. day, gender vs.

seat position, and/or day vs. seat position effects were seen,providing evidence that a VAS was an appropriate choiceto quantify the construct of discomfort. This was expecteddue to the high levels of scale reliability and responserepeatability between gender, days, and seat positioncontained within the ASDQ. Between-subject variability(standard deviation) suggests that subjects have differentperceptions of discomfort but are able to maintain thisperception over time. Also, mean subject response averages

ARTICLE IN PRESSD.R. Smith et al. / International Journal of Industrial Ergonomics 36 (2006) 141–149 147

were consistent throughout all analyses supporting thedefinition of discomfort to include ‘mild form of pain’ andas an appropriate measure for evaluating automotiveseating discomfort. A large body of literature in whichVAS scales were shown to be reliable and consistentsupports these findings (Giacomin and Quattrocolo, 1997;Streiner and Norman, 2003; Tijhuis et al., 2003; van derEijk et al., 2001).

Comparisons between the Automotive Seating ComfortSurvey and the ASDQ were positive. A relative scorecomparison between similar questionnaire variables re-sulted in an R2 ¼ 0:715. Two omitted Automotive SeatingComfort Survey variables (J, M) were shown to besignificant and used in all between questionnaire compar-isons. These findings support the use of a preliminaryversion of the Kolich (2000) Automotive Seating ComfortSurvey.

A factor analysis of experimental trial data identified‘Lumbar’, ‘Trim’, ‘Backrest’, and ‘Cushion’ as significantASDQ sub-scales. Sub-scale reliability (Pearson r40:8)was significant, confirming each sub-scale variable washomogeneous and allocated properly. However, a discre-pancy between ASDQ pilot test #3 and experimental trialdata was identified. Pilot test #3 identified 3 principalcomponents whereas experimental trial data identified 4.Sub-scales ‘Trim’ and ‘Cushion’ had similar results,however, sub-scale ‘Lumbar’ differed dramatically betweenthe 2 data sets. Pilot test #3 data resulted in backrestrelated seat components to be grouped into the ‘Lumbar’sub-scale. Low sample size and an experienced populationin pilot test #3 directly affected exposing only 3 out of 4eventual sub-scales.

Comparing data sets from the Automotive SeatingComfort Survey and pilot test #3 resulted in highcorrelation values and strong internal consistencies. Bothdata sets were also unable to show ‘Backrest’ and ‘Lumbar’sub-scales as distinct. However, when sample size wasincreased in the ASDQ and an inexperienced subjectpopulation was solicited, ‘Lumbar’ and ‘Backrest’ sub-scales were shown to be significantly separate and distinct.This supports that sample size and seat selection had animpact on resultant questionnaire content. The lack ofdivision between ‘Lumbar’ and ‘Backrest’ components andthe lack of a ‘Trim’ sub-scale greatly diminishes the detailpresented in questionnaire results.

A between questionnaire comparison resulted in similarmean responses and minimal percent mean differencesbetween ASDQ and Automotive Seating ComfortSurvey subject response data. A significant correlationwas found between comparable questionnaire sub-scales.These findings suggest that both questionnaires aremeasuring the same construct. However, differencesbetween 6 of the 10 compared source item variables werefound to be significant (po0:05). These differences arelikely due to sub-scale structure variation between ques-tionnaires. ASDQ VAS usage allowed subjects higher levelsof freedom when establishing their perceived level of

discomfort. The Automotive Seating Comfort Surveyforced subjects to select an anchored point, removingsubjective control and the freedom to provide an exactrepresentation of their perceived discomfort. Since sub-jective perception was pre-defined in the AutomotiveSeating Comfort Survey, when a subject felt a level ofdiscomfort between anchors, a forced decision wasprompted due to an intermediate value being unavailable.Supporting this claim, questionnaire response scoreswere higher on the Automotive Seating ComfortSurvey on 8 of 10 between questionnaire variablecomparisons. The ASDQ allowed subjects to respondaccording to their own discomfort perception definition ona continuous scale, where as the Automotive SeatingComfort Survey pre-defines this perception throughanchoring and forces the subject to respond in 1 of 7defined ways.It is being suggested that the physical spacing between

anchors in the Automotive Seating Comfort Surveyproduces a region of non-response that decreases the levelof precision with which a subject is able to evaluate theirperceived discomfort. This was seen to result in a lowerlevel of detail present, the difference in overall number ofvariables used (ASDQ ¼ 20, Automotive Seating ComfortSurvery ¼ 10), and overall questionnaire applicability andquality of results.

5. Conclusion

The ASDQ contains significant questions and wording,readability, face validity, VAS repeatability, and reliableand internally consistent sub-scales. The ASDQ is sensitiveto subjective perceptions of seating discomfort over time.The use of proper developmental methods resulted in astatistically significant, repeatable, reliable, and partiallyvalidated assessment tool for automotive seating discom-fort.A between-questionnaire comparison exposed differ-

ences between a previously established questionnaire andthe newly developed ASDQ. These differences supportmethodologies used and how the ASDQ is able to providemore detailed and comprehensive results than otherquestionnaires in the automotive seating industry.Therefore, it is concluded that the ASDQ is able to

measure the construct of automotive seating discomfort ina reliable and detailed way as well as provide informationon how areas of the seat contribute to overall occupantdiscomfort.

Acknowledgements

The authors would like to thank Schukra of NorthAmerica for their cooperation and funding throughout thisproject, Dr. Krista Chandler for statistical assistance.

ARTICLE IN PRESSD.R. Smith et al. / International Journal of Industrial Ergonomics 36 (2006) 141–149148

Appendix A. Automotive Seating Discomfort Questionnaire

(ASDQ) (not actual size)

For full questionnaire please contact the authors.

References

Blair, G.R., So, R., Milivojevich, A., van Heumen, D., 1998. Automotive

seating comfort: investigating the polyurethane Foam contribution-

phase 1. SAE Technical Paper 980656, Society of Automotive

Engineers, Inc., Warrendale, PA.

Crane, B.A., Holm, M.B., Hobson, D., Cooper, R.A., Reed, M.P.,

Stadelmeier, S., 2004. Development of a consumer-driven wheelchair

seating discomfort assessment tool (WcS-DAT). International Journal

of Rehabilitation Research 27, 85–90.

Dhingra, H.S., Tewari, V.K., Singh, S., 2003. Discomfort, pressure

distribution and safety in operator’s seat—a critical review. Agricul-

ture Engineering International 5, 1–16.

El Falou, W., Duchene, J., Grabisch, M., Hewson, D., Langeron, Y.,

Lino, F., 2003. Evaluation of driver discomfort during long-duration

car driving. Applied Ergonomics 34, 249–255.

Giacomin, J., Quattrocolo, S., 1997. An analysis of human comfort when

entering and exiting the rear seat of an automobile. Applied

Ergonomics 28 (5/6), 397–406.

Goonetilleke, R.S., Feizhou, S., 2001. A methodology to determine the

optimum seat depth. International Journal of Industrial Ergonomics

27, 207–217.

Gyi, D.E., Porter, J.M., 1999. Interface pressure and the prediction of car

seat discomfort. Applied Ergonomics 30, 99–107.

Jung, E.S., Choe, J., 1996. Human reach posture prediction based on

psychophysical discomfort. International Journal of Industrial Ergo-

nomics 18, 173–179.

Kolich, M., 2000. Ergonomic modeling and evaluation of automobile seat

comfort. Ph.D. Dissertation, Windsor, ON.

Kolich, M., 2003. Automobile seat comfort: occupant preference

vs. anthropometric accommodation. Applied Ergonomics 34,

177–184.

Kolich, M., 2004. Predicting automobile seat comfort using a neural

network. International Journal of Industrial Ergonomics 33, 285–293.

Kolich, M., Taboun, S.M., 2004. Ergonomics modeling and evaluation of

automobile seat comfort. Ergonomics 47 (8), 841–863.

Porter, J.M., Gyi, D.E., Tait, H.A., 2003. Interface pressure data and the

prediction of driver discomfort in road trials. Applied Ergonomics 34,

207–214.

Reed, M.P., Massie, D.L., 1996. Distribution of automobile trip durations

for studies of seat comfort. SAE Technical Paper 960476. Society of

Automotive Engineers, Inc., Warrendale, PA.

Reed, M.P., Saito, M., Kakishima, Y., Lee, N.S., Schneider, L.W., 1991.

An investigation of driver discomfort and related seat design factors in

extended-duration driving. SAE Technical Paper 910117. Society of

Automotive Engineers, Inc., Warrendale, PA.

Reed, M.P., Schneider, L.W., Ricci, L.L., 1994. Survey of auto seat design

recommendations for improved comfort. UMTRI Technical Report

94-6. The University of Michigan Transportation Research Institute,

Ann Arbor, MI.

Shen, W., Parsons, K.C., 1997. Validity and reliability of rating scales for

seated pressure discomfort. International Journal of Industrial

Ergonomics 20, 441–461.

Steinmetz, S. (Ed.), 1997. Random House Webster’s College Dictionary,

second ed. Random House, Inc., New York.

Streiner, D.L., Norman, G.R., 2003. Health Measurement Scales: A

Practical Guide to their Development and Use, third ed. Oxford

University Press, Oxford.

ARTICLE IN PRESSD.R. Smith et al. / International Journal of Industrial Ergonomics 36 (2006) 141–149 149

Tijhuis, G.J., Kooiman, K.G., Zwinderman, A.H., Hazes, J.M., Breed-

veld, F.C., Vlieland, T.P., 2003. Validation of a novel satisfaction

questionnaire for patients with rheumatoid arthritis receiving out-

patient clinical nurse specialist care, inpatient care, or day patient team

care. Arthritis and Rheumatism 49 (2), 193–199.

van der Eijk, I., Sixma, H., Smeets, T., Veloso, F.T., Odes, S., Montague,

S., et al., 2001. Quality of health care in inflammatory bowel

disease: development of a reliable questionnaire (QUOTE-IBD)

and first results. American Journal of Gastroenterology 96 (12),

3329–3336.