Presentation of International Conference by Kiwamu Maki

Presentation by Kiwamu MAKI at international conference

4)
K.Maki, S.Kobayashi, Y.Nakamura, M.Inui. The Hierarchy of Words Used in the Evaluation of Environments. 15th Conference of the International Association for People-Environment Studies (Eindhoven), 1998.

1. Introduction
The ability to predict environmental preference using variables that represent environmental characteristics would facilitate the creation of preferable environments. The approach often taken in research of this type has been to have subjects assess environmental preferences and their predictor variables, and then investigate the validity of said predictor variables.
For instance, Kaplan et al. (1972) found strong correlation between preference and complexity in the three categories of nature slides, urban slides and other slides, while Herzog (1989) has defined nine predictor variables (e.g., spaciousness, refuge, coherence and legibility) and examined their relevance to preferences for urban environments containing prominent natural elements.
The authors have performed the following experiment. First, parameters relevant to the preferability of streetscapes were identified in interviews. Next, subjects were asked to rate these parameters, after which their degree of relevance to preferability was indicated as correlation coefficients and regression coefficients in regression analysis. Nakamura et al. (1994), who published some of these results, report that regardless of the type of streetscape, preferability can be predicted with great accuracy from serenity and cheerfulness.
However, an inconsistency was discovered in this experimentation: Although an evaluative parameter (i.e., a parameter having relevance to evaluation) should, by definition, have an effect on evaluation, either by raising or lowering the evaluation, some of the parameters cited as justification for an evaluation had a correlation to preferability ratings and a partial regression coefficient both approaching zero - suggesting that the evaluative parameter had no effect whatsoever on preferability ratings.
One possible reason for such experimental results may be the fact that data of varying degrees of relevance were analyzed together. In the experimentation described by Nakamura et al. (1994), the same word was sometimes used to represent a parameter relating to streetscape preferability that had both positives and negative effects on evaluation (Table 1). If the direction (i.e., negative or positive) in which evaluation is affected differs in some cases, then the degree of relevance may have been underestimated due to the failure to keep those cases separate (Fig. 1). It is for this reason that in this experiment the authors decided to calculate the degree of relevance after separating parameters represented with the same word according to the direction of effect (i.e., negative or positive) as stated by the subjects. The first half of this paper reports on the result of this experiment. The second half uses those results to reveal the hierarchy that exists in the meaning of words used to evaluate environments, and then touches on the limitations of conventional research methods.
<Table. 1>
<Fig. 1>

2. Experimentation
First, a preliminary experiment was held in which subjects rated the preferability of sixty-eight slides of urban streetscapes in Tokyo and its environs and explained their justifications for those ratings. The justifications given were summarized separately for each slide.
In the main experiment, the sixty-eight slides used in the preliminary experiment were shown to the subjects and rated by them using ten 7-point semantic differential scales, one of which was "preferability". The subjects were then asked to select, from among the summarized justifications, those parameters they perceived when rating preferability. The subjects were then asked to state whether the parameters thus selected (i.e., parameters cited as justification for their evaluations) either raised or lowered their evaluation of preferability, or had no effect on evaluation.
An evaluative procedure such as this one makes it possible to ascertain not only which parameters are consciously regarded as affecting evaluation (i.e., as being relevant to evaluation), but also the direction (i.e., negative or positive) in which each parameter affects evaluation.
The subjects comprised 29 students: 10 men and 19 women.

3. Results
The degree of relevance between a parameter and preferability is often indicated as a correlation coefficient or regression coefficient. For this experimentation, however, the authors decided to express this relevance in the form of the mean rating of preferability for all cases in which a parameter was cited as justification. Because only cases of clear relevance to evaluation are used to calculate the degree of relevance, this method offers the advantage of greater accuracy in calculations. Here, the degree of relevance determined with this method is also referred to as the parameter's effect on evaluation.
Roughly similar trends were observed upon comparing the effects of parameters calculated from the authors' experimental data with the sizes of correlation coefficients and partial regression coefficients obtained in experimentation by Nakamura et al. (1994). This indicates that a mean rating representing a parameter's effect is a suitable scale for degree of relevance.
When instances of positive and negative effects were kept separate in the calculating of the parameters' effects, a degree of relevance was frequently ascertained even in the cases of parameters previously found to have a correlation coefficient or partial regression coefficient approaching zero.
Data termed "the character of the parameter" were calculated in order to determine the source of these differences in the direction of effect (positive versus negative). These data were the mean ratings (made using ten semantic differential scales) for scenes in which each parameter was cited as justification for the evaluation made. In other words, these data can be viewed as indicating the average image profile of all scenes in which the parameter was cited as justification. In this experiment, separate parameters represented with the same word were treated as separate data for the purpose of determining the differences between cases of positive and negative effects on evaluation. The results of factor analysis on these data are shown in Fig. 2.
<Fig. 2>

Each plotted point in the illustration represents the parameter's character: The character of a parameter can be read from the vector at the center of the illustration. The parameter "unity", for instance, means that coherence is perceived when the evaluation of a scene is positively affected (+) by that parameter, but that a lack of friendliness is perceived when evaluation is negatively affected (-). Likewise, friendliness and warmth are perceived to a considerable degree when the parameter "ambiance of daily living" raises a scene's evaluation, but are not perceived very much when the parameter lowers a scene's evaluation, in which cases a lack of unity is perceived. These are the types of conclusions that can be drawn from the illustrations. Thus, with parameters such as unity and ambiance of daily living, the parameter's character depends on whether that parameter has a positive or negative effect on evaluation.
Each parameter's effect on evaluation is represented by the projection vector on the "preferability" axis in Fig. 2. That values vary according to relevance to the parameter's character is interpreted not as meaning that the unity or ambiance of daily living perceived in the streetscape has affected the rating of preferability, but as meaning that evaluation has been affected by the coherence perceived in unity or by friendliness perceived in the ambiance of daily living.

4. Discussion
The "ambiance of daily living" to which the authors refer as justification for evaluation is most likely a verbalization of the primary factors that cause the subjects to perceive either friendliness and warmth or a lack of coherence. This interpretation is reinforced by the results of the aforementioned experiment, which show that it is the perceptions of friendliness and warmth or lack of coherence associated with daily existence that are affecting evaluation. If this is so, then there is not much significance in accurately measuring the quantitative degree to which a feeling of daily existence is perceived. Rather, it is the quality of daily living that must be represented - i.e., indicating which types of existence are perceived as having friendliness and warmth, or lacking coherence.
Fig. 3 contains the evaluative justifications cited regarding the ambiance of daily living. These evaluative justifications are each positioned somewhere in a hierarchy comprising the following four elements(Table. 2).
Characteristics: What is cognitively recognized in the object of evaluation
Condition: One's impression of the object of evaluation itself
Character: Feelings aroused in the evaluator by the object of evaluation
Evaluation: One's evaluation of the object of evaluation
<Fig. 3>
<Table. 2>

With evaluative parameters belonging to "character", changes in ratings correspond to changes in ratings of overall preferability. As reported by Nakamura et al. (1994), parameters belonging to this level of the hierarchy, such as serenity and cheerfulness, did not require the categorizing of objects of evaluation in order to describe the relevance between said parameters and overall evaluation in the case of streetscapes.
With parameters belonging to "characteristics" and "conditions", in contrast, the same word can have different connotations (i.e., "characters") when it represents two different parameters. In this experiment, parameters such as "unity" and "ambiance of daily living" are such parameters.
Concepts conventionally used as predictor variables of landscape preferability (e.g., refuge, legibility, complexity and nature) belong to these levels of hierarchy. One must avoid considering these predictor variables merely as explanatory variables of "preferability," and instead strive to determine the categories of objects of evaluation in which a parameter has a consistent direction of effect.



References
Herzog, T. R. (1985). A cognitive analysis of preference for waterscapes. Journal of Environmental Psychology, 5, 225-241.
Herzog, T. R. (1989). A cognitive analysis of preference for urban nature. Journal of Environmental Psychology, 9, 27-43.
Herzog, T. R. (1992). A cognitive analysis of preference for urban spaces. Journal of Environmental Psychology, 12, 237-248.
Kaplan, S., Kaplan, R., and Wendt, J. S. (1972). Rated preference and complexity for natural and urban visual material. Perception & Psychophysics, 12, 354-356.
Kaplan, S. (1987). Aesthetics, affect, and cognition; environmental preference from an evolutionary perspective. Environment and Behavior, 19, 3-22.
Kaplan, R., Kaplan, S., and Brown, T. (1989). Environmental preference: A comparison of four domains of predictors. Environment and Behavior, 21, 509-530.
Sanui, J. and Inui, M. (1984). Towards a phenomenological model of place evaluation: Paper presented at the IAPS 8th international conference in Berlin, W-Germany.
Sanui, J. and Inui, M (1986). Individual Differences in the Evaluation of Living Rooms: Paper presented at the IAPS 9th international conference in Haifa, Israel.
Y.Nakamura, K.Maki, M.Inui(1994). Relationship between the Classification of Urban Streetscape and Overall Evaluation: Proceedings of the IAPS 13th international conference in Manchester, England, 351-361.
R.Lamb, A.Purcell, E.Peron and S.Falchero(1994). Cognitive categorization and preference for places: Proceedings of the IAPS 13th international conference in Manchester, England, 405-416.