Means to measure the aesthetic properties of wood

L U L E A I U N I V E R S I T Y

O F T E C H N O L O G Y

2 0 0 0 : 2 6

D O C T O R A L THESIS

Means to Measure the Aesthetic

Properties of Wood

Means to Measure the Aesthetic

Properties of Wood

N . O l o f B r o m a n

Institutionen i S k e l l e f t e å / A v d e l n i n g e n för t r ä t e k n i k Luleå tekniska universitet

A k a d e m i s k a v h a n d l i n g

som m e d v e d e r b ö r l i g t tillstånd av Tekniska f a k u l t e t s n ä m n d e n v i d L u l e å tekniska universitet, f o r a v l ä g g a n d e av teknologie doktorsexamen i ä m n e t T r ä t e k n i k , k o m m e r att offentligt försvaras p å svenska, fredagen den 29 september, k l 10.00 i H ö r s a l e n , A-huset, S K E R I A , Skellefteå.

Handledare: Professor Anders G r ö n l u n d , Avdelningen för t r ä t e k n i k

Metoder att mäta estetiska egenskaper hos trä S A M M A N F A T T N I N G

Trä är ett material med estetiska egenskaper, som rätt utnyttjat kan ge fördelar framfor andra material. Kunskapen om kunders tycke och smak vad gäller träegenskaper för en given produktgrupp, där trä används synligt, är ofta bristfällig, och i de fall man vet vad kunderna föredrar vet man alltför sällan varför det är så.

Målen med arbetet har varit att utveckla metoder för att mäta människors preferenser för olika träutseenden samt att ange vilka träegenskaper som är viktiga. Syftet har också varit att visa på samband mellan människors preferenser och blandningen av

träegenskaper i en träyta. Enbart träslaget f u r u har studerats.

Resultat visar att folk föredrar olika utseenden på trä och att det är två kvalitativa skillnader som är av betydeke: den allmänna blandningen av träegenskaper i en yta och eventuell närvaro av avvikande egenskaper i en yta. Starkt avvikande egenskaper, t.ex. stor kvist, kvistansamling eller enstaka färgstråk, har störst betydelse (och oftast negativ). Finns inga avvikande egenskaper som stör upplevelsen av en träyta är det den allmänna blandningen av träegenskaper som avgör vad folk tycker.

Träytor som uppskattas skall väcka intresse och ha ett fräscht utseende. Naturligt nog ger kvistrena ytor ofta ett harmoniskt och elegant utseende. Intervjuer visar dock att sådana ytor b ö r ha så pass stor mängd variation i texturen, att de blir intressanta och spännande att se på. För ytor med kvist råder det motsatta. Frågor, huruvida ytan är i harmoni, lätt att se på eller har balans i texturen är därför viktigare än frågor rörande livlighet i texturen.

Kvantitativa studier visar att det går att mäta människors preferenser för trä med hjälp av en enkät bestående av frågor, där svaren anges på skalor med motsatsord på var sida. Övergripande frågor o m synintryck och detaljerade frågor om enskilda träegenskaper har testats. De övergripande frågorna ger en bättre förklaring (starkare modeller) av folks slutomdömen än de detaljerade frågorna. Tretton övergripande frågor har pekats ut som viktiga att använda vid framtida intervjuer.

Två metoder att visa trätexturer vid intervjuer har utvärderats. Den ena utgick från rena träytor och den andra datorbilder, där samma träytor applicerats på en exempelprodukt (köksluckor). Resultaten var likvärdiga och visar att datorbilder, som är mer praktiska att använda i en intervjusituation, är ett bra alternativ.

Samband mellan subjektiva preferenser och objektivt mätbara texturegenskaper mätta i digitala bilder har påvisats. Resultaten visar att det är ett mångsidigt problem och att det därigenom krävs många beskrivande mätningar av texturen för att beskriva folks tycke och smak. Det verkar vara enklare att hitta texturegenskaper som påverkar människors synintryck negativt än tvärt om. Bland de viktigaste variablerna var de som beskrev träegenskapers fördelning i en yta och speciellt de som mätte tyngdpunktsförskjutning orsakad av snedfördelning av egenskaper.

DOCTORAL THESIS

Means to Measure the Aesthetic

N . O l o f B r o m a n

Division o f W o o d Technology Skellefteå Campus Luleå University o f Technology Skeria 3, S-931 87 Skelleftå, Sweden

h t t p : / / w w w . t t . l u t h . s e / 2000

Means to Measure the Aesthetic Properties of Wood

A B S T R A C T

W o o d is a biological material w i t h inherent aesthetic properties w h i c h can give the final product a competitive advantage over other materials. I n this thesis, people's feelings and preferences f o r Scots pine (Pinus sylvestris) w o o d surfaces have been examined. T h e main objectives o f this thesis are to develop methods f o r measuring people's preferences toward different looks o f w o o d and to connect the subjective preference data w i t h objective measurements o f w o o d features. I t comprises b o t h qualitative and quantitative methods f o r collecting preference data. T w o projection methods, P C A and PLS, were used to analyze the quantitative studies.

Interviews made clear that people prefer different blends o f w o o d features. There are t w o qualitative differences that are o f importance f o r people's impressions and valuations o f w o o d : T h e overall blend o f w o o d features and

divergent features that mismatch i n a surface. I t was f o u n d that divergent features

are more important than the overall mixture o f features. B u t i f there are no defects that mismatch, the overall mixture w i l l then be the key to a person's appreciation o f a w o o d surface. W o o d surfaces should stimulate people's interest and be fresh l o o k i n g . A clear surface is naturally rather harmonious, elegant and easy to l o o k at. O n the other hand, a clear surface should be stimulating to l o o k at, should be exciting and i t should n o t look like an imitation. K n o t t y surfaces usually are less harmonious. Therefore, questions about harmony, easiness to l o o k at and balance are o f importance. Just as f o r clear surfaces, a k n o t t y surface should also stimulate people's interest, have a fresh look, be exciting and stimulating to look at.

The quantitative studies show that i t is possible to measure people's

preferences t o w a r d w o o d by a questionnaire technique. T h e results also show that there are differences i n people's judgments f o r surfaces with knots and f o r those without knots. There are stronger connections between questions regarding the overall impression and the final assessment than there are f o r detailed questions regarding certain w o o d features. T h i r t e e n questions were pointed out as relevant to use w i t h interviews. T w o interview methods were compared. T h e first m e t h o d used w o o d surfaces and the second used computer images w i t h the same w o o d surfaces applied to an example product, a kitchen cabinet. T h e results show that the t w o methods are comparable and useful. It was shown that i t is possible to some extent to connect subjective preference data w i t h objective w o o d feature measurements. Results show that a l o t o f w o o d feature measurements are needed and that i t is a multivariate problem.

w o o d surface, and especially those variables that detect a deviation i n center o f gravity.

Increased knowledge about people's preferences f o r the aesthetic properties o f w o o d w i l l lead to a better understanding o f w h i c h w o o d features should be measured and controlled i n the f u t u r e .

Keywords: w o o d texture, Scots pine, aesthetic properties, interviewing

methods, questionnaire, preference, prediction models, P C A , PLS, image analysis.

Means to Measure the Aesthetic Properties of Wood

P R E F A C E

This w o r k was carried out at the D i v i s i o n o f W o o d Technology, Luleå University o f T e c h n o l o g y , Skellefteå Campus, under the supervision o f Professor Anders G r ö n l u n d , to w h o m I owe sincere thanks. Financial support f r o m the Swedish National Board f o r Industrial and Technical Development made the w o r k possible and is gratefully acknowledged.

I w o u l d also like to thank all colleagues w h o have helped and inspired me over the years. Special thanks to Olle Hagman f o r broadening m y multivariate t h i n k i n g and to Johan Oja, f o r f r a i t f u l discussions and theoretical

development. Thanks also to Brian Reedy and m y father w h o have corrected m y English.

Finally, I w o u l d also like to express m y sincere gratitude to m y family w h o still loves me.

SkeUefteå, July 2000

L I S T O F P A P E R S

This thesis is based on w o r k reported i n the following six papers, referred to by roman numerals:

I Broman, N . O . , 1995. Visual Impressions o f features i n Scots pine w o o d surfaces: A quaHtative study. Forest Products Journal 45(3), 61-66.

I I Broman, N . O . , 1995. Attitudes toward Scots pine w o o d surfaces: A multivariate approach. M o k u z a i Gakkaishi 41(11), 994-1005.

I I I Broman, N . O . , 1996. T w o methods f o r measuring people's preferences for Scots pine w o o d surfaces: A comparative multivariate approach. M o k u z a i Gakkaishi 42(2), 130-139.

I V Broman, N . O . , 2000. Aesthetic properties w i t h i n k n o t t y w o o d surfaces and their connection w i t h people's preferences. Accepted f o r publication i n Journal o f W o o d Science.

V Broman, N . O . , G r ö n l u n d , A . , 1996. T h e end-users' requirements for the aesthetical features o f Scots pine w o o d . I n : Proceedings o f the second I U F R O W o r k s h o p W P S5. 01-04. South Africa, pp. 343-352.

V I Broman, N . O . , 2000. T h e measurement o f w o o d features i n knotty Scots pine w o o d surfaces and the connection w i t h people's preferences. Presented at the 4t h International Conference o n Image Processing and

Means to Measure the Aesthetic Properties of Wood C O N T E N T S A B S T R A C T i P R E F A C E i i i L I S T O F P A P E R S iv C O N T E N T S v 1. I N T R O D U C T I O N 1 1.1 Quality approach 2 1.2 Wood communication 3 1.3 The objectives of this thesis 3 1.4 Outline of the thesis 4

1.5 Limitations 4

2. M A T E R I A L S A N D M E T H O D S 6

2. 1 The starting point: a qualitative study 6 2.2 Screening of questions in a quantitative study 7 2.3 Comparing two methods for measuring people's preferences for wood 9

2.4 Aesthetic properties within wood surfaces 9 2.5 Connections between preference data and coarse manual measurements 9

2.6 Prediction qf people's preferences based on detailed wood feature

measurements 10

2.7 Projection methods 11 2.8 Discussion - Material and Methods 11

3. R E S U L T S A N D D I S C U S S I O N 13

3 A What people see and evaluate in wood 13 3.2 How to measure people's preferences toward wood 14

3.3 What questions to use? 15 3.4 Important questions for surfaces with knots 16

3.5 Important questions for clear surfaces 17 3.6 Detailed questions about specific wood features 20

3.7 People's preferences versus objective wood feature measurements 20

3.8 Similarities with Preference mapping 21

3.9 Limitations 21

4. C O N C L U S I O N S 23 5. F U T U R E W O R K 24 6. R E F E R E N C E S 25

1. I N T R O D U C T I O N

W o o d is a biological material w i t h inherent aesthetic properties w h i c h can give the final product a competitive advantage over other materials. I t is therefore important f o r w o o d manufacturing industries to take advantage o f those features o f w o o d that have great influence o n customers' choice o f product. Formerly, skilful craftsmen handled the aesthetic values o f w o o d . Today, w o o d products are produced at high speed and by modern technology and n e w methods. T h e risk o f losing valuable aesthetic features during the process is therefore great.

It is the final customer's willingness to buy a product that is the motive p o w e r f o r the w h o l e production process. Customer satisfaction depends u p o n the product's performance relative to a buyer's expectations, and high customer satisfaction creates high customer loyalty (Kotler et al. 1996). This applies to the production o f w o o d products also. I n a w o o d product chain, i t is c o m m o n that knowledge o f the preferences o f the final customers is poor. This w i l l often lead to suboptimization o f the utilization o f w o o d i n the production process. There are several reasons f o r this. O n e is the lack o f a j o i n t language throughout the w h o l e production chain, w h i c h w o u l d enable a correct

description o f the quahty aspects i n a product. A n o t h e r very important factor is that manufacturers often fail to translate the features wanted i n t o objectively measurable parameters. The question becomes h o w to investigate and measure people's preferences f o r different looks o f w o o d , and hence, h o w to

communicate them.

Every plank or board has its individual features causing possibilities and problems i n the w o o d production chain. For products w i t h visible w o o d , i t w o u l d be an advantage i f there were a way to communicate the aesthetic properties o f w o o d desired. Today, there is no such communication t o o l b o x i n use w i t h i n the market f o r w o o d products. Ordinary market studies applied to w o o d products o f t e n reveal people's preferences, i.e. what they prefer, but not the reason w h y they prefer a certain product. I t has been observed that although consumers are clear w h i c h products they like and dislike, they are not always able to describe specifically w h y they Uke or dislike a product. A n y description given by consumers also tends to be o f a hedonic nature and not descriptive. Knowledge o f h o w to transform or communicate i n f o r m a t i o n about a preferred product, especially its w o o d content, is o f t e n poor. I n a preference study f o r Pacific Northwest hardwoods, Swearingen et al. (1998) conclude that buyer/customer discrepancies may be due to the inability o f buyers to recognize some o f the w o o d attributes that their customers actually prefer.

Means to Measure the Aesthetic Properties oj Wood

I n w o o d processing i t is c o m m o n to talk about quality or grade, where quality means properties and w o o d features, and grade the level o f quality. There are many different wood-grading systems, b o t h specific systems f o r individual countries and internationally adopted systems. M o s t o f the appearance-grading systems are used f o r grading o f commodities, b u t there are also systems that are end-product specific. C o m m o n f o r all appearance-grading systems is, however, that they all to a large extent only reflect the producer perspective. End-user preferences are not taken i n t o account, primarily because they are not k n o w n . " I n sawmills, grading rules originally installed to ensure a certain level o f quality m i g h t actually Hmit quality. Some companies may try just to meet m i n i m u m specifications, and do not reaHze the specific quality expectations o f their customers" (Hansen and W e i n f u r t e r 1999). T h e grades often do not refer to customers' specific needs (Hansen and Bush 1996).

Apart f r o m the w o o d processing industry's usage o f quality, this thesis relies o n a standard quality definition: "Quality is the totaHty o f characteristics o f an entity that bear o n its ability to satisfy stated or i m p l i e d needs" (International Standard I S O 8402). T h e appearance o f the w o o d texture i n a product w i t h visible w o o d ought to affect people's propensity to buy i t , and hence affect the experience o f quahty.

1.1 Quality approach

A l o t o f effort around the w o r l d has been expended to define the concept o f quality. A m o n g the five approaches to the definition o f quaHty discussed by Garvin (1984), t w o o f them coincide w i t h the aims o f this thesis. These are (in summary):

Transcendent approach. " I k n o w i t w h e n I see i t " is a criterion o f the

transcendent quality. W h a t is beautiful must be j u d g e d and answered by the customer.

User-based approach. This is the approach that is closest to the n o w

most-used d e f i n i t i o n (ISO 8402: "ability to satisfy . . . needs") and indicates the customer-oriented attitude. I t is measurable, b u t y o u must ask the customer and use the answers as input to quality efforts.

Garvin (1984, 1988) has also surveyed eight different aspects o f quality: performance, features, reHability, conformance, durability, serviceability, aesthetics, and perceived quality. T h e importance o f a particular aspect o f a quality changes w i t h the nature o f the product and the needs o f the customer. Garvin claims that companies compete i n these eight dimensions o f quaHty,

Pakarinen (1999) investigated the success factors o f w o o d as a furniture material, and he showed that the f o u r most often mentioned attributes were rehable, environmentally friendly, g o o d - l o o k i n g and valued. Furthermore, Ozanne and Smith (1996) investigated h o w noneconomic b u y i n g criteria, such as the environment, affected customers and their purchase decision. B o t h these investigations i m p l y that by segmenting the market and by applying more customer-oriented product development and manufacturing, w o o d as an engineering material w i l l exploit a number o f competitive factors and achieve success i n the marketplace. I n a study o f the appreciation o f oak w o o d , Marchal and M o t h e (1994) claim that w o o d w i t h intermediate quality is not marketable because o f the lack o f genuine knowledge o f people's taste. I n detail, they conclude that all kinds o f tastes coexist, b u t there are different groups o f preference profiles. Marchal and M o t h e also show that the overall criteria that affect people's choice are knottiness, cut orientation, tint, and annual ring w i d t h .

1.2 Wood communication

T h e wish f o r a t o o l b o x f o r communicating the aesthetic properties o f w o o d w h i c h w i l l w o r k i n all situations at all times is o f course a b i t infantile. Nevertheless, its content reflects the need f o r i m p r o v e d market information and w o u l d be o f interest f o r the w o o d processing industry and f o r the

customers. Some components o f such a communication t o o l b o x that could be mentioned are:

- Methods f o r measuring people's preferences f o r the aesthetic properties o f w o o d .

- Methods f o r transforming subjective preferences i n t o objectively measurable parameters, i.e. desired aesthetic quality.

- Methods f o r communicating the desired aesthetic quality throughout the whole w o o d production chain.

Research regarding aesthetic features o f w o o d and people's preferences f o r different looks o f w o o d has so far been very limited. A l t h o u g h some m i n o r studies have been carried out i n the U S , Europe and Japan, the lack o f knowledge i n these fields is evident. I t is also n o t quite clear what methodologies should be applied i n this research area.

1.3 The objectives qf this thesis

This thesis is devoted to suggesting methods f o r measuring people's preferences toward different looks o f w o o d . Hence, the main objectives o f the w o r k are to develop methods f o r measuring people's preferences t o w a r d different looks o f w o o d and to connect the subjective preference data w i t h objective

Means to Measure the Aesthetic Properties qf Wood

measurements o f w o o d features. Increased knowledge about people's preferences f o r w o o d can be implemented i n research aimed at translating people's subjective preferences to objective parameters measured i n a w o o d specimen and can also be implemented i n market studies.

T h e objective o f Paper / was to describe and create understanding o f people's feelings and attitudes toward w o o d i n order to develop quantitative studies i n the future.

I n Paper II, the objective was to f i n d relevant questions to ask w h e n measuring people's preferences f o r different looks o f w o o d . T h e aim was also to reveal the relationship between the mixture o f features i n a w o o d surface and people's attitudes. A questionnaire specially developed f o r this purpose was tested. T h e objective of Paper III was to compare t w o interviewing methods and to f i n d relevant questions to ask w h e n measuring people's preferences f o r different looks o f w o o d .

I n Paper IV, the aim was to shed light o n h o w people's preferences toward different w o o d appearances containing knots can be described and measured. T h e objectives of Papers V and FTwere to see i f it was possible to connect people's subjective attitudes w i t h objective measurements made o n k n o t t y w o o d surfaces, as w e l l as to p o i n t out w h i c h texture properties affect people's l i k i n g , i.e. what is important to measure and communicate.

1.4 Outline of the thesis

T h e thesis is based on the papers outlined i n Figure 1 - showing the evolutionary progress o f this w o r k . A l l papers concern the issue o f h o w to communicate the aesthetic features o f w o o d f r o m the final customer back to the w o o d processing industry. T h e first f o u r papers concern h o w people's preferences toward w o o d can be described and methods f o r investigating them. A n important task was to point out relevant questions to ask. T h e last t w o papers investigate the possibilities o f connecting subjective preference data and objective wood-feature measurements. Those were based o n the results o f surfaces w i t h knots. As a complement, the analysis made on k n o t t y surfaces i n Paper I V was also done f o r clear surfaces and is presented i n this thesis.

1.5 Limitations

T h e results presented i n this thesis are all based o n interviews done i n Sweden. T h e most serious limitation was the small number o f w o o d surfaces used. Also,

i n t e r v i e w i n g techniques applied to other w o o d species, products and people. T h e w o r k presented here should be seen as an i n t r o d u c t i o n to a research area n o t yet f u l l y exploited.

Market preference for visible wood

Preference measurements

I . W h a t people see i n w o o d and h o w they express i t

£1

I I . Testing possible questions and a questionnaire

I I I . Comparing the usage o f w o o d surfaces and digital images when interviewing people

3

Communication

I V . A description o f two dimensions o f the phenomenon attitudes toward w o o d and what questions to use.

r ; \

V . Relation between preference data and simple manual measurements made on w o o d surfaces

Connections between Preferences and the blend O f w o o d features

1

Industrial production of visible wood

Means to Measure the Aesthetic Properties of Wood

2. M A T E R I A L S A N D M E T H O D S

I n the research process the six papers presented i n this thesis are closely related to each other. Therefore, a brief description o f the governing idea among the papers is given.

T h e first investigation (Paper I), a qualitative study o f what people see and evaluate i n a w o o d surface, built a platform f o r further studies.

W i t h results f r o m the first investigation i n hand i t was possible to develop a relevant questionnaire w h i c h was tested i n the second investigation (Paper II). This quantitative study sorted o u t the most important questions f r o m those o f m i n o r importance. I n order to separate the appearance o f a w o o d surface f r o m product influences, no product example was given.

I n Paper III an i m p r o v e d questionnaire was used and people were asked t o assess the l o o k o f different w o o d structures applied to an example product, a kitchen cabinet. I n the same study t w o interviewing methods were compared. A l l three o f the first investigations were done w i t h different people and at different places i n Sweden.

T h e f o u r t h study (Paper IV) used the interview results f r o m Paper I I f o r a descriptive analysis o f the phenomenon Attitudes T o w a r d W o o d , A T W . T h e important decision o f w h i c h questions to use i n further studies was made.

The connections between preference data and coarse manual measurements o f w o o d features were then examined i n Paper V. T h e interview data was taken f r o m the second study and the results f r o m detailed questions concerning certain w o o d features served as guidelines o f what to measure i n a w o o d surface.

Finally, the sixth study (Paper VT) describes the connections between people's preferences and detailed w o o d feature measurements extracted f r o m digital images. I t was based o n the interview data f r o m Paper I I , the questions pointed out i n Paper I V and indications o f what to measure f r o m Paper V .

2. J The starting point: a qualitative study

T o detect people's visual impressions o f w o o d surfaces made o f Scots pine, interviews were made i n a carpentry shop using 27 different surfaces o f sohd w o o d panel sized approximately 0.4 x 1.5 meters. W o o d surfaces were made

theoretical foundation i n Grounded T h e o r y (Glaser and Strauss 1967). T h e theory and the method used to summarize the data are described i n Paper I.

2.2 Screening of questions in a quantitative study

W i t h the results f r o m Paper / as a base, a questionnaire was developed and tested at the Skellefteå W o o d Festival i n Sweden. The questionnaire contained 54 questions about people's visual impressions and detailed questions about h o w people felt about different features i n a w o o d surface. For practical reasons, the number o f t e s t surfaces was decreased. Thus, 16 o f the 27 w o o d surfaces used i n the first investigation were used f o r this study, and they are shown i n Figures 2 and 3. The data were analyzed by three multivariate

techniques. A correlation analysis was carried out to detect simple relationships between questions and w o o d features, and a Principal C o m p o n e n t Analysis (PCA) was done to v i e w the more complex relations between variables (questions). Finally, to sort out important questions, a Partial Least Square (PLS) analysis was made. The correlation analysis and P C A were computed b y a program, J M P (Anon. 1994), and the PLS study by a program, S I M C A ( A n o n . 1998). I n Paper lithe materials and methods are described i n detail.

2.3 Comparing two methods for measuring people's preferences for wood

I n the t h i r d study t w o methods f o r measuring people's preferences were compared and the interviews t o o k place at a f u r n i t u r e exhibition i n Älvsjö, Sweden. A reduced questionnaire was used f o r the t w o methods. T h e first m e t h o d used w o o d surfaces, and the second m e t h o d used computer images w h e n people were interviewed about their preferences f o r different looks o f w o o d applied to an example product, a kitchen cabinet. T h e w o o d surfaces were identical w i t h those i n Paper I I . T h e computer images showed a kitchen cabinet w i t h the same w o o d surfaces used i n m e t h o d 1. T o create the images f o r the second method, a program, Adobe Photoshop, was used. T h e same program was used during the interviews. T h e statistical tools were the same as f o r Paper I I presented above. T h e methods are f u r t h e r described i n Paper III.

2.4 Aesthetic properties within wood surfaces

I n Paper IV the variations among the interview data f r o m Paper I I were again examined by a P C A to stress the importance o f using questions f r o m the first t w o dimensions o f the phenomenon A T W . O n l y surfaces w i t h knots were examined. O n e reason f o r not using clear surfaces was that results f r o m Papers I and I I showed that there are differences i n people's judgements f o r surfaces w i t h knots and f o r those w i t h o u t knots. A n o t h e r reason was that i t simplified the analysis. The first t w o principal components were used f o r a thorough description o f the variations among people's preferences. Questions (answers) showing great impact o n the P C A model were chosen and pointed out as important to use f o r further studies.

A similar analysis f o r clear w o o d surfaces has n o t been presented i n a paper. Therefore, as a complement, the results f o r clear surfaces are presented i n this summary (see Section 3.5 Important questions for clear surfaces).

2.5 Connections between preference data and coarse manual measurements

I n Paper V, the connections between interview data and coarse manual measurements o f w o o d features were studied. T h e measurements concerned only the knots. T h e number, size, type, shape and distribution o f knots were detected. T h e connections between the subjective preference data and the w o o d features were analyzed by correlation analysis and Partial Least Square analysis (PLS).

Means to Measure the Aesthetic Properties of Wood

2.6 Prediction of people's preferences based on detailed wood feature measurements

I n Paper VI the connections between preference data and detailed w o o d feature measurements were analyzed. F r o m digital images o f ten w o o d surfaces containing knots, 140 w o o d feature variables were extracted. The interview data and the w o o d surfaces were taken f r o m Paper I I . Results f r o m detailed questions regarding the appearance o f certain w o o d features investigated i n Paper I I were used as guidelines as to what to measure. Also, the output f r o m Paper V was used, showing possible connections between preference data and objective wood-feature measurements. N o t e that features f o r the clear w o o d areas were also detected. O n l y preference data f r o m those questions pointed out as relevant i n Paper I V were used. T h e overwhelming problem here was the very small number o f w o o d surfaces (10 pieces) together w i t h a large number o f measurements.

W i t h P C A there is a fundamental difference between a model's fit and its prediction ability (Eriksson et al. 1999). T h e problem w i t h goodness o f at (R2,

explained variation) is that w i t h sufficiently many free parameters i n a model,

R2 can be made arbitrarily close to the optimal value o f one (1.0). M o r e

important is the fact that the prediction ability ( Q2, predicted variation

estimated by cross validation) is not inflationary and w i l l not automatically come close to 1 w i t h increasing model complexity. Figure 4 shows the trade-o f f between gtrade-otrade-odness trade-o f fit, R2, and goodness o f prediction, Q2. I n Paper VI,

b o t h R2 and Q2 were used to evaluate the performance o f the models.

T o secure robust models based o n causality, an intermediate step i n the modeling procedure was conducted. Pre-models were established f o r all y-variables (answers) where the robustness o f the linear coefficients w i t h i n each model was calculated by a Jackknife procedure ( W o l d 1982, W o l d et al. 1984, E f r o n and G o n g 1983). Final models were established by ordinary PLS

technique were the 30 most robust x-variables (unique f o r each model) were used. T o further reduce the risk o f modeling noise, only the first PLS

component was calculated. T h e methods are f u r t h e r described i n Paper VI.

2.7 Projection methods

T h e research w o r k presented i n this thesis is to a high degree based o n the use o f multivariate projection methods. P C A and PLS are projection methods that are w e l l suited f o r consumer science oriented investigations (Eriksson et al. 1999). Those methods are also described i n Papers II—VI. I n this w o r k , projection methods have been used f o r summarizing and visualizing data and f o r f i n d i n g quantitative relationships among variables. W h e n many consumers are asked f o r their preferences i n regard to a product or a series o f products, i t often results i n large and multivariate data tables. Also, preference data often contain bias or noise and the data is often collinear. There are several motives f o r using projection methods f o r the analysis o f such data. T h e y cope w i t h multicolhnearity b y b u i l d i n g models that capture this phenomenon. P C A and PLS are robust against noise. The noise is separated f r o m regularities among the variations i n the data. Projection methods also handle matrices that are short and fat, and they cope w i t h missing data. A l l preference data analyzed i n

Papers II-VI were scaled to unit variance. T h e reason f o r this was that the

interviewees were not trained i n assessing w o o d textures by a questionnaire. O n e can assume that the questions used were evaluated different b y different people, despite u n i f o r m scales f o r every question. This is also recommended b y R i s v i k et al. (1997).

2.8 Discussion - Material and Methods

This thesis comprises b o t h quahtative and quantitative methods f o r collecting preference data. Qualitative investigations are often f r u i t f u l , especially w h e n knowledge about the problem under study is scarce. I n this case, the

qualitative interviews i n Paper I gave good insight into what and h o w people see the texture properties i n w o o d surfaces. I t also gave a lot o f ideas about what to measure i n attempting to detect people's preferences f o r different looks o f w o o d . Quahtative interviews, though, are often time-consuming and do not show the relative importance among the results, i.e. they are hard to quantify. T h e quantitative interviewing methods used i n Papers II and III were very similar. A questionnaire was developed o n the basis o f the qualitative

Means to Measure the Aesthetic Properties of Wood

interviews. The questionnaire w o r k e d well, and therefore no alternative technique was tested. This is one o f several loose ends w i t h i n this w o r k . The issue o f h o w to select the interviewed persons is important f o r the

generalization o f results i n ordinary market preference studies. D u e to the aims o f the w o r k presented here, the only criterion f o r the selection o f interviewed persons was that they were eager to participate. Paper I was different, as the interviewed persons were selected by a m a x i m u m variation m e t h o d (Patton 1980).

3. R E S U L T S A N D D I S C U S S I O N

A l l six studies concern interview results collected f r o m potential consumers o f visible w o o d , and the interviewees were n o t trained f o r the purpose o f the investigations. T h e results shed light on what consumers see and evaluate regarding different looks o f w o o d and o n h o w to measure and communicate this.

3.1 What people see and evaluate in wood

I n Paper I a qualitative study o f what people see and evaluate i n a Scots pine w o o d surface is presented. The result is a map showing what features people see i n w o o d and a description o f properties and aspects o f the phenomenon A T W (People's Attitudes Towards W o o d ) .

Interviews made clear that people prefer different blends o f w o o d features. There are t w o quahtative differences that are o f importance f o r people's impressions and valuations o f w o o d . First, there is the overall blend o f w o o d features and second, the presence or absence o f divergent features that

mismatch i n the surface. T h e latter factor can be divided into t w o sub factors: - Distinguishing natural defects.

- Technical defects made by machine or the carpenter.

I t was f o u n d that divergent features are more important than the overall mixture o f features. B u t i f there are no defects that mismatch, the overall mixture w i l l then be the key to a person's appreciation o f a w o o d surface.

T o comment o n people's evaluations o f w o o d surfaces shown i n Papers II and

III, i t seems that w o o d surfaces should stimulate people's interest and be fresh

l o o k i n g . A clear surface is naturally rather harmonious, elegant and easy to l o o k at. O n the other hand, a clear surface should be stimulating to l o o k at, should be exciting and i t should not l o o k like an imitation. K n o t t y surfaces usually are less harmonious. Therefore, questions about harmony, easiness to l o o k at and balance are o f importance. Just as f o r clear surfaces, a k n o t t y surface should also stimulate people's interest, have a fresh l o o k and be exciting and stimulating to look at.

Using images o f w o o d and stones, Nakamura et al. (1994) also studied h o w visual patterns and color variations affect people's impressions. T h e y suggest that pattern anisotropy is one o f the most important visual factors w h i c h influence the psychological images o f " w o o d l o o k i n g " . Those results support the findings i n Papers II and III. Pattern anisotropy, f r o m m y p o i n t o f view, may create an exciting surface w h i c h stimulates people's interest. Pattern anisotropy should also affect the balance i n a surface. " T o o m u c h " o f some

Means to Measure the Aesthetic Properties qf Wood

w o o d feature, f o r instance color variation, number o f knots etc., may cause imbalance i n a w o o d surface. Nakamura et al. (1993) drew similar conclusions w h e n they studied preferences f o r w o o d w a l l panels. T h e y showed that the smaller the k n o t ratio, the more "agreeable" a wall panel becomes. (The k n o t ratio was an area fraction between knots and panel).

I n Paper IV, the multivariate analysis o f the interview answers f r o m a large number o f questions shows that a balance between t w o main properties affects people's l i k i n g : the degree o f harmony and activity. T h e state o f disharmony is strongly connected w i t h l o w preference among interviewed people. M o r e about this i n Section 3.4 Important questions for surfaces with knots. The results f r o m Papers II and III c o n f i r m these findings.

3.2 How to measure people's preferences toward wood

I n Paper II the results f r o m the first paper were used to create a questionnaire f o r measuring people's preferences f o r different looks o f w o o d . T h e study shows that i t is possible to measure people's feelings f o r different looks o f w o o d surfaces quantitatively by questionnaire technique. T h e results show that there are differences i n people's judgements f o r surfaces w i t h knots and w i t h o u t knots.

T h e same k i n d o f questionnaire, but w i t h fewer questions, was also used i n

Paper III, where t w o methods f o r measuring people's preferences f o r w o o d

were compared. People were asked to assess different looks o f w o o d applied to an example product, a kitchen cabinet. M e t h o d 1 used w o o d surfaces and method 2 used computer images o f the same w o o d surfaces used i n method 1 but showing the structures i n an authentic environment. T h e results show that the t w o methods used are comparable and useful. Papers II and III also show the same pattern regarding the importance o f the questions, and hence i t can be concluded that the questionnaire technique is valid and useful. Comparing the t w o methods, w o o d surfaces and computer images, the model f o r w o o d surfaces was slightly stronger, but there were no major differences. M y impression f r o m the study at the furniture exhibition is that b o t h methods have their advantages and disadvantages. W h e n using w o o d surfaces, the interviewed person can really see the w o o d , b u t i t can be hard to imagine the look o f the final product. W h e n using computer images to interview people, the situation is the opposite. Here the person can see the look o f the w o o d i n a product, b u t he or she can not see the w o o d features i n detail. I t was concluded that method 2, computer images, is more practical to use than method 1, w o o d surfaces.

techniques and questions used cover the most c o m m o n properties and aspects o f people's attitudes towards w o o d .

I n analyzing people's backgrounds (Paper III) there was no relation f o u n d between age and people's preferences f o r different looks o f w o o d . There was also no relation between a person's experience w i t h w o o d and what he or she prefers. T h e amount o f i n f o r m a t i o n about the interviewees was limited, though. T h e results also show that people interviewed at the furniture exhibition have higher preference f o r knotty w o o d surfaces than f o r clear w o o d surfaces.

3.3 What questions to use?

I n Paper II a large number o f questions were tested. There were t w o kinds o f questions: questions that detected the overall impression and questions that detected preferences f o r w o o d features i n detail. T h e analysis o f Paper II showed stronger connection between questions regarding the overall

impression and the final assessment than f o r the detailed questions. Therefore the main track has been to analyze the questions that encompass the overall impression.

T h e motives f o r searching f o r the most important questions and thus reducing the amount o f questions used i n an interview situation are t w o f o l d . First, i f a lot o f different w o o d appearances are to be investigated, it's not rational to have too many questions, and second, the interpretation w i l l be easier i f there are f e w but significant answers collected. A rather difficult task w i t h i n this w o r k has been to b u i l d a theory around what questions to ask. I n Figure 5 ( f r o m Paper IV), a loading plot visualizes the variation among the answers f o r k n o t t y w o o d surfaces. I n b o t h Papers II and III i t was concluded that questions w i t h close connection w i t h the final assessment were suitable to use. Thus, the suggestion was to use questions far f r o m the origin o f coordinates and located near the horizontal axis i n Figure 5. However, i n Paper IV a change is proposed as to w h i c h questions to use. T h e motive f o r this theoretical change is that i t seems wise to collect i n f o r m a t i o n about h o w people experience a w o o d surface, n o t merely w h y they like i t or dishke i t . T h e conclusion was that questions situated far f r o m the origin o f coordinates and w h i c h cover b o t h dimensions i n the loading plot should be selected (in Figure 5).

Means to Measure the Aesthetic Properties of Wood CL 0.30 0.20 0.10 0.00 -0.10 -0.20 • gaudy • restless crude disharmoniuos "a s v r l " hard tc. Loo!<hgavy " •cheapr t^ not fresh • eventful • lively • contrasty • * rich imaginative jncommon interesting •warm • -exquisite genuine |i k e i t ü l s l lbB^fcticiiftsäe •artificial, "fragile . cold • boring "unexciting uninteresting empty J i f e f l g n a •rigid -u n e ^ Mitce • solid beautiful

• symmetricar ^leg a^\auant

• expens^^rrarrnonious balanced • common clean ive • light " restful

easy to look at ventful

• strict

-0.20 -0.10 0.00 0.10 0.20

p[1]

Figure 5. K n o t t y w o o d surfaces. Visualization o f the answers f r o m all questions

and their variation described by a P C A model. T h e i r loading, i.e. their importance, is represented by the distance f r o m the origin o f coordinates. P [ l ] and P[2] mean the loading on the first and the second principal component, respectively.

3.4 Important questions for surfaces with knots

Thus, questions (answers) situated far f r o m the origin o f coordinates, and fairly w e l l distributed around the circular swarm i n Figure 5 were chosen w i t h great care. The variation among the selected questions/answers is shown i n Figure 6. Variables, i.e. answers, located far f r o m the origin and near the horizontal axis are strongly correlated w i t h people's l i k i n g . T h e answers are circularly

distributed, and the nearer the vertical axis, the more the variable describes the blend o f w o o d properties i n the w o o d surface indifferent to people's

preferences. I n more detail, the questions like it/dislike it, beautiful/ugly and

nice /objectionable detect the final assessment and the questions balance/unbalance, harmony / disharmony, interesting/uninteresting and stimulating/boring describe the

0.30 0.20 0.10 CT 0.00 -0.10 -0.20 -0.30 Aunev A rigid A indifferent A empty ^uninteresting A boring strict snfful A restful A A harmonious balanced nice *iike_it hpaiirifnl ^bjjectionable • dislike it * unbalanced Adisharmoniuos A restless Agaudy A interesting A stimulating . k rich ^ontrasty * lively * eventful • 1 1 ! 1 1 1 -0.20 -0.10 0.00 0.10 0.20 0.30 p[1]

Figure 6. K n o t t y w o o d surfaces. Visualization o f the answers f r o m 13 selected

questions and their variation described by a P C A model. T h e i r loading, i.e. their importance, is represented by the distance f r o m the origin o f coordinates. P [ l ] and P[2] mean the loading on the first and the second principal component, respectively.

I n an attempt to simplify and to show the general pattern i n the interview data f o r the ten k n o t t y w o o d surfaces, the raw interview data were summarized together i n three new variables, degree o f acceptance, harmony and activity. For k n o t t y w o o d surfaces, people's preferences were shown to be dependent o n a balance between the degree o f harmony and activity. T h e results also indicated the importance o f focus o n a sustained degree o f harmony w h e n composing w o o d surfaces w i t h knots.

3.5 Important questions for clear surfaces

N o corresponding analysis to Paper regarding clear surfaces has yet been presented. That is because o f the small number o f clear surfaces used (six pieces). However, as a complement to the analysis i n Paper IV, the same procedure f o r selecting important questions was used f o r clear surfaces. T h e interview data and w o o d surfaces originate from Paper II (the same as i n Paper / F a l s o ) . The variation among the answers f o r clear surfaces was analyzed by P C A . A corresponding loading plot to Figure 5 is shown i n Figure 7. T h e

Means to Measure the Aesthetic Properties qf Wood

answers are again circularly distributed but here they are mirrored compared to

Figure 5. 0.201 0.10 CN Q. 0.00 -0.10 -0.20 ' clean ' restful e a s y j o j o o k j a t symmetrica

.

. balanced

®

h i c f r l r ø i #

fresh tiarmoriiöBs

ri<r

'

hard

riph jmaginativj-e " d i r ty . hard to look at

l i v e l y ^ ™ i g au d yr e s t l e s s

-0.20 -0.10 0.00 0.10 0.20

P[1]

Figure 7. Clear w o o d surfaces. Visualization o f the answers f r o m all questions

and their variation described by a P C A model. T h e i r loading, i.e. their importance, is represented b y the distance f r o m the origin o f coordinates. P [ l ] and P[2] mean the loading o n the first and the second principal component, respectively.

W i t h the same reason and m e t h o d as f o r k n o t t y surfaces, the most important questions were selected, and a new P C A model was computed. T h e result is shown i n Figure 8. O n l y the t w o first principal components are examined i n

Figure 8, and they account f o r 54 percent o f the variation w i t h i n the interview

0.30 0.20 0.10 ° - o.oo -0.10 -0.20 -0.30 lively *e\ ventful Arich ' exciting ' interesting * exquisite : likejt

beautiful genuine itu ' fresh ' harmonious 1 e a s y j o j o o k j a t * restful 1 gaudy " restless 1 hard_to_look_at Adisharmoniuos ugly Anot fresh disgusting ^.dislike it 1 artificial ~ 4 uninteresting A unexciting Aempty 4 rigid * uneventful * strict -t 1 1--0.20 -0.10 0.00 0.10 0.20 p[1] 0.30

Figure 8. Clear w o o d surfaces. Visualization o f the answers f r o m 14 selected

questions and their variation described by a P C A model. Their loading, i.e. their importance, is represented by the distance f r o m the origin o f coordinates. P [ l ] and P[2] mean the loading on the first and the second principal component, respectively.

Comparing the results i n Figure 8 and Figure 5, we see that the same kinds o f questions are pointed out as relevant to ask. T h e question balanced/unbalanced f o r k n o t t y w o o d surfaces has been replaced w i t h fresh/not fresh and to some extent easy to look at/hard to look at f o r clear w o o d surfaces. Exquisite/disgusting has replaced the question nice/objectionable and exciting/unexciting has replaced the question stimulating/boring. This shift o f questions primarily has to do w i t h the criteria f o r sorting out important questions. The aim was to select

questions (answers) situated far f r o m the origin o f coordinates, and fairly w e l l distributed around the circular swarm i n Figure 7. This shift o f questions f o r clear w o o d surfaces compared w i t h questions f o r k n o t t y w o o d surfaces can be seen as a sign that somewhat different questions are proper to use f o r clear surfaces. One question added that diverges f r o m the questions used f o r k n o t t y w o o d surfaces is genuine /artificial. This question is located near the questions detecting the final assessment and i t detects i f a clear surface looks artificial or not.

Means to Measure the Aesthetic Properties of Wood

3.6 Detailed questions about specific wood features

W h e n t r y i n g to detect the relation between people's preferences and detailed questions about the mixture o f w o o d features, the models were rather weak

(Paper II). I t was concluded that i t is hard to force people to assign different

w o o d features a certain rank. Classifying different w o o d features requires some training. Still, there is some valuable i n f o r m a t i o n . For k n o t t y surfaces i t was the blend o f k n o t shapes, knot directions, k n o t colors and k n o t sizes that was important. Also the number and the distribution o f knots was important f o r people's preferences.

For clear surfaces there was weak connection between people's preferences and questions about texture, colors etc. T h e only w o o d feature that was o f

importance was the question about whether the surface had a lively texture or not. Mazet et al. (1990) showed, however, that difference i n color is an important factor f o r classification o f oak veneers. T h e y also showed that lightness was a very appreciated characteristic f o r assessment by professionals. T h e y conclude that i t is possible to obtain some measured parameters w h i c h can objectively describe the qualities o f the appearance o f oak veneers.

3.7 People's preferences versus objective wood feature measurements

I n Paper I7 the relation between people's preference and coarse manual

measurements o f w o o d features was studied. T h e connections were weak, but i n the research process, the results w o r k e d as indications o f what to measure i n

Paper VI. Therefore, no result from Paper F i s stressed here. I n Paper VI

detailed measurements o f w o o d features were made o n digital images o f surfaces containing knots. B o t h the w o o d surfaces and the interview data were taken f r o m the second study (Paper II). O n l y those 13 questions pointed out as important i n Paper I K were used (Figure 6). F r o m a total o f 140 x-variables ( w o o d feature measurements), the 30 most robust variables were selected f o r each model. Prediction models were then estabHshed, and almost half o f them showed strong predictive ability (Table 2, Paper VI).

I t was shown that questions/answers describing the blend o f w o o d properties i n the w o o d surface (those answers far up and d o w n i n Figure 6) were easiest to predict. Hence, the answers describing the final assessment were harder to predict. Concerning the final assessments, stronger models were obtained f o r negative answers than f o r positive answers. A m o n g the preference models, different groups o f w o o d feature measurements were used. I t was shown that almost aU groups o f w o o d feature measurements were represented among the most important variables. A deviation i n center o f gravity caused b y variation

I n an attempt to show the sign o f covariance between preference data and the most important objective w o o d feature measurements, a j o i n t model f o r all 13 questions was computed b y PLS (Figure 5, Paper VI). This plot gives good i n f o r m a t i o n about h o w different w o o d features connect w i t h preferences. This m e t h o d o f displaying results has many similarities w i t h methods o f External Preference M a p p i n g reviewed below.

3.8 Similarities with Preference mapping

I n the area o f f o o d research, consumer preference f o r products is frequently investigated by Preference mapping ( M c E w a n et al. 1998). Preference mapping shows many similarities w i t h the approach and methods used i n this thesis. Preference mapping is a useful tool b o t h i n the study o f human choice behavior to determine w h i c h characteristics o f products influence preference and whether there is consumer segmentation o f preferences. I t refers to a range o f multivariate statistical methods that are used to relate sensory data to

consumer data. Preference mapping can broadly be divided into internal and external preference mapping. Internal preference mapping provides a map o f the products and the consumers together based purely on preference ratings. P C A is the major tool w i t h i n internal preference mapping and is often

combined w i t h cluster analysis. External preference mapping relates consumer preferences to a product map obtained f r o m other product information; this may be sensory, physical or chemical i n f o r m a t i o n about the products. Here, the preference data obtained f r o m each consumer is regressed to the product coordinates obtained f r o m the external data. O f t e n , principal component regression ( P C R ) forms the basis o f external preference mapping, b u t also PLS is recommended f o r external preference mapping.

N o n e o f the papers presented here f o l l o w the principles o f preference mapping i n detail, t h o u g h they are very close. Thus, the research area o f methods f o r measuring the aesthetic properties o f w o o d could benefit f r o m an application o f knowledge f r o m f o o d research. T h e way o f interviewing and collecting i n f o r m a t i o n f o r the segmentation o f customers and experiences f r o m different analytical methods are such examples.

3.9 Limitations

T h e w o r k presented i n this thesis is based o n only three investigations where n e w interview data was collected. T h e w o o d material used i n the three investigations comprised only softwood, namely Scots pine (Pinus sylvestris), and all three investigations t o o k place i n Sweden. This lack i n the research design may n o t be so critical, as the objective was to develop an interviewing technique f o r measuring people's preferences towards w o o d . A generalization o f the results o f people's preferences presented i n this thesis w o u l d not be

Means to Measure the Aesthetic Properties qf Wood

valid. T h e more interesting issue is whether i t is possible to use similar questions and interviewing techniques i n apphcation to other w o o d species, products and people. A l l value words used i n the questionnaire emanate f r o m deep interviews w i t h people w i t h different backgrounds (Paper 1). Thus, the author has not solely created the questions used, and they seem to cover the most c o m m o n properties o f people's attitudes toward w o o d .

4. C O N C L U S I O N S

Features that people see and notice i n a w o o d surface (Figure 3, Paper I) vary very little among people, though the evaluation and the visual impression o f the m i x t u r e o f w o o d features i n a surface differ. F r o m person to person the impression o f sight is processed, and the output, a synthesis o f what they feel, is presented. Interviews made clear that people prefer different blends o f w o o d features. There are t w o qualitative differences that are o f importance f o r people's impression and valuation o f w o o d . First there is the overall blend o f w o o d features and second, the presence or absence o f divergent features that mismatch i n the surface. I t was f o u n d that divergent features are more

important than the overall mixture o f features. B u t i f there are no defects that mismatch, the overall mixture w i l l then be the key to a person's appreciation o f a w o o d surface.

I conclude that a synthesis describing that a surface appears interesting, i n harmony and w a r m describes the reason f o r a person's preference f o r a w o o d surface. Value words c o m i n g f r o m such syntheses can then be the key to the relationship between the mixture o f w o o d features and people's preferences. There are differences i n people's preferences f o r surfaces with, knots and without knots. I t is important that a w o o d surface should stimulate people's interest and be fresh l o o k i n g . A clear surface is naturally rather harmonious, elegant and easy to l o o k at. O n the other hand, a clear surface should be stimulating to l o o k at, be exciting and i t should not l o o k like an imitation. K n o t t y surfaces usually are less harmonious. Therefore, questions about harmony, easiness to l o o k at and balance are o f importance. Just as f o r clear surfaces, a k n o t t y surface should also stimulate people's interest, have a fresh look, and be exciting to l o o k at.

Results show that questions that detect people's overall visual impression are more correlated w i t h the final assessments than are questions about specific w o o d features. I t was concluded that i t is hard to force consumers to assign different w o o d features a certain rank. For k n o t t y w o o d surfaces, however, i t was the blend o f k n o t shapes, k n o t directions, k n o t colors and k n o t sizes that was significant. For clear surfaces the analysis pointed out liveliness i n the texture as an important feature.

A l o t o f questions have been examined, and about 13 questions detecting the overall visual impression have been pointed out as relevant to use i n f u r t h e r studies. T h e questions suitable to use w i t h clear w o o d surfaces showed many similarities w i t h those pointed out as relevant f o r k n o t t y w o o d surfaces. T w o methods f o r measuring people's preferences f o r w o o d were studied. T h e analysis shows that using computer images showing different looks o f w o o d

Means to Measure the Aesthetic Properties of Wood

applied to a product w i l l give the same result as using the real w o o d surfaces. Using computer images instead o f w o o d surfaces is also more practical. I n the last study i t was s h o w n that i t is possible to some extent to connect subjective preference data w i t h objective w o o d feature measurements. Results show that many w o o d feature measurements are needed and that i t is a

multivariate problem. Concerning the measurements (extracted f r o m gray-scale images), i t seems easier to f i n d features that negatively affect people's l i k i n g than to f i n d the opposite. A m o n g the most important variables were those that detect different kinds o f feature distribution over a w o o d surface, and especially those variables detecting a deviation i n center o f gravity. This study gives a hint o f what w o o d features to measure i n further studies. In the area o f f o o d research, preference mapping is an accepted technique f o r the analysis o f consumer data (McEwan et al. 1998). T h e research approach i n this thesis and the methods f o r analyzing interview data have a l o t i n c o m m o n w i t h the concept o f preference mapping. B y using knowledge f r o m the f o o d research area, the process o f developing methods f o r measuring the aesthetic properties o f w o o d could benefit significantly.

5. F U T U R E W O R K

This thesis may serve as a platform f o r further studies to increase knowledge about market requirements f o r the characteristics o f visible w o o d and w o o d products. D e v e l o p i n g means to measure the aesthetic properties o f w o o d i t w i l l most certainly require an iterative process. For instance, an improved questionnaire w i t h the most important questions has to be validated by n e w interviews w i t h more people and other w o o d surfaces. Knowledge f r o m f o o d research o f methods f o r acquiring preference data and analyzing them should be used. Once again, b o t h real w o o d surfaces and computer images (displaying the w o o d textures i n an authentic environment) should be used f o r

comparison.

The objective measurement o f w o o d features should be complemented b y measures that seem lacking and those that are insignificant should be left out. The possibilities m o d e r n image-analyzing tools provide f o r the extraction o f variables are enormous: thus there is great potential f o r i m p r o v i n g the measurement o f w o o d features. Increased knowledge about people's preferences f o r the aesthetic properties o f w o o d w i l l lead to a better understanding o f w h i c h w o o d features should be measured and controlled.

6. R E F E R E N C E S

A n o n , 1994. J M P . Reference manual Version 3.0. SAS Institute Inc., N C , U S A , I S B N 1-55544-590-x.

A n o n , 1998. S I M C A . Reference manual Version 7.0. U m e t r i A B , B o x 7960, SE907 19 U m e å , Sweden.

E f r o n , B . , G o n g , G. 1983. A leisurely l o o k at the bootstrap, the jackknife, and cross-validation. T h e American Statistician. 37(1), pp 36-48.

Eriksson, L., Johansson, E., K e t t a n e h - W o l d , N . , W o l d , S. 1999. Introduction to M u l t i - and Megavariate Data Analysis using Projection Methods ( P C A & PLS). Umetrics A B , B o x 7960, SE907 19 U m e å , Sweden.

Garvin, D . , A . 1984. W h a t does "product quaHty" really mean? Sloan Management R e v i e w . FaU 1984, pp 25-43.

Garvin, D . , A . 1988. Managing quaHty. T h e Free Press, N e w Y o r k , U S . Glaser, B . G., Strauss, A . L . 1967. T h e discovery o f grounded theory:

Strategies f o r qualitative research. Aldine D e Gruyter, N e w Y o r k . 271 pp.

Hansen, E. N . , Bush, R . J . 1996. Consumer perceptions o f softwood lumber quality. Forest Prod Journal 46(10), pp 29-34.

Hansen, E. N . , W e i n f u r t e r , S. 1999. S o f t w o o d lumber quality requirements: examining the supplier/buyer perception gap. W o o d Fiber Science 31(1), pp 83-94.

Kotier, P., Armstrong, G., Saunders, J., W o n g , V . 1996. Principles o f marketing. Prentice Hall, L o n d o n , pp 441-448.

Marchal, R . , M o t h e , F. 1994. Appreciation o f oak w o o d f o r the trench consumer and w o o d professionals. Annales des Science Forestiere 51(3), pp 2 1 3 - 2 3 1 .

Mazet, J. F., Janin, G. 1990. T h e quality o f appearance o f oak veneers — color measurements and visual appreciation o f French and ItaHan professionals. Annates des Science Forestiere 47(3), pp 255-268.

M c E w a n , J. A . , Earthy, P. J., Ducher, C . 1998. Preference mapping: A review. R e v i e w N o . 6, Campden & C h o r l e y w o o d Food Research Association, Gloucestershire, U K .

Nakamura, M . , Masuda, M . , Inagaki, M . , 1993. Influences o f knots and grooves o n psychological images o f w o o d waH-panels. M o k u z a i Gakkaishi, 39(2), pp 152-160.

Means to Measure the Aesthetic Properties of Wood

Nakamura, M . , Masuda, M . , Hiramatsu, Y . , 1994. Visual factors influencing psychological images o f woods and stones. M o k u z a i Gakkaishi 40(4), pp 364-371.

Ozanne, L . K . , Smith, P. M . , 1996. Consumer segments f o r environmentally marketed w o o d e n household furniture. W o o d Fiber Science 28(4), pp 461-477.

Pakarinen, T . , 1999. Success factors o f w o o d as a furniture material. Forest Products Journal 49(9), pp 79-85.

Patton, M . Q . 1980. Qualitative evaluation methods. SAGE Publications Inc., Beverly Hills, C A .

Risvik, E., M c E w a n , J. A . , R ø d b o t t e n , M . , 1997. Evaluation o f sensory p r o f i l i n g and projective mapping data. Food Q u a l i t y and Preference. 8(1), pp 63-71.

Swearingen, K . A . , Hansen, E. N . , Reeb, J. E., 1998. Customer preference f o r pacific northwest hardwoods. Forest Products Journal 48(2), pp 29-33. W o l d , H . , 1982. Soft modehng. T h e basic design and some extensions.

Chapter 1 i n V o l . I I o f K . - G . J ö r e s k o g and H . W o l d , Ed.s. Systems under indirect observation. N o r t h Holland, Amsterdam, 343 pp.

W o l d , S., R u h e , A . , W o l d , H . , D u n n , W . J., 1984. T h e collinearity problem i n linear regression. T h e partial least squares approach to generalized inverses. S I A M Journal on Scientific C o m p u t i n g . 5(3), pp 735-743.

Visual impressions of features i n Scots pine

wood surfaces: A qualitative study

E r r a t a : P a p e r I

Page Present Corrected 63, left, 3 lines fr. b. ...0.7 b y 1.5 m i n size... ...0.4 by 1.5 m i n size...