To be published in Behavioral and Brain Sciences (in press)

© Cambridge University Press 2007

 

Below is the unedited, uncorrected final draft of a BBS target article that has been accepted for publication. This preprint has been prepared for potential commentators who wish to nominate themselves for formal commentary invitation. Please DO NOT write a commentary until you receive a formal invitation. If you are invited to submit a commentary, a copyedited, corrected version of this paper will be posted.

 

A study of the science of taste:  On the origins and influence of core ideas    

Robert P. Erickson      

Departments of Psychology and Neurosciences, and Neurobiology, Duke University, Durham, N.C. 27708, USA

eric@psych.duke.edu

 

Abstract: Our understanding of the sense of taste is largely based on research designed and interpreted in terms of the traditional four “basic” tastes: sweet, sour, salty and bitter, and now a few more. This concept of basic tastes has no rational definition to test, and thus it has not been tested. As a demonstration, a preliminary attempt to test one common but arbitrary psychophysical definition of basic tastes is included in this paper; that the basic tastes are unique in being able to account for other tastes. This definition was falsified in that other stimuli do about as well as the basic words and stimuli. To the extent that this finding might show analogies with other studies of receptor, neural, and psychophysical phenomena, the validity of the century-long literature of the science of taste based on a few “basics” is called into question. The possible origins, meaning and influence of this concept are discussed. Tests of the model with control studies are suggested in all areas of taste related to basic tastes. As a stronger alternative to the basic tradition, the advantages of the across-fiber pattern model are discussed; it is based on a rational data-based hypothesis, and has survived attempts at falsification. Such “population coding” has found broad acceptance in many neural systems.

Keywords: across-fiber pattern, basic tastes, distributed neural coding, population coding, psychophysics, receptors, reductionism, systems biology, taste

 

“Science is made up with facts as a house is made from stones. But a collection of facts is no more a science than a pile of stones is a house.”

Poincaré, Hypotheses in Physics

 

1. Introduction    

Over a century of research and understanding in the field of taste has been largely contained within a four-taste model, now referred to as the “basic tastes,” using the words sweet, salty, sour and bitter, or stimuli representing these words. This model holds that there are only these few tastes, and each taste has receptor and neuron types specifically serving it. At the neural level this model is equated with the “labeled-line” idea that each taste has its own dedicated neural line into the brain. There is a very different approach that does not involve basics, but posits a continuum of tastes served by “across-fiber patterns” of activity over populations of neurons.

 

There has been a long-standing controversy between the basic tastes and the across-fiber pattern positions. Why has it lasted so long rather than being resolved by a few experiments early on?  The position is taken in this paper that the concept of “basic tastes” is not a proper scientific model because it is not potentially falsifiable. This model has been convenient enough to be easily adapted to any experimental situation, receptor, neural or psychophysical. But the vagueness and flexibility which provide this convenience prevent it from being adequately defined, which is a fundamental fault in science. A proper definition of the idea is taken here to mean that it is based on an underlying data-based and rational hypothesis that generates precise definitions of the terms used, and thus provides the context for tests that could possibly falsify the idea (Popper 1963). The basic tastes model does not include these provisions, and therefore such tests have not been carried out. Under these conditions, there can be no resolution of the controversy. On the other hand, the across-fiber pattern model meets these two requirements of science, a rational data-based definition and testability. This model has been supported in a variety of attempts at falsification.

 

The major thrust of this paper is to critically examine the across-fiber pattern and basic tastes models of the organization of taste. The across-fiber pattern model will be cast in the format of Young’s theory of color perception, and the basic tastes model will be drawn on the backdrop of Hanig’s famous experiment on the distribution of taste sensitivities across the tongue.

 

The argument will be made that, although the idea of basic tastes has directed our research for over a century, we don’t know what basic tastes are; we don’t know the origins, rationale or definition of this idea, and thus we do not explicitly know how to test it.

 

2. “Basic Tastes”

 

2.1. Background

 

The origins of the idea that there are a few basic tastes predate modern science by at least two millennia. The evolution of the concept is rooted in ancient cultural and philosophical thought long before the advent of the scientific method. It also appears to be derived from the characteristics of our cognitive abilities. These may be the primary root reasons for the development of the idea of basic tastes, but it is clear that this model has never been construed as a scientific hypothesis to be tested.

 

2.2. Basic Tastes from antiquity

 

In China during the early Han dynasty, the first two or three centuries BC, two ways of organizing experience were formalized: these were the Yin-Yang dichotomy, and the five ways of wu xing (Unschuld 1993).  Prominent in the latter were the five tastes, sweet, sour, salty, bitter, and acrid. Both of these aspects of knowledge are evident in modern thought.

 

The Yin-Yang idea points to the importance of differences, meaning that a concept cannot exist without something different from it; “day” would not exist without “night,” “high tide” needs “low tide,” good and bad depend on each other, male and female, red depends on green, and there would be no “salty” if there were no other taste. This differentiation provides the logical basis for taste discrimination and its quantification in the across-fiber pattern theory (Section 6.4 Quantification).

 

2.3. Basic tastes and information theory   

 

But why did the Chinese create the idea of five tastes?  The reason is not immediately obvious. Perhaps it is clear from simple sensory experience that five tastes are distinct perceptual entities, like pigs, pineapples, and planets. Or it may be that the ancient Chinese had jars of spices on the table that included a sugar and a salt, a citrus juice, bitters, and something acrid (pepper?); they needed - and we need - words to communicate about these important substances.  Still the main reason for the idea of five tastes may have been that the traditional Chinese method of organizing experiences, the “five ways” demanded by wu xing. But then where did these five ways come from?

 

By the 20th century the idea of basic tastes had not evolved much beyond asking how many there are; Boring stated that the general consensus was that there are four (1942). Of relevance here, in an information theory approach, Miller showed that human memory limits the contents of facile categorizations to about 7 members each (1956). His research provides one very practical rationale both for the invention of the five ways of Chinese thought, and for our four or five basic tastes. Our limited ability to remember the contents of a category containing more than “the magical number seven plus or minus two” of unrelated items certainly could have influenced our understanding in the field of taste, as it does in many other aspects of our description of the world.

 

As examples of the ubiquity of such categorizations, we have seven days in the week, seven sins and seven virtues, and Newton’s seven colors - all arbitrary man-made rubrics. The many geological time periods are packaged into groupings and sub-groupings of seven or less each. The standard format of scientific papers in that they have about six sections - Introduction, Methods, Results, Discussion, Summary and Conclusions - which is very useful. But why is this format so useful? We have not been able to deal with the 365 days of the year without dividing that number into small categories; four seasons of three months each, each month consisting of about four weeks, each week of seven days.  The year is a natural phenomenon - once around the sun - but these other terms are not. Aristotle - the great classifier of all things into short lists, proposed seven tastes. These short lists seem to make good intuitive sense. But are there really seven seas and seven continents? And is the Mediterranean Sea really made up of seven mini-seas? Should we anticipate seven tastes?

 

3. Hanig and basic tastes

 

3.1. Hanig sets the stage

 

The modern empirical history of basic tastes is herein taken rather arbitrarily as starting with a seminal experiment by Hanig (1901). He demonstrated that the areas of sensitivity of the tongue to the basic tastes differ, with sweet largely in the front, bitter in the back, and salty and sour on the sides. Although their points of maximum sensitivity differ, their total distributions overlap greatly. It has been data of this sort that has supported the idea that we can sense only these “basic” tastes - the main assumption of the basic tastes doctrine. Subsequent research in taste largely followed his format. In the present paper Hanig’s work is used as a short-hand rubric for the idea of basic tastes.

 

            An attempt to define the idea of basic tastes would be helpful here, although it will become clear throughout this paper how nebulous and indefinable this idea is.  But at this point, the idea of basic tastes may be put simply: The basic tastes are all we can sense, thus all tastes are reducible to only these few separate and distinctly different sensations.

 

3.2. Following Hanig

 

The idea of four basic tastes gave us a very convenient structure for research. The job then was just to find evidence for them. And we did.

 

3.2.1. Psychophysics. The field of psychophysics followed Hanig in general form; subjects can describe tastes in the terms of the basic words (Halpern 2002a). Textbook treatments often state that it is generally accepted, or that it is a fact, that there are only four tastes for humans (e.g. Bartoshuk 1988; Bernstein et al 2006; Geldard 1964; Graham 1990; Kandel et al 1991; Moncrieff 1956; Osgood 1956; Pfaffmann 1951; 1954; Shepherd 1994; Smith and Davis 2000; Smith and Scott 2001; Steward 2000). Throughout these texts, definitions of “basic tastes” are notably absent. But the influence of the idea of a few basic tastes has been so strong that Hanig’s broadly overlapping areas of sensitivity have sometimes been redrawn into four discrete and non-overlapping areas on each side of the tongue, and these drawings make excellent illustrations for text books (e.g. Bear et al 1996; Graham 1990; Kandel et al 1991; Shepherd 1994; Steward 2000). It should be pointed out that Hanig’s data have been reproduced accurately in some cases (e.g. Geldard 1964; Purves et al 2001), and properly corrected in another (Smith and Margolskee 2001).

 

3.2.2. Receptor organization. Much research on taste receptors has been presented in tacit or explicit support of the basic tastes idea; experiments have been routinely designed and interpreted in terms of a small number of “basic” typologies. For example, when appropriate techniques had been devised, there began a protracted search for the basic taste receptor mechanisms or cells, such as for “sweet” etc. (Birch 1977; Eggers et al, 2000; Shallenberger and Acree 1967, 1971; Tateda 1965). Current data on receptors using molecular genetic techniques (Chandrashekar et al 2006; Zhang et al. 2003) have been taken to support basic tastes in that there are receptor mechanisms that are described as specific to two of the basic tastes, sweet and bitter (Zhang, et al. 2003; Zhao et al. 2003), with similarly specific neural pathways (Sugita and Shiba, 2005) and behaviors related to these receptors and pathways (Mueller et al. 2005; Zhang, et al. 2003; Zhao et al. 2003). A neutral search without the bias to find basic receptors would have been preferable.  It might be that with at least 40 years of research directed to finding the receptors for the basic tastes, acceptable receptors for each would eventually be found.

 

3.2.3. Neural organization. The history of the search for neuron typologies to correspond with the basic tastes began with the first recordings from individual taste neurons over 60 years ago (Pfaffmann 1941) and continues to the present (e.g. Boughter and Smith 1998; Boughter et al. 1999; Lemon and Smith 2005, 2006; Scott and Giza 2000; Smith and Scott 2001; Smith et al 2000; St. John and Smith 1999).  The “labeled-line” coding idea in taste holds that each neuron represents only the basic taste to which it responds best, giving us four groups of neurons (Frank 1974, 2000). Mathematical techniques have been used to categorize data on neural responsiveness into a few types to support this position (Scott and Giza 2000; Smith and Davis 2000; Smith and Margolskee 2001; Smith and Scott 2001; Smith and St. John 1999; Smith and Vogt 1997; Smith et al. 2000). As with basic taste receptors, there is a lack of an a priori rationale, a rational definition of what neuron groups are, and critical testing with control groups.

 

3.2.4. Summary.  For about a century the field of taste has rather uncritically been guided by the idea of four basic tastes. We went in with Hanig’s orientation, and then showed that there were indeed four basic tastes. It is not easy to estimate the degree of bias that this assumption may have brought to the field of taste, and the corollary degree of bypassing of data and theory to the contrary.

 

3.3. Hanig omitted control groups

 

Unfortunately, Hanig’s format did not provide the essential perspective of control tests demanded by science. If his findings were to have any relevance to the basic tastes model, it would have been necessary to learn if other tastants, for example KCl and CaCl2, proline and lysine, also had differentially distributed sensitivities across the tongue. If they were distributed in this way, the implications of his data for the basic tastes idea would have been nullified. No one has completed his research with the required control studies. Why not?

 

Since Hanig, there has been a routine omission of control groups in research using the basic tastes for its organization. This omission is notable in the three areas of research noted just above: psychophysics, and receptor and neural organization. The effect of omitting control groups is to support the basic tastes model, rather than to question it. This format is not helpful in science.

 

4. Extra-curricular support for Hanig

 

Two of our approaches to knowledge - our language and our experimental methods - have guided us toward Hanig’s position.

 

4.1. Language

 

4.1.1. Cognitive characteristics of humans. The limits of human cognition mentioned above (Miller 1956) give a logic to the idea that there would be just a few named tastes, up to six or seven, but not 20. If there were indeed these higher numbers of basic tastes of some sort, they might be too many for us to be aware of them - or even to know to look for them. The five rather than 105 ways of Chinese thought might have derived in part from this cognitive constraint. We cut up reality into a few words in order to understand and communicate. To follow such an arbitrary linguistic force in science is a problem worth study. 

 

4.1.2. Words as hypotheses? The taste words have served as scientific formats in that we follow them. The Eskimos have multiple words to communicate about snow. But these important words do not necessarily invoke a science about multiple kinds of snow. Analogously, our use of a few basic taste names does not indicate more than that we need such words to communicate about taste. While necessary for communication, there is no reason that these words should have served as meaningful hypotheses about the nature of taste.

 

4.1.3. Words versus reality. There is the very strong problem that we tend to believe in the reality and validity of our words. Our language is necessarily limited to discrete words that usually seem to have clear and straight-forward meanings. May we assume that the words “liberal,” “love,” “intelligence,” “clear,” “understand,”  “define,” or “reality” are clearly understood and defined, and are accurate representations of reality? And “salty” or “basic taste?” Our words are designed to serve us, but are not necessarily the best scientific representations of the nature of things. We may think that we get words from the dictionaries, but they are merely compendiums of the words we use, such as the taste words. Allowing scientific theory to be constrained or led by the terminology in common use would have made our current understanding of mathematics or physics impossible.

 

Many scientists have been critical of the influence of our words on our understanding.  Speaking about geological time periods, S. J. Gould (1994) discussed words used to invent categories thus: “The human mind seems to work as a categorizing device…  This deeply (perhaps innately) ingrained habit of thought causes us particular trouble when we need to analyze the many continua that form so conspicuous a part of our surrounding world.  Continua are rarely so smooth and gradual in their flux that we cannot specify certain points or episodes as decidedly more interesting, or more tumultuous in their rates of change, than the vast majority of moments along the sequence.  We therefore falsely choose these crucial episodes as boundaries for fixed categories, and we veil nature’s continuity in the wrappings of our mental habits…  We must also remember another insidious aspect of our tendency to divide continua into fixed categories.  These divisions are not neutral; they are established for definite purposes by partisans of particular viewpoints.”  The close correspondence between Gould’s comments and the probable continua in taste, including inflection points as the corners of Henning’s tetrahedron (1916) is of note. Henning described the world of gustation as a continuum of tastes in three dimensions - a tetrahedron, defined by “sweet, sour, salty and bitter” at its four corners.

 

Gould could have been writing about Aristotle, the great categorizer. In addition to his seven tastes, Aristotle wrote that there are five elements, earth, air, water, fire and aether; this was a good beginning as he identified the major forms of matter - solids, gases, liquids - and energy. Aether was the astronomical material. He even categorized categories into 10 kinds - such as substance, quality, quantity, relations etc.

 

Another critic of the uncritical belief in words, G. Kimble, wrote (1996) “Language is the agent of cognition, the currency of thinking, the tool-box of communication and the custodian of culture.  To be useful, it must map onto the world with some precision.  Unfortunately, however, the fact that it does so encourages the faith that the fit is perfect and that truth is in the dictionary: If there is a word for it, there must be a corresponding item of reality.  If there are two words, there must be two realities and they must be different.” 

 

William James (a foundational psychologist and brother of the author Henry James, who may have been a greater but less systematic psychologist than William) recognized the problem (1890):  “Whenever we have made a word…to denote a certain group of phenomena, we are prone to suppose a substantive entity existing beyond the phenomena. (And) the lack of a word quite often leads (to the idea) that no entity can be there…” The relevance of these comments for the naming of tastes is clear.

 

Proper concern has occasionally been voiced over the problematic adequacy of the “basic” words to represent tastes (Beidler 1974; Purves et al 2001). And eloquently presaging the words of Henning (1916), Brillat-Savarin (1886) pointed out that the four basic tastes “…are only a convenient way to talk about a much larger array of taste sensations.” 

 

It is interesting to note that Richard Feynman did not discover certain revolutionary aspects of quantum mechanics in terms of words, and then decide how to interpret the words. He just imagined the atom, and “watched” its behavior. But he did have to translate these discoveries into words - although they were clearer in graphs. We may well follow his direction in taste - to lead, not to follow, the words. In the field of physics it has been realized that our common words are not always reliable or sufficient, and may be misleading, so neutral words such as “quark” are sometimes invented.

 

Beyond these, it might be useful to consider that even our most basic and secure word-concepts, mass, time, distance and energy, are based on man’s conception of the nature of things.  They seem too good to be less than totally true in their general or particular usages, and quite real independent of our naming them.  But they are only our words, rubrics facilitating our understanding, and our communications (or miscommunications) with each other.  Our understanding of even these “basic” physical concepts are changing; they are not as simple and real as these distinct words make them seem.

In summary, it might be helpful to worry about allowing common usage words such as “sweet” and “sour” guide our research and understanding. These words could provide a problem rather than the solution.

4.1.4 Why these taste words?

Imagine the wordsmiths of ancient China developing a communication system for the tastes of things; what words would they choose? Which foods would they name? They might devise names for the good things to distinguish them from the bad: sweet versus sour or bitter. And names for that all-important substance, salt; wars have been fought over the possession of salt as a seasoning and as a preservative. These four would very probably be at the top of any culture’s list of taste words. Other words might be invented by culture, depending on the local foods available.

            In fact, words for the four “basics” appear to be found quite uniformly in all languages - evident in English as well as foreign dictionaries. These include at least Afrikaans, Arabic, Albanian, Chinese, Danish, Dutch, English, French, German, Greek, Hebrew, Hindi, Hungarian, Indonesian, Irish, Italian, Japanese, Latin, Maori, Nepali, Oromo, Papua, Persian, Polish, Portuguese, Russian, Samoan, Sanskrit, Scotch, Serbo-Croatian, Slovak, Spanish, Swahili, Swedish, Tagalog, Urdu, Vietnamese, and Yiddish. There are some exceptions, such as omissions of “salty,” even in the presence of the noun “salt” and the verb “to salt,” and sometimes confusions between “sour” and “bitter;” but these may be errors in the editing of the dictionaries. And, as in English, there are very many culturally idiosyncratic taste words beyond the basic four in each language. In Korean there are words for bland and spicy. In Basque there is one word for bitter/sour. The Indonesians have the word “pahit,” which has no translation in English. A thorough cross-cultural study of taste words and their meanings might prove useful.

Does this apparent ubiquity of usage of the four basic taste words indicate something very obviously true, four separate sensations that sum up all tastes? Or could they represent four specialized physiological process? These propositions lack testable rationales that are independent from the invention of words for communication. It might be doubted that scientists in each of these cultures independently proposed these words as scientific hypotheses.

 

4.2 Our methods fit Hanig’s orientation.

 

Of course, all of our knowledge comes in terms of what we can measure, and thus what we can measure can determine how we think. What we measure depends on our techniques of measurement. Thus the techniques of measurement that are available can set the arena into which our discourse falls. These techniques include the biological and mathematical methods at hand. Then the characteristics of our language bias our understanding of the data.

 

4.2.1. Biological techniques. All of our biological techniques have forced our thinking into a search for a neural structure for a named function; we can scarcely avoid taking this position (Erickson 1978). As examples of the techniques at hand, the earliest of the modern biological techniques were those of Gall and Flourens at the beginning of the 19th century (Erickson 2001). Gall, a comparative neuroanatomist, attempted to measure variations in the relative sizes of human brain areas to relate to the talents of these individuals. His contemporary, Flourens, ablated various parts of animal brains to assay the function of each. Because of the nature of their techniques, their answers were necessarily in terms of what “word-as-function” existed in which “word-as-structure.” Gall opted for a strict localization of many functions; math, language, sex, avarice, and memory as examples. Flourens won that argument for the time stating that, on the contrary, the brain was something of a bowl of relatively homogeneous soup. Both were probably correct in part, Gall presaging the labeled-line point of view, and Flourens the distributed models. Later, section of the dorsal and ventral roots showed that the former had a sensory function, and the latter were motor. Then electrical stimulation of the brain also dictated locations for functions, for example that the various muscles or movements were represented in different points in the brain. The electrical recording methods - gross and microelectrode - were, again, necessarily locationist in their conclusions. Brain imaging, borrowing a clinical tool that favors locationist data, shows that the various searched-for functions are relegated to differing named brain structures. But why not search for avarice or patriotism or fishing? Why expect worded functions to be in circumscribed and easily namable localized structures? Chemical tracing of neural pathways is also locationist with the pathways being the structures considered. In the labeled-line model, the basic tastes become the functions to be identified with the structures that are individual receptors or neurons.

 

It is probable that none of these new localizing techniques produced false or trivial data. But the proper interpretation of each species of data turned out to be more complicated than had been acknowledged. Does each new technique, now the molecular/genetic studies, show us a new reality? The position taken here is that any of these techniques might have given us a very advanced answer if we had known exactly what the question was. Otherwise the techniques may at least temporarily direct our ideas rather than vice versa. We cannot know the best technique, or even an appropriate one, until we know what we are looking for.

 

In physics on the other hand, techniques are continually being developed to meet the explicit issues raised by their formal hypotheses. Particle accelerators were invented to address very specific issues, topics such as the structure and mechanics of atoms and their components. Scientific physics, although not clearly older than the science of taste, might be considered more advanced.

 

4.2.2. Mathematical techniques. Various mathematical methods have been used to provide support for the basic tastes idea, rather than test it. Prominent among these has been multi-dimensional scaling, cluster analysis, and factor analysis. While these techniques can be used to suggest directions for further hypotheses and research (Uslander, in Kruskal and Wish 1978), they cannot prove or disprove the reality of any groupings, including those used to support the basic tastes model (Aldenderfer and Blashfield 1984; Sarle 1987; Sokal and Sneath 1963). This issue depends on a prior formal definition of what groups are (Erickson et al. 1993) that might then be tested by these mathematical techniques. To use these techniques to determine if groups exist is like asking a road map to tell us where we are going, which they cannot do, rather than to legitimately ask it about the lay of the land for our consideration.

 

The exemplary presentation for the use of multi-dimensional scaling uses linear distances between cities to recover a correct map; importantly, this provides no conclusions about groups of cities - such as the DC-Boston corridor. Such linear distances are also possible in handling taste data in multi-dimensional spaces (Erickson 1986, 1993; Schiffman and Erickson 1993). Factor analysis, although used to support the idea of groupings, is not relevant to this issue; this technique only indicates the number of factors underlying the data - for example three strong factors might be used to successfully represent data of no groupings, or of many groupings.

 

But correlations have been routinely used (e.g. Geran and Travers 2006; Gilbertson et al. 2001; Lemon and Smith 2005; Scott and Giza 2000; Smith and St. John 1999) instead of the linear distances between objects for which multi-dimensional scaling methods were devised. Correlations are indicative of non-linear degrees of relationships, incorrectly displacing the points into a “groupy” appearance towards the edges of the now non-linear space (Erickson et al. 1993). Correlations are completely inappropriate for multi-dimensional scaling, but still they are used to support the basic tastes model. This adds to the previous problem that these plots are incorrectly used to provide conclusions about groupings.

 

Even random data sets will always contain apparent structure, appearing as grouped in both multi-dimensional scaling and cluster analyses whether based on linear distances or correlations (Erickson et al. 1993). Indeed, the published graphs purporting to establish the presence of groups with these methods could just as well be representing random or regularly arranged “checkerboard” data. These groupings are only in the eye of the beholder.

 

In summary, basic taste “groupings” or “typologies” derived from these mathematical methods are completely unfounded (Erickson et al. 1993), but have been used as key support for the basic tastes doctrine. Why?

 

5. Following Hanig has been problematic.

A variety of problems attendant on following Hanig’s lead are evident, but ignored.

 

5.1. Lack of controls

 

As mentioned in Section 3.3, Hanig set the stage for omitting control groups. This, in itself, disqualifies most data since then used to support the basic tastes position.

 

5.2. Undefended assertions

 

There is a general uncritical acceptance of the reality of basic tastes (Section 3.2.1). The many textbook statements that there are four basic tastes do not define what a basic taste is. They do not discuss the lack of the necessary underlying testable hypotheses or the absence of controls, although ad hoc reasons - such as control of food intake - may be offered in support of the idea. This procedure of only supporting the basic tastes position has become routine (Amrein and Bray 2003; McBurney 1974, 1978; McBurney and Gent 1979; Scott 2004) instead of fairly testing it. When both sides of this issue are represented, the lack of support for the basic tastes position becomes evident (Schiffman and Erickson 1980).

 

By contrast, in the field of physics, some textbooks, and certainly discussions of the history of physics, point out where the idea of atoms came from, what data support that idea, how falsifications were attempted, and in what ways it is a heuristic idea. Could that be done in taste? Probably not with the present structure in this field.

 

5.3. Data are ignored

 In order to follow Hanig, much data and logic contrary to the basic tastes position have had to be jettisoned.

5.3.1. Receptors and neurons. In studies supporting the few basic tastes position, the data which falsify that position are often not addressed. As a recent example, data from molecular studies define several receptor mechanisms. Conclusions were drawn that taste receptor cells and neurons are not broadly sensitive to a variety of tastants (Chandrashekar et al. 2006; Zhang et al. 2003; Zhao et al 2003). Without denying the data about taste receptors, the conclusion of narrow tuning is clearly falsified. That receptor cells and neurons are broadly responsive constitutes one of the clearest and most substantial arrays of data in taste stretching over 50 years (e.g. Brasser et al. 2005; Caicedo and Roper 2001; Caicedo et al 2002; Dahl et al. 1997; Erickson 1963, 2000, 2001; Gilbertson et al. 2001; Herness 2000; Lemon and Smith 2005, 2006; Pfaffmann 1941, 1955; Rolls 2005; Stapelton et al 2006). In an important analogy, color vision receptors may be described in terms of single photo-chemicals, but they each respond to very broad arrays of wavelengths; likewise, that chemoreceptor mechanisms exist does not indicate that the receptors and neurons each respond only to one taste. It would be useful if these molecular data were rationalized with what is known about taste. When such important and relevant data are ignored (e. g. Amrein and Bray 2003; Scott 2004), the analysis in support of basic tastes does not fail - and perhaps cannot fail.

 

One issue should remain clear. The molecular data show that there are at least two types of taste receptors, and this is to be expected. The breadth of responsiveness of receptor cells and taste neurons is a different issue.

 

5.3.2. Psychophysics: How many tastes? That there are only four tastes, and that these four will completely account for all other tastes, has been stipulated as the core assertions of the basic taste model. This model, seldom defined, is perhaps given its most articulate presentation by Halpern (2002a). There may be no more, or other, than these four tastes. But if these untested assumptions are removed, any taste has equivalent opportunity to serve as a basic. In English there are at least 31 words used to describe tastes without resorting to non-taste adjectives (www.yoakumisd.net/junior_high/bartosh/bartosh10.html). These words are all potential candidates for basic tastes in that several of them might account for other tastants as efficiently as do the basics. But such tests with other than the basics have not been performed. A preliminary example showing how such tests could be done is presented in Section 7.

 

And there is evidence that there are other than the four basic tastes, or that the four basics will not account for all tastes. Umami is the foremost example. Metallic (Critchtey and Rolls 1996; Lawless et al 2005) and fat (Chale-Rush et al 2007; Gilbertson et al 1997; Mattes 2005; McCormack et al (2006) are possible candidates as tastes separate from the basics. However, due to the lack of a definition of what a taste is, fat may serve only as a flavorant (Pittman et al 2006) for the reasons given by Halpern for umami (2002b). That water might be a separate taste has a long history, now with some empirical support (Gilbertson 2002). Di Lorenzo et al. (1986) showed that in rats a conditioned taste aversion to EtOH did not generalize to any of the basic tastants or combinations of any two of these, except that it did generalize to a mixture of sucrose and QHCl; thus EtOH appears to be a separate non-basic taste, and the taste of a sucrose/quinine mixture is not clearly reducible to the basics. If a continuation of such findings are readily absorbed into the basic tastes model, it will be a negation of its major raison d’etre - that there is a defined and limited sets of basic tastes.

 

Spurious suggestions for new basic tastes might occur due to the lack of a clear definition of what a basic taste is. To be noted here is Halpern’s cogent argument that umami is not a taste, but a flavorant (2002b) importantly affecting the tastes of other stimuli. Although MSG has a taste - disagreeable and not clearly a combination of the basics - umami is a flavor that only appears when small amounts of MSG are added to foods. What exactly is a taste, or even a basic taste?

 

It should also be noted that although Boring (1942) did not touch on the lack of a definition for basic tastes, he did correctly comment that Hanig’s data did not limit the number of tastes, for example to four.

                                                                                                   

5.4. Logic is ignored.  

 

The primary logical fault with the basic tastes position is that - as previously mentioned - it does not have a rational and testable definition. We don’t know what “basic tastes” are, but that does not appear to be a topic of concern. This fault runs through the following discussion.

 

In addition to not addressing contrary data, papers supporting the basic taste idea do not address the logical strength of coding in populations (e.g. the across-fiber pattern theory) and the corresponding logical weaknesses of the labeled-line basic tastes position (Erickson 1968, 1977, 1982a, 1982b, 1984, 1985a, 2000, 2001; Erickson et al. 1990; Erickson and Covey 1980; Ishii and O’Mahoney 1987; O’Mahoney and Ishii 1987; Schiffman and Erickson 1980, 1993). It is to be noted next that there is a general acceptance of some form of population coding throughout the field of neurobiology (see Erickson 2000, 2001).

 

The evidence for receptor or neuron types is held as a falsification of, or at least awkward for, across-fiber patterning (Scott and Giza 2000). However the original presentation of across-fiber patterning was in terms of a few receptor and neuron types (Erickson 1963). This model requires receptor types in non-topographic sensory systems such as color vision and taste (Erickson 1968). It is odd that groupings have been used as an argument against this theory (Smith and Scott 2001).

 

The basic tastes assumption has provided an illogical format for neuron categorization. The fact that each taste neuron or receptor is more sensitive to one of the basic tastants than others (Frank 1974, 2000) is logically inescapable; how could a neuron not be more sensitive to one tastant than others? And this must be as true for any other set of tastants as for the four basics, qualifying any set of four stimuli as basics; but, again, these control studies of trying other tastants are absent. And in Frank’s model the side-band responses to the non-“best” stimuli are considered to be noise, or the tastes of the basic stimuli must be assumed to be a combination of several tastes (Bartoshuk 1988) - both unparsimonious ad hoc assumptions apparently devised to support the basic tastes model. The fact that larger amounts of data, from more than the four basic stimuli, can be expressed along a continuum in the same form as Frank’s plots (Erickson 1967; Scott and Erickson 1971) is not addressed in the basic tastes model.

 

Although both the rationale and control experiments are missing, Frank’s data have been cited as strong support for the idea of labeled lines, and thus basic tastes. Other methods of obtaining data in support of neuron types or groupings have been undertaken, also without independent rationale and definition (e.g. Smith et al. 2000; St. John and Smith 1999).

 

The similarity of the specific taste receptor data to the characteristics of color vision provides another area for an unbiased critique of the basic taste model. Adaptation, or the genetic absence, of the long-wavelength visual receptors, results in a strong diminution or absence of responsiveness to a range of “reds,” analogous to the taste effects seen in mice missing a taste receptor (Zhang et al 2003; Zhao et al. 2003). In this situation, responses of cells to other wavelengths may vary from not affected to very affected depending on their wavelength. The across-fiber pattern model would accommodate increases in saturation in the perception of shorter wavelength stimuli along with decreases in saturation in longer wavelengths. Similar effects in taste are evident, and not surprising. Resolution of this issue in taste would require close attention to the responses to systematically varied stimuli other than the basics (analogous to variations in wavelength), and a rationalization of the conclusions of specificity in taste with the contradictory historical findings.

 

The assumption that there are only four tastes has a strong but illogical effect on psychophysical data collection (Erickson et al 1990). The finding of “suppression” between tastes (Bartoshuk 1988) required the prior assumption of the existence of a few perceptually distinct basic tastes that cannot each be modified in mixtures. In that idea, only the basics should be sensed, so only changes in the strength of the basic tastes in the mixtures could be accepted as data. Any change in taste quality, such as synthesis of different tastes (Erickson et al 1990; Schiffman and Erickson 1980), was not admissible. If the subjects are only given the basics for use, then they can respond only in terms of the basics. But given other standards for use, they possibly would have used those - there was an omission of this key control group. But if the doctrine of basic tastes is not assumed, then mixtures of “basic” or other tastants could be understood to produce a changed across-fiber pattern with some similarity to, or correlation with, each of the components. A new and intermediate taste, no longer a basic taste, should be sensed, usually similar to both components, but not just suppressed versions of basics. For perspective, if the descriptions of colors are limited to only a few basic terms or color samples, then the ratings for colors intermediate to these “basic” are reduced (Erickson 1977); these could be inappropriately considered “suppressed.”

 

So in addition to the prior assumption of basic tastes, the idea of suppression requires an additional assumption that these tastes are somehow related while also being distinct and separate. No additional assumptions are required in the across-fiber model as the continuum requires that tastes are related. Parsimony has proven useful in science.

 

The lack of linking hypotheses is one of the most serious of the problems with the logic of basic tastes. This model encompasses four unrelated events at various levels of meaning. First, it holds that only four tastes are perceivable. This is extended without explanation to the neural level where four different sets of sensory neurons each exclusively represent one of these tastes.  It is also held for the receptors, and then receptor mechanisms, as well as for stimuli, without articulating the relationships between these levels. 

 

5.5. Convenience

 

It may be that the most generally persuasive drawing power of the basic tastes model is that without a formal restraining definition this model is very convenient in designing experiments, and it easily categorizes any data - especially from experiments cast in the four tastes format. The design is thus very straight-forward and is supported by our words and techniques. And this procedure, as Hanig’s, seems to validate the basic tastes idea without test. Arguments in favor of the basics position flow easily from this context, encouraging a lack of attention to contrary matters of data and logic. It is certainly hard to think of a simple model different from that provided by the basic tastes to guide our research and understanding - and against which the basic model might finally be tested. It is improbable that an ultimately successful science was ever built on convenience.

 

5.6. Formality and science

 

In the remainder of this paper the term “formal” will refer to the use of data-based, rational and heuristic hypotheses that provide clear enough definitions of their terms to be potentially falsifiable. Hanig’s work does not fit into this definition. The core orientation herein is that formal definitions are required for taste to qualify as a science just as they are in physics. One aspect of this process is that it provides the essential groundwork for control experiments as illustrated in Section 7. But first some previous experiments with controls will be discussed.

 

6. Previous control tests of taste models    

 

The major tenet of this paper is that models of taste must be testable, which means potentially falsifiable. There have been a few unbiased tests of differential predictions between the basic tastes and across-fiber pattern models: the topics are the required singularity of basic tastes, the presence of tastes other than the basics, taste as a continuum rather than four separate tastes, and the possibility of quantification. The findings uniformly falsified the basic taste model and/or supported the across-fiber pattern model. A brief discussion of these will lead to an illustrated proposal (Section 7) for a test of more generally accepted aspects of the basic taste model, the ability held for only the basics to account for other tastes, and that only the basics are perceived in other tastes (Halpern 2002a).

 

6.1. Singularity of tastes

 

If there are only four (or a few more) basic tastes, then each should be sensed as being singular since each is not more than one taste (Erickson 2000; Erickson and Covey 1980; Erickson et al. 1990; O’Mahoney et al. 1983). On the other hand, if each stimulus that was not basic were composed of several of the basics, it should each be perceived as “more-than-one” taste. The across-fiber pattern model indicates that any tastant, basic or non-basic, might be sensed as singular (Erickson 2000; Erickson and Covey 1980; Erickson et al. 1990; O’Mahoney et al. 1983). This is testable. Previous studies (Erickson and Covey 1980) have demonstrated that the basic tastes are judged as only relatively singular (by around 80% of the subjects). This research also showed that the non-basics are somewhat less singular than the basics (about 50% of the subjects reporting singularity), but not clearly “more-than-one” as would be predicted in the basic tastes model.

 

            And certainly mixtures of tastes should contain more than one basic taste, and thus be rated as more-than-one. However, it was shown that mixtures of several tastants can be perceived as being relatively singular (Erickson 1982b; Erickson and Covey 1980; O’Mahoney et al. 1983; Schiffman and Erickson 1980). Indeed mixtures of two basics can appear to be as singular as the individual basic components of that mixture; for example, the singularity of mixtures of HCl with either NaCl or Q2SO4 were equivalent to the singularity of those basics when tasted alone.

 

            As controls on these methods (Erickson 1982b), individual tones were routinely rated as one stimulus, and multiple tones were rated as more-than-one; this complexity increased with the number of tones in the “mixture.” That is, the tonal stimulus appeared to be “analyzed” into its components, all components remaining clear and unchanged as predicted for tastes in the basic model. However, increasing the number of components of a taste mixture does not show this increase in complexity. Instead of being analyzed into its components, a new somewhat singular but non-basic taste was “synthesized” in mixtures of two, three or four components. Like taste, a non-basic color resulting from the combination of two “basic” colors (Erickson 1977; Erickson and Covey 1980), such as a blue-green, was also rated as singular - this being an example of the synthesis of a sensation different from the either of the “basic” components (Erickson et al 1990). “Analysis” is predicted for topographic systems such as auditory tones, and location on the skin or retina. “Synthesis” is predicted for the non-topographic systems such as color, taste, the vestibular sense, kinesthesis, and olfaction. These unbiased methods could have falsified the across-fiber pattern theory, but instead supported it and falsified this aspect of the basic tastes position. 

 

            On the other hand, and in partial support of the basic tastes model, mixtures of sucrose or fructose with any one of the other three basic stimuli were rated as strongly less singular (around 39% to 64%) than either of the components alone (about 85% each for the sugars), suggesting that these sensations do indeed remain somewhat independent and singular. Also, when these sugars were mixed with any non-basic stimulus, the combinations were rated as less singular than either component, an indication of semi-independence. Also, quinine behaved somewhat like the sugars in that, in mixtures with sucrose or NaCl, the singularity of the mixture was greatly reduced - as it should be in the basic tastes doctrine; however the combination of QHCl with HCl remained singular. These data are suggestive of the receptor specificity claimed for sweet and bitter (Zhang et al. 2003). It was concluded that although NaCl, HCl and QHCl were interrelated, sucrose might well stand further apart - perhaps not being on the same dimension as the others (Erickson and Covey 1980). That sucrose and quinine might lie outside the continua of the other tastants had been suggested (Erickson et al, 1965). Further control studies could provide clarity here.

 

6.2. Tastes other than the basics

 

The singularity of non-basic tastes and mixtures, noted just above, is indicative of the synthesis of a different taste, this taste necessarily being something other than and in addition to the basics (Erickson 1977; Erickson et al 1990; O’Mahoney et al. 1983; Schiffman and Erickson 1980).

 

The question of whether the basic tastes are the only tastes perceived in mixtures, as held in the basic tastes model, has also been tested (Erickson 1982b; Erickson et al 1990). Subjects very rarely fail when asked to identify one of the four basic tastes alone. But when mixed with another basic taste, the failure rate in identifying a stimulus increased dramatically, indicating that that basic was not clearly and separately experienced in the mixture. In a control study, a simultaneous “mixture” tone did not generate this confusion in the identification of a given tone.

 

These findings together suggest that there are more than just the four basic tastes.

           

6.3. A continuum of tastes, not just four

 

As proposed by Henning (1916), many tastes comprise a continuum in the shape of a tetrahedron defined at its four corners by sweet, sour, salty and bitter; this continuum is explicitly composed of many singular tastes, not just those tastes at the four corners. But the four corner-tastes were later held to be the only tastes possible, and called the “basics” (Bartoshuk 1988). Henning’s continuum is a clear contradiction of the basic tastes position, while being a core aspect of the across-fiber pattern theory.

 

That multi-dimensional scaling solutions are possible indicates that the spaces are continuous. If the data were unrelated, solutions of low dimensionality would be impossible. This invalidates the basic tastes idea that the few tastes are independent of each other. Inter-relatedness along a continuum is a core part of the across-fiber pattern model.

 

That taste is probably organized as a continuum also shows up in reports of neural activity. For example, Frank’s data (1974, 2000) on the responses of individual neurons to the various basic tastes derives from a continuous organization presented as her baseline (Woolston and Erickson 1979); she demonstrates that neurons will show decreasing responsiveness with increasing distance from the “best stimulus,” a “neural response function” along these continua (Erickson 1967, 1968, 1974, 1978, 1982, 1984, 1985, 1986; Doetsch and Erickson 1970; Doetsch et al 1969; Erickson et al 1965).  This is contrary to the basic tastes position and in accord with the across-fiber pattern theory (Erickson 1967; Erickson et al 1965; Scott and Erickson 1970). But her data have been accepted as strong support for the basic tastes position instead. Why?

 

In this basic tastes orientation, the fact that many of these neurons respond best to stimuli other than the basics is not considered since the topic is restricted to basic stimuli. The desired results are in the main mathematically assured once the best stimulus of these four is found; that the other three stimuli are less effective is thus not surprising. However this relatedness may be extended to provide the same coherent organization for at least 16 stimuli across a continuum when the requirement is dropped that only the four basic stimuli be considered (Doetsch and Erickson 1970; Erickson 1965, 1967, 1968; Erickson et al. 1965; Scott and Erickson 1971). Plots similar in form to Frank’s were obtained, but for many stimuli. These are the smooth, bell-shaped neural response functions that describe the responses of individual neurons along the taste continua, as they appear for color vision receptors and neurons. These much more extensive and powerful neural response functions are neither expected nor checked for in the quadripartite logic.

 

It is important to note that even Hanig’s data are what would be expected from a continuum (Schiffman and Erickson 1980), as would Frank’s groupings (Schiffman et al 1979; Woolston and Erickson 1979).

 

6.4. Quantification of neural information

 

There has been no definitive attempt to quantify taste information in the quadripartite model. On the other hand, the across-fiber pattern theory has provided a testable method to quantify the differences or similarities between any stimuli, including tastants. The general idea for the representation of the differences in quality and intensity for any two discriminable tastants is that they must provide neural responses that are somehow different from each other. This is the meaning of the Chinese Yin/Yang differentiation required for our understanding of any concept; to exist, “sweetness” must have another taste different from it. People with only the “red” receptors can see no colors at all. So as the most bias-free measure (no prior assumption of what constitutes the difference, such as groupings) the different across-fiber patterns set up by two tastants are superimposed, and the absolute difference between them, say in impulses/second, is measured.

 

The quantification given in the across-fiber pattern model may be best presented first by example, and then formalized. The across-fiber patterns shown by Lemon and Smith (2005) provide the first demonstration. In their Fig. 3, the basis for the quantitative differences between two stimuli is given by the differences between the pairs of curves. The difference between sucrose and fructose is small, as would be the difference between KCl and HCl. The difference between HCl and sucrose is large. As absolute differences in impulses/time, these would be linear distances, and could be appropriately used for multi-dimensional scaling. Other examples that could be analyzed this way may be seen in the across-fiber patterns presented by Boughter and Smith (1998), Boughter et al. (1999), Dahl et al. (1997), Schiffman et al. (1979), Smith and Li (1998, 2000), and Smith et al. Fig. 2 (2000). This approach has been found to be a useful heuristic (Dahl et al 1997; Erickson 1982a, 1986; Schiffman et al 1979).

 

Stated in general terms, the total amount of activity set up by a given stimulus, here a tastant, is its neural mass. These are the areas under each across-fiber pattern curve discussed above. (Lashley [1931] noted importance of  “neural mass” in the handling of high levels of information of importance to a species.) The absolute difference in impulses/second (or some other measure) between this profile and the profile set up by a second stimulus is called the neural mass difference (Erickson 1986). In the across-fiber pattern theory, the prediction is made that the perceived differences between two stimuli is a positive function of the amount of this neural mass difference. As an example, two sodium salts of similar concentration would produce small neural mass differences, and would thus be predicted to be poorly discriminable from each other. 

 

The total neural mass produced by a stimulus is taken to be indicative of intensity. If the neural masses produced by two stimuli are different, that indicates a difference in intensity. Both intensity and quality give neural mass differences information about discrimination between stimuli. If it is desired to inspect only taste quality differences, leaving out intensity differences, the neural masses - representing intensity - can be brought to the same level by multiplication through one of them.

 

As examples, just noticeable differences in the color of a stimulus, or the location of a visual or somesthetic stimulus, would require the same neural mass differences. In the topographic systems (Erickson 1968), such as visual or somesthetic location, very many neurons are available; therefore they may be narrowly tuned and still evoke enough neural mass to differentially define the event. For usable events in non-topographic systems such as taste or color, few neurons are available, so to produce sufficient neural masses with their concomitant neural mass differences, they must be broadly tuned. This broad tuning would bring many neurons into play to provide the necessary neural mass (Erickson 1968, Figures 2 and 3).

 

7. A more definitive control?

 

7.1. A proposed control experiment

 

A model might be evaluated by the degrees to which it can encompass and predict large amounts of useful data. Such tests, described in Section 6 above, appear to controvert the basic tastes doctrine. An even stronger test would be to inspect the standard text-book definition of the basic tastes; are the “basics” the only ones that can account for other tastes (Pfaffmann 1951, 1954), or even are they more efficient at this than other tastants or words? And are the basic tastes the only tastes perceivable in other tastants and mixtures (Bartoshuk 1988; Halpern 2002a)? The following sketch, an example extended from the control formats in Section 6, is presented as a tentative illustration of how such an experiment might be done, and the rich vein of information to be expected from such experiments.

 

7.2. Introduction to the illustration

 

The key aspect of this demonstration was that control groups composed of tastants not considered to be basics were included to compare with the performance of the basic tastes. If the few-basic tastes position is correct, then the experimental basic words and stimuli should account very strongly or completely for various other taste stimuli, and the four controls using non-basic stimuli should completely fail, or nearly so. If there were no clear differences between the experimental and control groups in effectiveness in accounting for other stimuli, the basic tastes position in this definition would be considered falsified - and this is how the results came out.

 

7.3. Materials and Methods

 

Two experimental groups of subjects used the basic taste words and stimuli as “Standards” to account for the “Comparison” stimuli (Fig. 1, panels E1 and E2), and four control groups used from two to five “Standard” stimuli not considered to be basic (Fig. 1, panels C1 to C4) to account for the same Comparison stimuli.

 

The subjects were Duke University undergraduate students earning experimental credit in a course. Each subject participated in only one of the six groups. There were 31 subjects in the basic words (Fig. 1, E1) and 26 in the basic stimuli (E2) groups, and 22, 18, 12 and 13 in the 2, 3, 4, and 5 non-basic controls (C1-4). Using these Standards, each subject was asked to account for the tastes of the 10 Comparison stimuli (Fig. 1 abscissae). To avoid biasing the subjects towards use of the basic tastes model, it was expressed only in the first control (E1) in which the subjects were asked, but not taught, to use the words sweet, sour, salty and bitter as Standards. The sip and spit method was used with partially filled 3 oz. paper cups, and distilled water rinses. The subjects were instructed to taste each Standard and Comparison stimulus - with rinses - as many times before and during the experiment as necessary to remain familiar with them. They recorded the percent to which each Standard word or tastant accounted for the tastes of the Comparison stimuli.

The stimuli were 0.002 M quinine HCl (QHCl); 0.0025 M sodium saccharine (Na-sacch); 0.01 M monosodium glutamate (MSG); 0.03 M HCl and l-phenylalanine (phen); 0.05 M l-tryptophan (trypt); 0.1 M l-lysine, n-acetylglycine (ace-gly) and NH4Cl; 0.2 M NaCl; 0.3 M KCl; and 0.6 M sucrose and l-proline.

 

7.4. Main Results

 

The main issue was to compare the performance of the two experimental basic taste groups (Fig. 1, panels E1-2) with the four control non-basic taste groups (C1-4). Contrary to the basic tastes idea, the data for the control groups overlap those of the experimental groups.

As examples of the data, the subjects rated the taste of NaCl (Fig. 1, E1, fourth column) as 0.3% sweet, 75.6% salty, 2.8% sour and 5.8% bitter for a total of 84.5%, leaving 15.5% of its taste not accounted for.  On the other hand, the four basic taste stimuli (E2) accounted for a total of 93.7% of the taste of NaCl.

 

Averaged over all the Comparison stimuli, the basic words (Fig. 1, E1) were relatively ineffective as Standards, accounting for 69% of the Comparison stimuli, and the basic stimuli (Fig. 1, E2) were relatively effective, accounting for 84%. Although the non-basic Standards should not have accounted for the Comparison stimuli, they were competitive with the basics; the 2 non-basic stimuli accounted for 62% of the Comparison stimuli (Fig. 1 C1), the 3 non-basics for 81% (C2), the 4 non-basics for 73% (C3), and the 5 non-basics for 78% (C4).

 

To be noted is the overlap in effectiveness between the basic (E1-2) and non-basic (C1-4) Standards, and the general increase in effectiveness as the number of non-basic Standards was increased from two (C1) to five (C4).

 

Individual subjects occasionally accounted for more than 100% of a Comparison stimulus. For example, one subject rated KCl as 30% accounted for by lysine, 50% by tryptophan, and 50% by phenylalanine (total = 130%), and another rated MSG as 30% lysine, 10% tryptophan and 80% phenylalanine (120%). These aspects of the data are hidden within the averaged data in Fig. 1, but were strong enough to appear in the averaged data as 109% of HCl accounted for in Fig. 1, C 4. This is discussed further in “A Completion Effect,” Section 7.8

 

 

Fig. 1.  Accounting for 10 Comparison taste stimuli with basic and non-basic Standard tastes; the experimental design and results. In the 6 panels, the first two, E 1-2, give the results from the two experimental “basic tastes” groups, and C 1-4 each give the results from one of the four “non-basic tastes” control groups. In panels E 1 and E 2 the four basic words (sweet, sour, salty and bitter) and stimuli (QHCl, HCl, NaCl and sucrose) were used as Standards to describe the tastes of the 10 Comparison stimuli. In the rest of the panels (C 1-4), from two to five stimuli not considered to be basic were used in the same role as Standards. 

                            

7.5. This psychophysical definition of basic tastes was falsified.

 

The definition that basic tastes (Fig. 1, E 1-2) will account for the tastes of a variety of taste stimuli, and that non-basic controls (C 1-4) cannot, was falsified; each of the experimental and control groups accounted for considerable and overlapping proportions of the Comparison stimuli. This is illustrated in Fig. 1 in that the columns have substantial height in both the experimental and control groups; no basic group (E1-2) achieves 100% in accounting for any Comparison stimulus, and no non-basic group (C 1-4) completely fails with columns at 0%. For this finding of broad overlap no statistical assay was needed. If a rational statistical definition of basic tastes were offered, for example if there were given a reason why basic tastes should do better than non-basics at p < .01, or that the basics should be at 100% and the non-basics at 0%, those predictions might be testable. At any rate, it is doubtful that the basic words could be found more efficient than the 3, 4 or 5 non-basic stimuli, as the latter have higher raw data averages.

The present format is different from Hang’s (1901), and those in other psychophysical papers supporting this model that followed him, in that it used controls, and could have supported or falsified the basic tastes model.

7.6. An inconvenient truth.

 

After the experiments it became disconcertingly clear that the subjects were biased towards the idea that the four basic tastes should account for the Comparison stimuli! We failed in our efforts to keep this issue balanced between the experimental and control groups, and so the data are probably skewed in favor of basic tastes. Many of the subjects commented on this issue in debriefing, such as “Aren’t there just four tastes?” They reported that they had learned from textbooks, in middle school through college, that there are four basic tastes. This unexamined bias may be unavoidable in all psychophysical studies involving the basic tastes. The future researchers in the field of taste will be drawn from such students.

 

This bias could account in part for the use of basic words and stimuli herein, and the relative singularity of the basics, and would raise the issue of whether the use of basic words is learned, or a sensory verity. These questions might be approached experimentally, but the cultural “common truth” of basic tastes would be hard to null out.

 

7.7. Further tests of the basic tastes model

 

7.7.1. A definition of “basic tastes.” To test the four tastes model in somewhat more detail, as clear a statement as possible of the idea was sought. Two relevant tenets of the model at the psychophysical level, as cogently summarized by Halpern (2002a), are a) that the four basic tastes completely account for all tastes, and b) that the basics are clearly and separately perceivable, unchanged, as the only components of mixtures and other stimuli. (He also points out that in this model each receptor and neuron is specifically tied to only one of these tastes). It is clear from his paper that the terms “taste” and “basic tastes” have not been formalized beyond functions a) and b), and no control experiments have been done to test other words or stimuli as basics, other numbers of nominee basic tastes, or whether they are indeed clearly discernable in mixtures or other stimuli. To the extent that Halpern’s two statements might be taken as possible tests of the model, it is further falsified as follows:

 

7.7.2. The accounting was not total. Halpern’s tenet a). First, the basic tastes did not completely account for the Comparison stimuli. But must the basic tastes be held to exactly a 100% accounting of other tastes in the present format? Would 95% do? It could easily be that there is an inherent lack of clarity of tastes, say in comparison with our abilities to report that a shape is a square, a straight line or a point, that would somewhat degrade the subjects’ acuity in tasting. Perhaps the performance of the basic tastes herein is the best the subjects could be expected to do, and that these data are equivalent to a perfect accounting. But the non-basic Standards accounted for the Comparison stimuli at about the same level as the basics; either way, the control and experimental groups are roughly equivalent in this limited set of data.

 

7.7.3.   Tastes other than the basics were perceived. Halpern’s tenet b). The psychophysical basic tastes model holds that the basic tastes are clearly and separately perceivable, unchanged, as the only components of other taste stimuli and mixtures. This did not hold in the present data. After the experiment proper the subjects in the four and five non-basic controls (Fig. 1, C 3-4) were asked to describe their non-basic Standard stimuli freely, that is without the instructions to use the basic words. In the basic tastes model, only the basic words should have been used - but only 65 % of the words used were basic, and 35 % were not. These non-basic words included alcohol, alkaline, aspirin, bad, bad soda, baking soda, bland, cottony, chalk, citrus, crap, dentist work, fresh, ginger ale, medicine, metally, milky, mineral, mouth wash, nasty, not good, Nutra Sweet, old, pastry, plastic, spicy, Sweet ‘n Low, and tangy. Acid, lemon, and vinegar were counted as basic words for sour. The fact that words other than the basics can account for tastes has been shown (Schiffman and Erickson 1971; Schiffman and Dakis 1975).  A more fulsome treatment of this issue would be interesting, perhaps using some non-basic words; and such a procedure would provide a reasonable control for the basic words protocol (Figure 1, E1). The present data suggest that the basic tastes do not come naturally as the only verbal descriptors subjects find appropriate; and this in turn suggests that their exclusive use in psychophysical studies may have resulted, to some degree, from directions to use only these words, guaranteeing a successful outcome in terms of basic tastes.

 

7.7.4. Inconstant accountings. And, as was true in describing the non-basic Standard stimuli with words (7.7.2), the Comparison stimuli did not appear to be perceptually made up of the basic stimuli.  The basic tastes model requires that the non-basic stimuli should be completely describable in terms of the basic stimuli (Halpern’s tenet a). For example, the taste of proline should be composed of the basic tastes - mostly sweet and bitter (Figure 1, E1, column 3). Further, the amounts of each basic taste in proline should be definite to unequivocally define the taste of proline (Halpern’s tenet b). If this is so, then accountings for Comparison stimuli by proline should also be definite, or fixed, because it represents fixed amounts of the basic stimuli. When used only with lysine, proline accounts for an average of 33% of each Comparison stimulus (Fig. 1, C3). If proline is composed of definite and fixed amounts of the basics, it should continue at 33% no matter how many Standards are used along with it. But this accounting drops progressively from 33% when there are only two Standards to 24%, 20% and 20% of each Comparison stimulus when more Standards were added (not illustrated independently, but discussed next with reference to Fig. 2). Lysine also drops progressively in this series from an average of 28% to 30%, 25% and 16% when used with one to four other Standard stimuli (Fig. 1, C1-4). So proline and lysine, as examples, are evidently not perceptually composed of definite amounts of the basic tastes as required in the basic tastes doctrine.

 

A similar effect might be found using various numbers of the five basic stimuli. For example the degree to which NaCl accounts for another stimulus or mixture might depend on whether it was used alone, or along with one to four other basics. If the accountings by NaCl decreased in this series, both of the basic taste’s primary definitions (Halpern’s tenets a and b) would be negated. Such control experiments have not been done.

 

7.8. A “completion” effect

 

This failure to find fixed amounts of accounting for the Comparison stimuli by individual Standards is not in line with the basic tastes model. A reason for this failure may be provided by a look at the overall averages for the non-basic Standards shown in Fig. 2. The average decreasing effectiveness of each of the individual Standards as the number of Standards increases from 2 to 5, appears in the lower curve. On average, when two Standards were used they each accounted for about 31% of the taste of the Comparison stimuli; but with three Standards this decreased to 26% each, to 18% each with four, and to 16% for each of five Standards. But this curve should be level if each non-basic Standard were indeed made up of definite and fixed amounts of four basic tastes as held in the basic tastes doctrine.

 

 

Fig. 2. The degree to which each non-basic Standard stimulus accounted for the Comparison stimuli depends on how many Standards were used (from Fig. 1 C 1-4). The lower curve shows that each Standard’s account was not constant but decreased (averages) as the number of Standards increased. The upper curve shows that the total amount accounted for by all Standards may be approaching an asymptote as the number of Standards increases. The upper curve equals the lower curve times the number of Standards.

 

The reason for this steady decline (lower curve, Fig. 2) may be seen in the totals accounted for with each set of Standards (top curve). The implication is that these totals may be rising towards “completion,” which would be close to 100%. With only two Standards the total amount of a Comparison stimulus accounted for is 62%; this rises to 78% with three Standards, 72% with four, and 80% with five Standards. It might be concluded that completely accounting for the taste of a Comparison stimulus was an implicit objective for the subjects. To do this, they would have to reduce the value given for each Standard when there were more of them; otherwise the total could quickly rise to more than 100%. If each Standard were held to the 31% seen for two Standards, then when 5 standards were used, 155% of the Comparison stimulus would be accounted for. This might be perceived by the subjects as an excess to be avoided, a potentially strong effect in previous research of this nature. Those controls were missing.

 

In summary, subjects in psychophysical procedures may take on two conflicting goals; to veridically report how well each Standard accounts for a Comparison stimulus, and to completely account for the Comparison stimulus with whatever Standards are available. The direction the subjects take in these kinds of experiments might depend heavily on the nature of the instructions given to them. Since the subjects were clearly aware of the idea and role of basic tastes before the experiment, their reports could be biased towards the completion usually required in studies of this sort (Section 7.6). Support for this idea is seen in that subjects very rarely go over 100% in accounting for a Comparison stimulus (Figure 1 C4, HCl). Further discussion and testable speculations on the “completion” effect are contained in Appendix I.

 

This problem may be summarized thus: If the subjects believe in four (or five etc.) basic tastes, and are given this number of stimuli to use, then they might for this reason complete the accountings with four (or five etc.) stimuli.

 

7.9. Singularity, or what is A taste?

 

As found in previous tests of singularity (Section 6.1), neither were the basics singular (one taste) nor the non-basics more-than-one (of the basic tastes) as required in the basic tastes doctrine. This was confirmed in the present data.

 

The perceived singularity of each of the 10 Comparison stimuli was determined in the basic words and stimuli controls (Fig. 1, E1 and E2) after the experiment proper. It is seen in Fig. 3 that only 70% of the subjects found the basics NaCl, QHCl and sucrose to be singular. The taste of the basic stimulus HCl was rated as less singular (43%) than the other basics, and also less singular than some of the non-basic stimuli. The non-basics were rated as singular to some degree; for example, KCl was rated as singular by 48% of the subjects. To the extent that the basic tastes model requires singularity (the model is not clear on this), these data falsify the model.

 

These are not statistical statements; as for Fig. 1, the point of Fig. 3 is not that one stimulus is statistically more singular than another, but that the data are intermingled between basic stimuli and the non-basics. Also it seems clear, as non-statistical statements, that the basics do not rest at complete singularity (left column) and the non-basics do not rest at completely more-than-one (right column), as might be held in the basic tastes doctrine.  It is hoped that this conclusion will be tested in future studies.

 

Some of the singular ratings of the basics may be accounted for by the fact that the subjects were biased towards the verity of basic tastes, as mentioned above (Section 7.6).

 

 

 

 

 

Fig. 3.  Comparison stimuli rated as “singular” or “more-than-one;” data from the basic tastes protocols (Fig. 1 E1, E2). The left column indicates how the ratings would appear for a hypothetical stimulus that was rated as singular by all subjects, and the right column illustrates the data for a stimulus that was always rated as more-than-one.  No stimulus is completely either singular or more-than-one, the only two outcomes that perhaps would be predicted in the basic tastes model. 


7.10. Other comments on the experiment

It could be argued from the basic tastes model that the non-basic stimuli could account for the Comparison stimuli because each was composed of set amounts of basic tastes. If this argument were made, it must equally be argued that the basic tastes might function because they were composed of various amounts of the tastes of the non-basics. To remain unbiased, the argument cannot be made in one direction only, in support of the idea of basic tastes. The basic tastes position may not be assumed valid while testing its validity.

 

            If words are not trustworthy vehicles of communication between what we perceive and what we report, should less than the expected basic tastes outcomes be taken very seriously? Might non-verbal methods, not using the basic words, give different, more valid results?

 

Were the protocols appropriate for the questions asked? It could be argued that the present procedures did not properly examine the basic taste model - but it must be admitted that the effort was made. This then raises the good question of how this idea might be more properly tested. This is not as clear as it should be. It is hoped that other more substantial and appropriate experimental protocols will be pursued.

 

The present data are not as forceful as the previous findings in other psychophysical studies in supporting the basics view. This may be because herein the subjects were not instructed on how to use the basics. That the data are as supportive as they are may be due in part to the fact, mentioned in Section 7.6, that the subjects expected that their responses should be in terms of the basic tastes. This issue could be tested.

 

7.11. Do these data falsify the Basic Tastes model?

 

These data and discussion falsify the basic tastes model as it has been presented thus far. But, strictly speaking, without a clear rationale and definition of what “basic tastes” are, the model can neither be falsified nor supported. It is in a fatal limbo. This may be seen as a serious problem by some in the field of taste, and beyond. It is hoped that a clear and testable underlying hypothesis and the resulting definitions will be elaborated and tested. These could lead to a clearer view of the nature of taste.

 

8. A promising alternative to basic tastes: Thomas Young

 

The previous portions of this paper have addressed the many serious shortcomings of labeled-line models of neural information, and