Clustering of Assessors The cluster

Clustering of Assessors
The clustering analysis on the subject x subject (1-Rand) matrix underlined three clusters of consumers.
The first cluster gathered 56,3% of the subjects (219 persons) and represents the mean perception of the products (first plan of the global analysis). The configuration (Fig. 5) is clearly driven by TDS. Tap water 6 is isolated on the third axis (not presented) because of its off-flavor. The preferred water samples are still those with balanced mineralization.
The second cluster gathered 36,2% of the subjects (141 persons). In this cluster, the tap water samples are identified on the first axis (Fig.6) as being tasteless and also preferred. The third axis (not presented) continously separates the water samples according to TDS, and this axis corresponds in fact to the second axis of the first cluster of subject while the second axis in Fig 6 is similiar to the third axis in Fig 4.
The third and last cluster gathered 7,5% of the subject (29 persons). This group is mainly one-dimensional (Fig 7). Bottled water samples 5 and 6 are isolated but the configuration is relatively simple in information. These people are less consistent in terms of perception of water so that this cluster gathers heterogenous perceptions.
Finally, most of the subjects were unable to distinguish between the tap and bottled water samples (63,8% of the subjects, clusters 1 and 3) and seem to like tap water as much as bottled water.
This classification was linked to the consumers characteristics via a x2 test (table 3). It was expected that sensory perception of water could be influenced by the subjects charateristics (gender, age, place of residence or consumer habits concerning water), but in fact, no link was found between this classification and the characteristics of the subject. Thus, the differences between the subjects with regard to perception may be essentially due to physiological characteristics and not to consumer habits.

Consumers preferences and consumers behavior
After having encoded the preference verbatim as described in the third part of the data analysis section, a multiple comparison test (LSD, P=0.05) was performed in order to extract the significant differences between the means obtained by each water.
Results are presented in Table 4 where the mean preference of the water samples (ranging from -1 to 1) is reported together with the TDS. In this table, the two water samples with the same lettter (A to E) should be considered as equally liked by the consumers. It is clearly confirmed that maximum preference is obtained for balanced mineralization (around 300mg/L). the further the water is from this medium mineralization, the more it is disliked (except for tap water 6, which is disliked because of its off-flavor). the three most disliked water samples were the extreme types of water: bottked water samples 5 and 6 on the one side (highly mineralized) and bottled water 1 on the other side (poor in minerals).
Finally, appreciations of the subjects who declared being the consumers of the bottled water samples were isolated. In spite of the small number of subjects who declared to regularly drink bottled water, the results presented in Table 5 show that the consumers do not systematically drink the water they like. In fact, except for the drinkers of medium-mineralized water (bottled water samples 3 and 4), the drinkers of highly mineralized water (bottled water samples 5 and 6) and the drinkers of low-mineralized water (bottled water samples 1 and 2) do not drink the water they prefer. Indeed, the drinkers of bottled water samples 5 and 6 seem to prefer low-mineralized water samples (bottled water samples 1 and 2) and noted that bottled water samples 5 and 6 as the worst. These subjects seem to be more indulgent with the taste of the water (higher menas). Finally, the drinkers of the bottled water samples 1 and 2 seem to prefer balanced water.

Discussion and perspectives

About the methodology
The free-sorting task is an efficient methodology to determine sensory perception of the drinking water. Indeed, this procedure is natural and easy to understand for untrained consumers and makes it possible to highlight the main sensory differences among the samples.
Nevertheless, categorization may include bias in other psychological process, notably an amplification of the perceived differences between groups associated with a reduction of the perceived differences within groups (Gold-stone et al 2001) and could influence hedonic judgment that way. In addition, the number of missing values induced by tis protocol in terms od preference (see part 3 of the data analysis section) is too large to enable analysis of individual differences. For these reasons, this study cannot be consideredas a real preference study and have to completed with a dedicated protocol with monadic presentation of the samples (one after the other), including quota considerations for recruitment of the consumers.
Is it also worth to note that the subjects were not chosen so as to obtain a representative group. However, it is assumed that the subjects were not used to tasting water in sensory analysis experiments (“naive” subjects) and the results are supposed to be consistent and representative of a mean preference for the product
Finally, free sorting is a useful tool to grasp the sensory space of water, but this methodology cannot be used as aroutine procedure as the projection of new samples on the map needs a new evaluation of the samples already tasted (this method does not enable “data aggregation”)
About the water samples
Three main tastes of water have been highlighted based on the level of minerals in water. However, it is worth to note taht no water sample with TDS content between 400 and 2000 mg/L was tasted in this study. However, other studies conducted by the authors with other water samples howed a continuum of taste between the tasteless balance water samples and the salty and astringent highly mineralized water samples. Moreover, it would have been interesting to add duplicates in the samples in order to evaluate the discrimination power through tis continuum.
The potential contribution of the specific mineral contents of the different water samples was investigated. However, in this study, the quantity of most addition, sensorydifferences between the water samples were too small to highlight the influence of some specific minerals in the taste of water.
Finally, the study highlighted the impact of TDS in the taste of water, but other factors such as contaminants or minerals, individually, can also influence the perception of the taste of water.

About the subjects
It has been demonstrated in the study that most consumers cannot distinguish between water and tap water when the latter is chlorine-free (see also Wells 2005). However, 36% of the subjects were found able to ditinguish between tap water and bottked water and the reasons were not identified.
In addition, it was not possible to investigate individual preference because of the missing values induced by the protocol. This should be studied through a dedicated experiment.
Finally, the gap between preferences and behavior has been emphasized and tend to prove the influence of other factors than those linked to sensory perception. Beyond physiological or habit considerations, information and education could be a way to potentially change consumers perception, preference and behavior.

Conclusion
Several topics have been addresed in this study, which is the first step of a larger project aiming at understanding sensory perception, preference and behavior of consumers toward drinking water. First of all, it is demonstrated that when presented with no chlorine, tap water is globally perceived in the same way as bottled water. The sensory characteristics of water (taste differences) are mainly due to their total mineral contents, and three main tastes have been highlighted: the bitter taste of water with low mineralization, the neutral taste of medium-mineralized water (around 350 mg/L) and the astringent and salty taste of highly mineralized water.
Water liking was also found to be influenced by mineral content, and water being medium in minerals waas preferred because it is perceivedd as tasteless and cooler.
Finally, this study also highlighted the gap between consumers preferences and consumption habits. Somme consumers drink water they do not like.

Results (Indonesian) 1:

Clustering of Assessors The clustering analysis on the subject x subject (1-Rand) matrix underlined three clusters of consumers. The first cluster gathered 56,3% of the subjects (219 persons) and represents the mean perception of the products (first plan of the global analysis). The configuration (Fig. 5) is clearly driven by TDS. Tap water 6 is isolated on the third axis (not presented) because of its off-flavor. The preferred water samples are still those with balanced mineralization. The second cluster gathered 36,2% of the subjects (141 persons). In this cluster, the tap water samples are identified on the first axis (Fig.6) as being tasteless and also preferred. The third axis (not presented) continously separates the water samples according to TDS, and this axis corresponds in fact to the second axis of the first cluster of subject while the second axis in Fig 6 is similiar to the third axis in Fig 4. The third and last cluster gathered 7,5% of the subject (29 persons). This group is mainly one-dimensional (Fig 7). Bottled water samples 5 and 6 are isolated but the configuration is relatively simple in information. These people are less consistent in terms of perception of water so that this cluster gathers heterogenous perceptions.Finally, most of the subjects were unable to distinguish between the tap and bottled water samples (63,8% of the subjects, clusters 1 and 3) and seem to like tap water as much as bottled water.This classification was linked to the consumers characteristics via a x2 test (table 3). It was expected that sensory perception of water could be influenced by the subjects charateristics (gender, age, place of residence or consumer habits concerning water), but in fact, no link was found between this classification and the characteristics of the subject. Thus, the differences between the subjects with regard to perception may be essentially due to physiological characteristics and not to consumer habits.Consumers preferences and consumers behavior After having encoded the preference verbatim as described in the third part of the data analysis section, a multiple comparison test (LSD, P=0.05) was performed in order to extract the significant differences between the means obtained by each water. Results are presented in Table 4 where the mean preference of the water samples (ranging from -1 to 1) is reported together with the TDS. In this table, the two water samples with the same lettter (A to E) should be considered as equally liked by the consumers. It is clearly confirmed that maximum preference is obtained for balanced mineralization (around 300mg/L). the further the water is from this medium mineralization, the more it is disliked (except for tap water 6, which is disliked because of its off-flavor). the three most disliked water samples were the extreme types of water: bottked water samples 5 and 6 on the one side (highly mineralized) and bottled water 1 on the other side (poor in minerals). Akhirnya, apresiasi subyek yang dinyatakan menjadi konsumen air botol sampel diisolasi. Meskipun jumlah kecil subyek yang dinyatakan secara teratur minum air Kemasan, hasilnya disajikan dalam tabel 5 menunjukkan bahwa konsumen tidak sistematis minum air yang mereka suka. Pada kenyataannya, kecuali para peminum menengah-mineralized air (air botol sampel 3 dan 4), para peminum sangat mineralisasi air (air botol sampel 5 dan 6) dan para peminum rendah-mineralized air (air botol sampel 1 dan 2) jangan minum air yang mereka suka. Memang, para peminum air botol sampel 5 dan 6 tampaknya lebih suka sampel rendah-mineralized air (air botol sampel 1 dan 2) dan mencatat bahwa air botol sampel 5 dan 6 sebagai yang terburuk. Mata pelajaran ini tampaknya akan semakin memanjakan dengan rasa air (lebih tinggi menas). Akhirnya, para peminum air botol sampel 1 dan 2 tampaknya lebih suka air seimbang.Diskusi dan perspektifMengenai metodologi Tugas gratis-menyortir adalah metodologi yang efisien untuk menentukan sensorik persepsi air minum. Memang, prosedur ini alami dan mudah untuk memahami untuk konsumen tidak terlatih dan menjadikannya mungkin untuk menyorot sensorik perbedaan utama antara sampel. Nevertheless, categorization may include bias in other psychological process, notably an amplification of the perceived differences between groups associated with a reduction of the perceived differences within groups (Gold-stone et al 2001) and could influence hedonic judgment that way. In addition, the number of missing values induced by tis protocol in terms od preference (see part 3 of the data analysis section) is too large to enable analysis of individual differences. For these reasons, this study cannot be consideredas a real preference study and have to completed with a dedicated protocol with monadic presentation of the samples (one after the other), including quota considerations for recruitment of the consumers. Is it also worth to note that the subjects were not chosen so as to obtain a representative group. However, it is assumed that the subjects were not used to tasting water in sensory analysis experiments (“naive” subjects) and the results are supposed to be consistent and representative of a mean preference for the product Finally, free sorting is a useful tool to grasp the sensory space of water, but this methodology cannot be used as aroutine procedure as the projection of new samples on the map needs a new evaluation of the samples already tasted (this method does not enable “data aggregation”)About the water samples Tiga utama selera air telah menyoroti berdasarkan tingkat mineral dalam air. Namun, hal ini layak untuk dicatat bahwa tidak ada sampel air dengan TDS konten antara 400 dan 2000 mg/L itu terasa dalam studi ini. Namun, studi lain yang dilakukan oleh para penulis dengan sampel air lain howed sebuah kontinum dari rasa antara sampel air tawar keseimbangan dan asin dan zat sangat mineralized sampel air. Selain itu, ia telah menarik untuk menambahkan duplikat dalam sampel untuk mengevaluasi kekuatan diskriminasi melalui tis kontinum. Kontribusi potensi isi mineral tertentu dari sampel air yang berbeda itu diselidiki. Namun, dalam studi ini, jumlah tambahan kebanyakan, sensorydifferences antara sampel air itu terlalu kecil untuk menyorot pengaruh beberapa mineral tertentu dalam rasa air. Akhirnya, penelitian menyoroti dampak TDS dalam rasa air, tetapi faktor-faktor lain seperti kontaminan atau mineral, secara individual, juga dapat mempengaruhi persepsi rasa air.Tentang mata pelajaran Ini telah ditunjukkan dalam studi yang kebanyakan konsumen tidak dapat membedakan antara air dan air keran ketika yang terakhir bebas-klorin (Lihat juga Wells 2005). Namun, 36% dari mata pelajaran yang ditemukan dapat ditinguish antara air keran dan bottked air dan alasan yang tidak diidentifikasi. Selain itu, hal itu tidak mungkin untuk menyelidiki preferensi individu karena nilai-nilai hilang yang disebabkan oleh protokol. Ini harus dipelajari melalui percobaan yang berdedikasi. Finally, the gap between preferences and behavior has been emphasized and tend to prove the influence of other factors than those linked to sensory perception. Beyond physiological or habit considerations, information and education could be a way to potentially change consumers perception, preference and behavior.Conclusion Several topics have been addresed in this study, which is the first step of a larger project aiming at understanding sensory perception, preference and behavior of consumers toward drinking water. First of all, it is demonstrated that when presented with no chlorine, tap water is globally perceived in the same way as bottled water. The sensory characteristics of water (taste differences) are mainly due to their total mineral contents, and three main tastes have been highlighted: the bitter taste of water with low mineralization, the neutral taste of medium-mineralized water (around 350 mg/L) and the astringent and salty taste of highly mineralized water. Water liking was also found to be influenced by mineral content, and water being medium in minerals waas preferred because it is perceivedd as tasteless and cooler. Finally, this study also highlighted the gap between consumers preferences and consumption habits. Somme consumers drink water they do not like.