Physiology & Behavior, Vol. 56, No. 1, pp. 81-93, 1994 Copyright © 1994 Elsevier Science Ltd Printed in the USA. All rights reserved 0031-9384/94 $6.00 + .00
Pergamon 0031-9384(94)E0025-Y
Perceptual Characteristics of Binary, Trinary, and Quaternary Odor Mixtures Consisting of Unpleasant Constituents DAVID
G. L A I N G , *l A N D R E W
EDDY'~ AND
D. J O H N
BEST1:
*Faculty of Science and Technology, University of Western Sydney, Bourke Street, Richmond NSW 2753 Australia, tCSIRO Division of Food Science and Technology, PO Box 52, North Ryde NSW 2113 Australia, and ~:CSIRO Biometrics Unit, Institute of Animal Production and Processing, PO Box 52, North Ryde NSW 2113 Australia R e c e i v e d 2 S e p t e m b e r 1993 LAING, D. G., A. EDDY AND D. J. BEST. Perceptualcharacteristicsof binary, trinary,and quaternaryodormixturesconsisting of unpleasant constituents. PHYSIOL BEHAV 56(I) 81-93, 1994.--Among the most obnoxious stimuli that the population at large is exposed to during everyday life are odorous emissions from sewage treatment plants. Such emissions are complex and contain many different types of odorants that vary in quantity depending upon the contents and efficiency of treatment processes. Because little is known about how individual odorants in complex mixtures affect the perception of each other, it is difficult to develop mathematical models that can predict the pleasantness, strength, and quality characteristics of an emission at different distances from a source. In the present study, the interactions of the four major types of odorants emitted by treatment plants worldwide, namely, hydrogen sulphide, isovaleric acid, butanethiol, and skatole, were investigated by measuring the perceived intensity of individual odorants alone and in mixtures, and the overall perceived intensity, unpleasantness, and qualities of mixtures. In addition, models for predicting odor strength were investigated. The results indicated that (i) the perceived odor intensity (odor strength) of mixtures of the odorants was equal or greater than that of any of the individual constituents, but less than the sum of their intensities. However, as the number of components in a mixture increased, the intensity of the most dominant component provided a good approximation of the intensity of the mixture. (ii) The vector model of intensity summation also satisfactorily predicted the odor intensity of mixtures containing two, three, or four of the odorants investigated. (iii) In no instance was the intensity of one odorant enhanced by another, i.e., no synergistic interactions occurred; the greater the number of odorants in a mixture, the greater was the degree of suppression of the individual constituents. (iv) The greater the number of constituents in a mixture the more difficult it became to identify individual constituents. (v) Hydrogen sulphide was the least frequently suppressed constituent, and isovaleric acid and skatole were the most frequently suppressed constituents in mixtures. (vi) The unpleasantness of mixtures was usually greater than that of the individual constituents, indicating that models used for predicting complaint levels in communities affected by sewage odor and based on assumptions related to a single odorant, e.g., hydrogen sulphide, will underestimate the number of complaints. Even mixtures with low but above threshold concentrations of these odorants are likely to generate complaints. Odorants
Odor mixtures
Hydrogen sulphide
Skatole
ONE of the most obnoxious stimuli experienced by humans is that due to the odorous constituents of sewage odor. Like most of the odorous stimuli encountered during everyday life, sewage odor consists of dozens o f odorants, many of which are unpleasant and are a source of annoyance to the general community. Accordingly, odors from sewage treatment plants have become increasingly important as sources of complaints from the public as residential development continues to encroach onto buffer zones around plants, and public expectation in relation to odor nuisance control continues to rise. Effective odor control strategies, however, depend in part on understanding the interactions of various odorous constituents when they mix in the atmosphere following emission from a treatment plant. Unfortunately, little
Isovaleric acid
Sewage odor
is known, for example, whether the strength of odorants such as hydrogen sulphide, a c o m m o n constituent o f sewage emissions, is increased or decreased by other volatile constituents, or whether its "rotten e g g " characteristics are altered to become more or less unpleasant. Accordingly, it is difficult to predict the odorous characteristics of a mixture that will be encountered by the public from a knowledge of the composition of emissions. Studies of the effects that constituents of mixtures have on each other have been reported since the beginning o f this century (10); however, very few have dealt with mixtures containing more than two constituents. Thus, although there is some knowledge of the types of effects odors have on the perception of each other, there is very little information on the action of a third or
To whom requests for reprints should be addressed. 81
82
fourth component, etc. Furthermore, because studies of volatile odorants require that the odorants be mixed in gaseous streams, sophisticated olfactometers are required for meaningful data to be obtained. This technical requirement has been an influential factor in reducing the number of mixture studies reported and together with the very unpleasant and difficult to handle components of sewage emissions that include fatty acids, sulphides, thiols, and amines, few studies of mixtures containing one or more of these odorants have been reported (1,2). From studies with two component mixtures, the major effect produced when odorants are mixed is the suppression of the perceived intensity of one or both odors (6). If the suppression of an odorant is substantial, it can be reduced in strength to a level that renders it undetectable. Although the strength of an odor can be increased by the presence of another, an effect commonly referred to as synergism, the latter is rarely seen with two-component mixtures. Substantial synergistic effects, however, have been reported by Laska and Hudson (8) with a complex mixture of many subthreshold odorants (no individual odorant was detectable when sniffed alone). This is an important finding and could have considerable relevance to the levels of odors encountered from emissions. However, along with other reports concerned with three-, four-, and five-component mixtures where the addition of further components to binary mixtures altered the characteristics of the mixtures (9), the data of Laska and Hudson suggest that the interactions of odorants in mixtures with more than two components can produce phenomena for which we have very little knowledge and understanding. Regarding the interactions of unpleasant odor components, the most relevant studies are those by Berglund (3), who reported the outcomes of mixing the odorants hydrogen sulphide, dimethylsulphide, dimethyldisulphide, and methylmercaptan. Unfortunately, in that study only the overall intensity of mixtures of these components was measured and there was no description of the odor qualities of the mixtures, of the perceived intensity of the individual constituents when presented in mixtures, or of the pleasantness of mixtures. Nevertheless, this study showed that the odor intensity of two-, three-, and four-component mixtures of these substances only slightly exceeded the intensity of the single odors. Berglund et al. (1,2) also developed a mathematical model, the Vector Model of Intensity Summation, to predict the overall intensity of mixtures. However, although the model adequately predicted the intensity of two-component mixtures, it underestimated the intensity of more complex mixtures. Indeed, when it was used to predict the intensity of pulp mill odor for which the above constituents were major components, the predictions were out by more than 50% (2). The authors proposed that unknown neural mechanisms underlying the perception of these odors appeared to be the cause of the difference rather than any other factor related to the sensory measurement method or physicochemical properties of the odorants. The lack of methods for predicting the odor quality, strength, and hedonic characteristics of complex mixtures, including environmental odors, is a major problem confronting pollution authorities as well as researchers in the flavor and fragrance industries and those concerned with unravelling the complex odor messages used in chemical communication between animals. Mathematical models used by a number of water and environmental bodies throughout the world to predict the strength of emissions, for example, are generally unsatisfactory, partly because they are based on nonolfactory assumptions. In addition, these models do not account for the suppression or enhancement of the strength of individual odor constituents due to odor interactions, or for changes in the characteristics of mixtures that can
LAING, EDDY AND BEST
alter the degree of unpleasantness and annoyance level of environmental mixtures. In view of our lack of information on the effects produced when more than two odorants are mixed, in particular when the constituents are unpleasant, the present study aimed to investigate the odor interactions, odor intensity, pleasantness, and quality characteristics of mixtures containing four of the major types of odorants commonly found in sewage, namely, hydrogen sulphide (rotten eggs), butanethiol (rotting cabbage), skatole (feces), and isovaleric acid (rancid, sweaty, sour). To achieve these aims the study was conducted in three parts: 1. Selection of the intensity of single odorants to provide four concentrations of each to cover most of the range of perceived intensities encountered in the environment. The four concentration levels were approximately matched in perceived intensity over the four odorants. 2. Determination of the total perceived intensity and pleasantness of stimuli consisting of one, two, three, or four components. 3. Determination of the perceived odor quality characteristics and intensities of the individual constituents of stimuli consisting of one, two, three, or four components. GENERAL METHOD The odorants used were hydrogen sulphide (ex CIG), butanethiol, isovaleric acid, and skatole (all ex Fluka AG and were of the highest purity available). Subjects (between 23 and 51 years old) were scientific staff of the CSIRO Division of Food Processing and had some experience with sensory tests. To obtain quantitative data on gaseous mixtures of the odorants, a computer-controlled, 16-channel, air-dilutionolfactometer was constructed. This device produced four concentrations of each odorant that could be presented singly or in mixtures consisting of two, three, or four components. Hydrogen sulphide was delivered from a cylinder and its concentrataion in the olfactometer was controlled by dilution with oxygen-free nitrogen and carbon-filtered air from a Teflon bearing compressor. Butanethiol and isovaleric acid were maintained undiluted in W-shaped glass vessels in which the diameter and length of each of the four arms of the W were 1 x 15 cm, respectively. Skatole, a solid, was dissolved (10% w/v) in "odorless" propylene glycol (AR Grade BDH), in a similar vessel. The glass vessels were sited in a water bath maintained at 20 +__I°C. Different concentrations of the three odorants in glass containers were produced by first passing a stream of oxygen-free nitrogen (ex CIG) over the surface of the odorants and diluting the resulting odorous nitrogen streams with a low flow rate of nitrogen. The odorous nitrogen streams emerging from the vessels were saturated with vapor from each odorant at the flow rates used. Each diluted odor stream was directed to a Teflon solenoid valve that allowed passage of the stream either to an exhaust or to a fast flowing stream of air (50 1 min-~). The odor stream and fast flowing air stream were immediately mixed in a Teflon chamber before passing into a Perspex sampling chamber where a subject could sniff the odor. Flows of nitrogen, air, and hydrogen sulphide were controlled by pressure regulators, flow meters, and flow controllers, and the direction of flow through the solenoid valves was controlled by a computer (Hypec 286). The final concentrations of each odorant are given in Table 1. Measurement of odorant concentrations was achieved using several methods. Hydrogen sulphide concentrations were obtained using a high sensitivity (ppb) Jerome hydrogen sulphide
PERCEPTION A N D O D O R M I X T U R E S
83
TABLE 1 ODORANT CONCENTRATIONS Concentration Levels (ppb) Odorant
1
2
3
4
Hydrogen sulphide (A) Butanethiol (B) Isovaleric acid (C) Skatole (D)
18.7 206 65.5 2.4
43.2 527 111 8. t
91.2 1660 231 15.7
170 3990 336 30.0
sensor. This device was inserted into the sampling chamber and measurements were taken at the position where a subject's nose was located during tests. Butanethiol and isovaleric acid levels were calculated from nitrogen and air flows, saturated vapor pressures of the two odorants, and peak areas from a gas chromatograph. However, because of the need for special chromatographic columns for these substances and necessary derivatisation of the acid for its measurement, butanol was substituted and used to determine the dilutions produced by each of the nitrogen and air streams. Skatole concentrations were calculated from peak areas in gas chromatograms of known (injected) concentrations in the nitrogen and air streams in the olfactometer before dilution with the 50 1 min- ~ air stream. The latter measurement technique was necessary because concentrations of skatole in the sampling chamber were too low to be detected by the gas chromatograph. The computer performed a number of functions. As described above, it controlled which odor streams were delivered to a subject. It also gave instructions via a monitor to a subject in an adjacent room and monitored their responses in the various tests. Specific details of each of the test procedures are given below with each phase of the study. The test environment was an air-conditioned mobile laboratory consisting of two rooms; one housed the olfactometer and the experimenter, and the other contained the sampling chamber of the olfactometer and a subject. PART 1: SELECTING PERCEIVED INTENSITIESOF THE FOUR ODORANTS Method Following the initial choice of four concentrations for each odorant by the experimenter to provide four matching sets of intensities, five subjects attended several sessions where they were instructed to rate the perceived intensity of each of the odorants on a graphic rating scale. At each session, an odorant was presented 10 times at each of the four concentrations to each subject. Order of presentation of the 20 stimuli (four concentrations of each odorant and four samples of air) was randomised for each subject and session. When all subjects had completed a session, the mean panel score for the perceived intensity of each concentration was calculated, Where matching of intensities within a level had not been achieved, the concentrations were adjusted by the experimenter in an attempt to achieve a match. The following day the adjusted concentrations were presented for assessment until the mean intensity score of the panel for each concentration provided four levels of approximately equal intensity odorants. When assessing the perceived intensity of stimuli, subjects were seated before a single Perspex sampling chamber that had a large sniffing port into which the subjects inserted their face until an approximate seal was made with the edges of the port.
When the head of a subject was in this position, their nose was located directly above a 50-mm tube that was the outlet from the olfactometer through which the stimulus flowed at 50 1 min -~. Once in position, the subject sniffed the odor stream and judged the strength of the stimulus. Subjects registered their response using a mouse to place a mark on a 195-mm graphic rating scale shown on the monitor screen. The scale had 75 integer divisions not seen by the subject and had the words " n o o d o r " and "extremely strong" at its ends. There was an intertrial interval of 45 s. After this time an instruction appeared on the monitor to indicate to a subject when to commence sampling; a period of 15 s was allowed for insertion of the face into the sampling port and sampling of the odorant. Results and Discussion The mean perceived intensities of each of the odorants at the four concentration levels selected are shown in Table 2. Despite the fact than an iterative procedure for selecting concentrations was used, it was unusually difficult with these four odorants to produce mean intensity levels for the panel that provided approximately equal increments between the concentration levels. Difficulty in assessing the intensities by the panel may have arisen because of the restricted range of concentrations we wished to use to keep the intensities within environmentally relevant limits. Approximately similar levels of perceived intensities, however, were achieved for most of the odorants. Note that the scale ranged from 0 (no odor) to 75 (extremely strong). Thus, the highest intensities of each of the odorants were between 51.2 and 63.0, indicating they were moderate to strong. At the other end of the scale, intensity values of approximately 20 were recorded for the lowest concentrations of three of the odorants, indicating they were of weak intensity. Odorant intensities, therefore (except for that of isovaleric acid), covered the range normally encountered in the environment. PART 2: INTENSITY AND PLEASANTNESSOF MIXTURES Method In this part of the investigation, the same five subjects who participated above were required to rate the perceived intensity and pleasantness of the stimulus presented using the same interactive procedure with the computer using a mouse as in Part 1. However, in this instance, the stimulus could consist of one, two, three, or four odorants. Subjects assessed the overall intensity of the stimulus first using the same scale as in Part 1 and were asked to ignore the intensity and identity of any individual odors perceived. Once they had rated the intensity of the stimulus on the monitor, the intensity scale was automatically replaced with another graphic rating scale (195 mm) that had the words "exTABLE 2 PERCEIVED INTENSITY OF INDIVIDUALODORANTS* Concentration Level 1 2 3 4
Air
Hydrogen Sulphide (A)
Butanethiol (B)
lsovaleric Acid (C)
Skatole (D)
9.5 9.5 10.4 8.2
19 24.3 35.2 51.2
24.8 39.0 55.0 60.7
38.5 38.5 44.5 54.9
21.6 29.9 40.2 63.0
* The 195-mm graphic rating scale with 75 (hidden) integer divisions was used. Values are related to the 75 divisions.
84
LAING, EDDY AND BEST
tremely unpleasant" and "extremely pleasant" at the ends. Subjects were instructed to assess the overall pleasantness of the stimulus by using the mouse to place a mark on the rating scale. In brief, on instruction from the monitor, subjects placed their face in position in the sampling port, sniffed the odor stream, assessed perceived intensity, withdrew, then sniffed the stream again and estimated the pleasantness of the odor. Subjects were allowed a 15-s sampling time, an unlimited response time, followed by a 45-s interval. The 60 odor stimuli comprising single odors and mixtures are given in Table 3. Each odor stimulus was presented eight times to the subjects over 16 test sessions, and 30 samples were assessed at each session. All stimuli were sampled once before any was given for a second time. At each session, in addition to receiving the odor stimuli, one stimulus was air (blank) to check the background level of the air stream that was used to dilute the odors and to provide a baseline from which to determine if the intensity ratings of the lowest concentration samples were above those for air. Odorants, at low concentrations, for example, can suppress each other to undetectable levels (8). Results and Discussion Odor intensity. The data from this part of the investigation were analysed in three ways. 1. For each of the four concentration levels of each odorant and for each of the combinations of odorants showed in Table 3, mean values of the overall intensity of each single odorant and mixture were calculated. These values are shown in Fig. l(a-k), where the intensities of the constituents of each mixture and the overall intensity of the mixture are plotted. Analyses of variance (ANOVAs) were used to check for overall differences between odorants, odor levels, subjects, and replicates, and to see if the response curves were parallel, i.e., to test whether there was an interaction between odorants and their levels. Note that all effects described as significant throughout this study have an alpha level of 0.05. The most important effects identified by the ANOVAs were significant differences between the intensity levels of each of the odorants, between the responses of the subjects, and between responses to the different stimuli. The absence of an interaction between odorant and replicate was an encouraging outcome with each combination of odors, supporting the reliability of the data. Clearly, however, the most striking effect observable in the plots is the very limited addition of the intensities of the constituents when they are in mixtures. In every instance, the function representing the mixture, e.g., AC (hydrogen sulphide and isovaleric acid), is very similar in shape and magnitude to the functions of the constituent odors. Strong suppression of intensity is therefore the major feature of mixing this particular group of odorants. This finding is reinforced in the other two types of analyses given below. The strong suppression and lack of additivity is also witnessed by the similar values of the maximum intensities recorded for mixtures of two, three, or four components at each of the four concentration levels (Fig. 1). Thus, increasing the number of components was not paralleled by a corresponding increase in intensity. 2. To obtain a clearer description of the extent of an interaction (suppression or synergism) that occurred when the odorants were mixed, the ratio IA+IB was calculated, where lab = overall perceived intensity of a mixture of the odorants A and B, Ia = perceived intensity of odorant A before mixing, IB = perceived intensity of odorant B before
TABLE 3 TEST STIMULI Air (blank)
ml A2 A3
A4 B, B2 B3
B4 Ci C2 C~ C4 D, D_, D~ D4
A~BI A2B2 A~B~ MB4 A,ct A~C2 A~C~ A4C4 A~D~ A~D2 A3D~ A4D4 B~C, B,C2 B~C~ B4C4 B,D, B~D2 B~D~ B4D4 C,D, C~D2 C~D~ C~D4
A,B,Ct A2B~C2 A3B~C3 A4B4C4 A,B,D, A2B2D2 A~B3D3 A4BaD4 A~C~D~ A2C2D2 A3C~D3 A4C4D4 B,C,D, B2C202 B3C3D3 B4C4D4
A~BtC,D~ A2B2C2D2 A~B3C3D3 A4B4C4D4
A, hydrogen sulphide; B, butanethiol; C, isovaleric acid; D, skatole. Subscripts indicate the concentrationlevel, with 1 the lowest and 4 the highest level. See Table 1 for details of the concentrationlevels.
mixing. The value of the ratio gives an indication of the type and degree of interaction between the odorants. Values that are > 1, 1, or < 1 indicate synergism, perfect additivity, and suppression, respectively. Values of the intensity ratios from each mixture are plotted in Fig. 2(a-k) for each subject. In addition, three-way ANOVAs were conducted (subjects, replicates, and concentration levels as the factors) with each mixture type, e.g., AB, ACD, ABCD. Using the error term from the ANOVAs, 95% confidence bands around a value of 1.0 (perfect additivity) were constructed, as shown in Fig. 2(a-k). The most striking feature of the plots in Fig. 2 is the dramatic decrease in the value of the intensity ratios as the number of constituents in a mixture increases [compare Fig. 2(a), (g), and (k)], indicating that greater interaction between the components, namely suppression of intensity, occurred. With every trinary and quaternary mixture the value of the intensity ratio was below that of the lower confidence band and there was no incidence of synergism with any mixture. Interestingly, the decreases in ratios were largely independent of odor concentration, and in the case of binary and quaternary mixtures were independent of odorant type. Whether the reduced ratio values were due to mutual suppression between odorants or to the suppression of only one odorant cannot be determined from this data; however, this was investigated in Part 3. Another important feature of the results is the good agreement between the ratios produced by different subjects. Importantly, the magnitude of the differences between subjects was most obviously reduced with the high concentrations of most binary and trinary mixtures, and the largest reduction occurred with three of the four quaternary mixtures. Thus, the variability between subjects was reduced as the number of constituents in mixtures increased, and generally less variability was observed with
[
g
F
g
0
8
N"
N
N"
g
~..~ g
~,'g.
10
,50
OC)
m >
nM(~.I,,
3O ),
Percmv~l Intensity (ram)
5"
0
.
50
0
|
.
n
*
30
10 a
,
50 O0 C) O
a
Perceived Intensity (ram)
.
.
.
.
30
10
Perc~vod tntlmsity
.IL
n
|
30 n
n
SO |
30
GO
0
Perce*ved Inten-W 10
n
10
P'et~t~vld tt~mty
A
|
|
3O |
|
50 |
30
50 o
PercmwKl Intens4ly 10
*
10
Pemmwd Inwn,,zy
A
f~o
IL
5O
10
30
GO
Perc~ved IntensW
3O
P e e v e d IntenWty 10
PO
.
.
30
n
50 i >
30
50
O
Perceived Intensity 10
.
.
Perce*ved Intensely 10
O"
86
LAING, EDDY AND BEST (a)
O
(b)
V
o
;" :
T
i
......
.......... .. ......... g
I
.......
..... ~
....... 41- ........ -~ .......... ,~ .......
.....
i
i
i
!
i
,
i
2
3
4
1
2
3
4
(d)
(c) v
V ._&_ 13
o
V
~
[]
=
o .=.
~
. ..... ..v....................
.~ .............
IE
,
i
i
,
i
=
|
|
I
2
3
4
I
2
3
4
(~
(e) [] 0
[] ..... ~
;
....... ~ . . . . . . . . . ~
......... ~ .......... ~
.
.
.
.
.
.
........ ~ ......... .~.
........ .. ......
.
-
Mean
v 1
2
3
4
Concentration Level
1
2
3
4
Concentration Level
FIG. 2. Plots of the ratio of total perceived intensity and sum of the intensities of the individual constituents of mixtures at four ascending concentration levels. See Tables 1 and 3 for odorant concentrations and names, respectively.
mixtures consisting of the higher concentration levels of the odorants. The data involving only binary mixtures indicated that very different interactions occurred with each pair of odorants. For example, in no instance did the mean intensity ratio of mixtures containing skatole (D) lie outside the confidence bands. In con-
trast, there was a significant interaction at three of the four concentration levels between isovaleric acid (C) and butanethiol (B) and with hydrogen sulphide (A). The latter two odorants interacted with each other only in mixtures containing the highest concentration levels. The greatest departure from total additivity of intensities, therefore, occurred with isovaleric acid (C) in its
PERCEPTION AND ODOR MIXTURES
87
(g)
(h)
........................................................
i= 1
2
3
4
1
i
w
i
2
3
4
(i)
(J)
q
+
•
,¢
V
o~ v
i
i
I
2
3
4
|
w
i
i
1
2
3
4
Concentration Level
(k) .......................................................
--
Mean
o •
Subject 1 Subject 2
+
*
Subject 3
!
v
Subject 4 Subject 5
i 1
2
3
4
Concentration Level FIG. 2. Continued
interactions with the two sulphur-containingodorants, only partial additivity being recorded. On the other hand, no significant departure from total additivity of intensities occurred with mixtures containing skatole (D). 3. The intensity of odor mixtures was also examined using the Vector Model of Intensity Summation (1,2). This model was proposed for predicting the odor intensity of a binary mixture lAB
from the perceived intensity of the unmixed components IA and IB by means of the equation
lAB =
(12 + 12 +
21AIBcos.)I/2
where a is a constant related to the qualifies of the components A and B. Derivation of a is achieved from this equation in the special case where the intensities of IA and IB are approximately
88
equal. Although it has been proposed that a ratio scaling procedure rather than the present graphic rating scale be used for obtaining intensity estimates for Vector model calculations, satisfactory values of cosc~ and of correlation coefficients have been reported using a category scale (6). Because the vector model has also been shown to give a modest fit to data from three-, four-, and five-component mixtures (3) (a constant angle between vectors is assumed), with some tendency to underpredict the intensity of mixtures with four or five components, the model was used in the present study to determine its suitability for predicting the total intensity of mixtures of the particular odorants used here. Calculations based on the Vector model are shown in Table 4, and indicate a very good fit between the predicted and observed values of odor intensity [r0.05(2df) = 0.95]. The values of a for binary mixtures fall comfortably within the range of reported values [102 ° (1), 108 ° (4), 109-115 ° (6), 109 ° (7)]. Only with the quaternary mixtures is there a significant departure from the predicted value for the slope of the vector (r = 0.95 is required for p = 0.05). Otherwise, the Vector model appears to be a suitable vehicle for predicting the intensities of mixtures of the four odorants. Odor pleasantness. The mean pleasantness ratings of the panel for each single odorant at the four levels of concentration and for each of the test stimuli listed in Table 3 are plotted in Fig. 3(a-k). Four-way ANOVAs were employed to determine the effect of replication, subjects, odor concentration, and odorant type on the pleasantness of the stimuli. The most important effect that emerged was the significant increase in the unpleasantness of odor stimuli as concentration and perceived intensity grew (p < 0.001 with each stimulus), regardless of whether the stimulus was a single odorant or a mixture. Furthermore, in most instances the level of pleasantness of a mixture was lower than the pleasantness of the individual (unmixed) components. In brief, mixtures were more unpleasant than their individual components and the unpleasantness of all stimuli was greatest at the highest concentrations used. These are important findings for this group of sewage-related odorants and they suggest that complaints from a community will not only increase as their perceived intensity grows, but will be in greater numbers than predicted by mathematical models if the odor perceived consists predominantly of two or more of the types of odorants used here. Even simple binary mixtures appear to follow this general rule of unpleasantness. Thus, a mixture consisting predominantly of hydrogen sulphide (A) and skatole (D) [Fig. 3(c)] will smell more unpleasant than either of the components alone. Another important result was that the level of unpleasantness recorded, particularly with the highest concentration levels of odorants, was close to the "extremely unpleasant" rating. Indeed, even with the weakest of odorants, for which concentrations were not far above recognition threshold, the mean ratings were below the midpoint of the hedonic scale, indicating that even at the weakest levels the mixtures were unpleasant. Accordingly, as regards community responses to these odorants, complaints can be expected even at these low levels. In regard to the other factors investigated with the ANOVAs, there were significant differences between the responses of subjects (p < 0.001), as is common in hedonic studies. In this instance, however, the variation was confined to the "unpleasant" end of the rating scales. In other words, although there were differences between the responses of different individuals there was general agreement that the odor stimuli were unpleasant. There were also significant differences in pleasantness resulting from different odorants (p < 0.001). The measurements in Part 2 precluded identifying which odorant contributed most to the
LAING, EDDY AND BEST
TABLE 4 VECTOR MODEL OF INTENSITYSUMMATION: CALCULATED PARAMETERS Odors
a
Slope
Intercept
r
AB AC AD BC BD CD ABC ABD ACD BCD ABCD
109° 117° 106° 107° 107° 111 °
1.13 1.01 1.04 1.01 I. 12 0.87 0.85 0.98 0.92 1.04 0.40*
- 3.50 -0.16 - 1.06 -0.23 - 3.70 4.06 - 3.81 -0.73 -4.21 - 5.08 -5.48
0.95 0.99 0.97 0.99 0.99 0.99 0.99 0.99 0.98 0.99 0.95
A, hydrogen sulphide; B, butanethiol; C, isovaleric acid; D, skatole. * Indicates the slope of the vector is significantly different to 1. unpleasantness of a stimulus, but this is discussed further in Part 3, where identification of odorants is investigated. Nevertheless, it is clear from Fig. 3 that regardless of the composition of mixtures, the odors of mixtures were almost always more unpleasant than the constituents. There seems little doubt, therefore, that foul-smelling odorants in combination produce even foulersmelling mixtures. PART 3: IDENTIFICATIONAND INTENSITYOF MIXTURE COMPONENTS
Method Subjects were advised that each stimulus could consist of zero, one, two, three, or four odorants. Their task was to select which odorant(s) was present and estimate its perceived intensity. Each trial commenced with the appearance of four boxes on the computer screen with the name of one of the four odorants in each box. When sampling a stimulus, a subject used the mouse to indicate which odorants were present. As each box was marked with the mouse, the box was transformed into a graphic rating scale that had the words " n o odor" and "extremely strong" at the ends. The subject then used the mouse as in Parts 1 and 2 to indicate the perceived intensity of each component detected. The time allowed for sampling and decision making on each trial was the same as that described in Part 2. The 60 stimuli used were those in Table 3 and all were presented in random order over two consecutive sessions before the next replicate occurred. Each was presented eight times to the five subjects over 16 test sessions. As in Part 2, a control (air) was presented once at each session.
Results and Discussion To fully investigate the data obtained, the latter were analysed and presented in three ways. 1. Identification of individual constituents when presented alone and in mixtures. Chi-square analyses were used to assess whether the level of identification differed when an odor was presented alone or in mixtures. The percentage of hits (correct identifications) is plotted for each odorant in each stimulus in Fig. 4(a-d). The plots show that in most mixtures identification of hydrogen sulphide (A) [Fig. 4(a)] and butanethiol (B) [Fig. 4(b)] was not significantly impaired when either was part of a
l=u
<
g
N
~,~
8~
20
30
OI"II~'
ItlE
i
10
40
Pleasantness (ram)
.<_
Q
I-
5"
i (o
r.,)
h~
i
,
i
30
o o
40
10
20
30
40
Pleasantness (ram)
20
10
Pleasantness
.~...
S-
10
10
30 ,
i
30
20
~tness
20
l~emirllnell
0
40
40
,b
po
10
i
i
i
30
30 i
20 i
Plessan[ness
20
10
0
O
40
40
.lh
f,,)
2O i
i
i
20
10
i
10
i
30
i
30
Pl414ianlneN
4O
GI
40
¢z.
A
W
lb
10
10
|
30
:~
30
l ) ~ S $
i
20
PteeSWllne14 40
40
.*
~r"
90
LAING, EDDY AND BEST
TABLE 5
mixture compared to when they were sampled alone. In contrast, identification of isovaleric acid (C) was significantly reduced in mixtures (on 11 out of 28 occasions), particularly at the two highest concentration levels employed [Fig. 4(c)]. Difficulties encountered in identifying isovaleric acid (C) in mixtures seem to have stemmed from the large suppression of the perceived intensity of the acid by the other odorants. The intensity of isovaleric acid (C) in mixtures presented at the two highest concentration levels is reduced in most instances to less than its intensity when presented alone or in mixtures at the two lowest concentration levels [Fig. 6(c)]. Thus, the intensity of the acid recorded at the upper concentration levels in most cases may have been too low for it to be identified, or its identity may have been lost as a result of blending perceptually with other odorants, a common event in flavors and fragrances. Similarly, in the three instances where there is a significant reduction in the identification level of skatole (D) in mixtures, the intensity of skatole (D) is significantly reduced, e.g., in BD, ABD, and BCD at the highest concentration level (see Fig. 6). In general, the identification of skatole (D) was
PERCENT IDENTIFICATION OF ODORANTS IN MIXTURES Concentration Level Mixture
1
2
3
4
Alone* Binary Trinary Quaternary
47.5 50.0 46.9 43.8
68.1 56.3 52.5 53.8
75.0 67.3 59.8 56.9
89.4 72.3 62.7 63.1
* Indicates percentage of hits averaged over all odors and concentrations when each odor was presented alone.
not frequently impaired in mixtures [Fig. 4(d)]; significant reduction occurred in only four out of 28 stimuli. Further analyses of correct identifications showed that significantly less hits (poorer identification) were recorded as the num-
4a
o 1 2 3 4
o 1 2 3 4
o 1 2 3 4
o 1 2 3 4
1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4
4b 8
8 ~
1 2 3 4
1 2 3 4
1 2 3 4
1 2 3 4
2
1 2 3 4
1 2 3 4
1 2 3 4
1 2 3 4
3
4
~
I
2
3
4
o
1
2
3
4
1 2 3 4
1 2 3 4
4d
°
o 1
2
3
4
° I
2
3
4
I
2
3
4
I
2
3
4
I
2
3
4
I
2
3
4
I
2
3
4
Concentration Level FIG. 4. Comparison of the percentage of correct identifications of an odorant alone (open bars) and in mixtures (filled bars) with two, three, or four components at four ascending concentration levels. For example, in (a) open and filled bars represent percent correct identifications of A alone and in mixtures of AB, AC, AD, ABC, etc. In (b) the bars represent percent correct identification of B alone and in mixtures of BA, BC, etc. An asterisk signifies that there was a significant difference at an alpha level of 0.05 between the identification level of an odorant alone and in a mixture. See Tables 1 and 3 for odorant concentrations and names, respectively.
PERCEPTION AND ODOR MIXTURES
91
TABLE 6 PERCENT CORRECT IDENTIFICATIONS OF EACH ODORANT IN MIXTURES Odorant
Mixture
Hydrogen Sulphide
Butanethiol
isovaleric Acid
Skatole
Alone* Binary Trinary Quaternary
71.9 64.3 58.8 55.0
62.5 60.8 56.9 57.5
80.0 63.1 56.3 58.1
65.6 57.5 50.0 46.9
* Percent correct identificationswhen an odorant was presentedalone. bet of constituents in a mixture increased (see columns in Tables 5 and 6). In contrast, significantly more hits occurred as the concentration level increased (see rows in Table 5), but there was no significant difference between the hits recorded for each odorant when the data were combined over all concentration levels and mixtures (60.6% hydrogen sulphide; 58.7% butanethiol; 59.5% isovaleric acid; 52.8% skatole). 2. Identification levels of each component in mixtures. As indicated above, there was very little difference between the percent identifications of the individual odorants over all mixtures. Nevertheless, comparison of the hit levels of the individual constituents in particular mixtures [Fig. 5(a-k)] indicates that within a number of mixtures one odor was identified significantly more often than another. Figure 5(a-f) shows that with binary mixtures one odorant is identified more often than the other on 9/24 (37.5%) instances. Skatole (D), for example, appears to be less identifiable in mixtures with hydrogen sulphide (A) [Fig. 5(c)] and butanethiol (D) [Fig. 5(e)] than with isovaleric acid (C) [Fig. 5(f)], where it was identified significantly more often than the acid at the highest concentration level. The decreased hit levels recorded with skatole in mixtures with hydrogen sulphide (A) and butanethiol (B) parallel the significant reduction in the perceived intensity of skatole (D) in these mixtures (Fig. 6). In trinary mixtures [Fig. 5(g-j)] significant differences in the identification levels of the constituents occurred with 7/16 (43.7%) mixtures. However, no consistent identification patterns emerged and no odorant was identified more often than another in any regular way. Thus, the pattern of hits at each concentration level for a particular mixture, e.g., BCD [Fig. 50)], was not consistent over the four concentration levels, and no odorant emerged as the one most frequently identified. As with binary and trinary mixtures, no consistent pattern of hits emerged with the quaternary mixtures [Fig. 5(k)] over the four concentration levels, and no single odorant dominated the mixtures. Over all the mixtures, regardless of whether they contained two, three, or four constituents, the ease of identifying the individual odors varied depending on the particular odors in a mixture and the concentration level. 3. Perceived intensity of odorants alone and in mixtures. ANOVAs were used to determine whether there were differences between the intensity of an odorant when it was presented alone or in mixtures. The mean perceived intensity recorded by the panel for each odorant and the results of the analyses are shown in Fig. 6a-d. The most striking feature of this figure is the high frequency of suppression of the intensity of each odorant. Of the 28 mixtures studied with each odorant, there were 15/28 [hydrogen sulphide (A)], 19/28 [butanethiol (B)], 26/28 [isovaleric acid (C)], and 20/28 [skatole (D)] instances where significant suppression of an odorant occurred.
Suppression of hydrogen sulphide (A) occurred most frequently at the two upper concentration levels, with few instances at the two lower levels. In contrast, there was only one instance (BACD) when butanethiol (B) was suppressed at the highest concentration level, but suppression occurred with 18/21 of the mixtures containing this odorant at the lower levels. As regards the other two odorants, the intensity of isovaleric acid (C) and skatole (D) was significantly reduced in 26/28 and 20/28 mixtures, respectively. The intensity of isovaleric acid (C), therefore, was clearly the most consistently affected by the other constituents. However, the intensity of all odorants was suppressed in most mixtures, hydrogen sulphide (A) being the least affected. The frequency of occurrence of suppression of individual odorants increased as the number of constituents in a mixture increased. Thus, suppression occurred on 62.5%, 75.0%, and 87.5% of occasions with binary, trinary, and quaternary mixtures, respectively. However, the difference in the occurrence of suppression at low and high intensities was not so clear-cut, with overall occurrences being 57.2%, 71.4%, 82.1%, and 75.0% for concentration levels 1-4, respectively. GENERAL DISCUSSION
This investigation had several major goals. One of these was to establish the type and extent of olfactory interactions that characterise mixtures of four unpleasant odors that are usually the major groups of volatiles that comprise sewage odor. It was not previously known whether these odorants in combination enhanced or suppressed the perception of each other, or affected the overall intensity of a mixture. By studying the perception of these substances at different concentration levels in mixtures, it became very clear that, as with the majority of mixtures of pleasant or unpleasant odors studied elsewhere (1,4,6,7), suppression was the dominant perceptual effect. Indeed, in common with those other studies, in no instance was enhancement or synergism observed. The results showed that the overall intensity of a mixture (Fig. 2) could not be predicted on the basis of the simple addition of the intensifies of the constituents. As shown in Fig. 2, the extent of additivity decreased as the number of constituents in mixtures increased. Prediction of the intensity of mixtures containing up to four components (but not of the individual components) was achievable using the Vector Model of Intensity Summation (1,2), in contrast to an earlier report that indicated significant deviations occurred with three- and four-component mixtures. The applicability of this model to more complex mixtures containing these and other odorants from sewage remains to be determined. However, a simpler basis for predicting intensity is apparent from Fig. 1, which shows that the intensity of a mixture is close to that of the most dominant (most intense) component. Thus, if the dominant odorant of a sewage emission is known, it should be possible to estimate the intensity of a mixture from its concentration. A second goal of this study was to establish the quality characteristics of mixtures consisting of two, three, or four of the odorants. To this end it was significant that the decrease in additivity of component intensities as the number of components increased (Fig. 2) was parallelled by a reduction in the intensity of the individual components (Fig. 6) and in the ease of their identification (Figs. 4 and 5, Tables 5 and 6). The decreased ability of the panel to identify the constituents in mixtures was not so apparent from a comparison of percent hits when an odor was perceived alone and in mixtures (Fig. 6), but, as shown in Table 5, it was apparent from the total percentages of odorants identified in mixtures. Although the identification of all odorants was substantially reduced in mixtures (Tables 5 and 6), identification was reduced most often when the odor was isovaleric acid (Fig. 4).
92
LAING, EDDY AND BEST
(a)
1
(b)
2
3
4
1
(c)
2
(d)
3
4
1
2
(e)
3
4
3
4
(~
m 1
2
,.-~
3
4
1
2
(g)
3
4
1
2
(h)
0
(i)
*
•
g_ 1
2
3
4
1
2
3
(j)
1
2
4
1
2
3
4
(k)
3
4
1
2
3
4
Concentration Level FIG. 5. Comparison of the percentage of correct identifications of the constituents of binary, trinary, and quaternary mixtures at four concentration levels (see Table 1). The bar on the left-hand side of each set of bars is designated as that of the first odorant listed on the ordinate, e.g., the open bars in the top left-hand diagram (AB) represent odorant A and the black bars odorant B. An asterisk signifies when one odorant was perceived significantly more often than another at an alpha level of 0.05. The names of the odorants are given in Table 3.
Whether the reduced levels of identification arose as a consequence of the lowered perceived intensities of the odorants, from blending of the odorants, or was due to both of these factors is not known. Nevertheless, it can be concluded that there is a substantial loss of identity of individual odorants in mixtures, in agreement with the finding that humans can rarely identify more than four odorants in a mixture (5), and have considerable difficulty in identifying three or four odorants. The latter discussion, however, is incomplete without reference to the third major goal of this study, which was to determine if there was a relationship between the composition and unpleasantness of mixtures. The results indicated that the mixtures were almost always more unpleasant than any of the individual constituents, and this suggests that the above loss of identity was not always due to the reduced intensity of a component, but could have been due to the blending of one or more constituents to produce an even fouler smell. The finding that mixtures generally smell more unpleasant than any of the individual constituents has significant implications for models that are currently used to predict annoyance or complaint levels in communities encountering sewage odors or other unpleasant environmental odors. It is common practice, for example, to
base calculations and predictions on the number of ODUs (odor dilution units) emitted by a source, or by assuming, in the case of sewage odors, that the emission is solely hydrogen sulphide, regardless of its composition. However, both of these calculations neglect the amplified unpleasantness of sewage odor, which is normally comprised of most if not all the odor types studied here, and will underestimate the impact of odor emissions on communities. From the above discussion, three features emerge that impinge directly on estimates and predictions of complaints and annoyance in the community. First, because of the phenomenon of odor suppression, combinations of odorants do not appear to be of much greater intensity than the most dominant component. The intensity of the dominant component can, therefore, be taken as a good approximation of the overall intensity of an emission. Second, mixing odors reduces the chance of identifying individual constituents and possibly the source of an emission. However, if blending produces a smell that is commonly identified as sewage odor the source will be readily identified and the level of annoyance and complaints is likely to increase. Thirdly, mixing odors increased the overall unpleasantness of the odors and this too is likely to increase complaint levels. Whether the effects of suppression (reduction of odor intensity) and increased unpleas-
PERCEPTION AND ODOR MIXTURES
93
6a
1
2
3
4
I
2
3
4
I
2
3
4
I
2
3
4
I
2
3
4
I
2
3
4
I
2
3
4
4
I
2
3
4
I
2
3
4
I
2
3
4
I
2
3
4
I
2
3
4
1
2
3
4
4
I
2
3
4
I
2
3
4
I
2
3
I
2
3
4
I
2
3
4
I
2
3
4
4
I
2
3
4
1
2
3
4
1
2
3
I
2
3
4
I
2
3
4
I
2
3
4
6b
E E
v
ffl ¢,-
I
2
e-
O_
3
6c
~
8
1
2
3
6d
1
2.
3
4
Concentration Level FIG. 6. Perceived intensity of the individual constituents in mixtures at four ascending concentration levels (see Table 1). The open bar in each set of bars represents the target odor alone, the black bar represents the target odor in a mixture. An asterisk signifies that there was a significant reduction in the perceived intensity of an odorant when it was in a mixture at an alpha level of 0.05. The names of the odorants are given in Table 3.
antness tend to cancel each odor's impact on the level of annoyance and complaint remains to be determined. Finally, the findings in the present study that identification of individual odorants becomes more difficult as the n u m b e r of constituents in a mixture increases, and mixtures o f unpleasant odors are more unpleasant than the individual constituents, strongly suggest that blending or fusion o f odorants perceptually occurs even in simple mixtures. Accordingly, there is a
clear need for the m e c h a n i s m of blending to be resolved, if an understanding of odor mixture perception is going to be achieved. ACKNOWLEDGEMENTS The authors wish to thank Ms Maree O'Sullivan for statistical and graphical assistance and the Urban Water Research Association and Water Board, Sydney, for financial support.
REFERENCES 1. Berglund, B.; Berglund, U.; Lindvall, T.; Svensson, L. T. A quantitative principle of perceived intensity summation in odor mixtures. J. Exp. Psychol. 100:29-38; 1973. 2. Berglund, B.; Berglund, U.; Lindvall, T. Perceptual interactions of odors from a pulp mill. Proc. 3rd Int. Clean Air Congr. A40-A43; 1973. 3. Berglund, B. Quantitative and qualitative analysis of industrial odors with human observers. Ann. NY Acad. Sci. 237:37-51; 1974. 4. Cain, W. S. Odor intensity: Mixtures and masking. Chem. Senses Flay. 1:339-352; 1975. 5. Laing, D. G.; Francis, G. W. The capacity of humans to identify odours in mixtures. Physiol. Behav. 46:809-814; 1989.
6. Laing, D. G.; Panhuber, H.; Willcox, M. E.; Pittman, E. A. Quality and intensity of binary odor mixtures. Physiol. Behav. 33:309-319; 1984. 7. Laing, D. G.; Willcox, M. E. Perception of components in binary odor mixtures. Chem. Senses 7:249-264; 1983. 8. Laska, M.; Hudson, R. A comparison of the detection thresholds of odor mixtures and their components. Chem. Senses 16:651-662; 1991. 9. Moskowitz, H. R.; Barbe, C. D. Profiling of odor components and their mixtures. Sensory Processes 1:212-226; 1977. 10. Zwaardemaker, H. C. (1900) Die compensation von Geruchsempfindungen. Translated in Perf. Ess. Oil. Rec. 50:217-221; 1959.