Amino acids as taste stimuli. II. Quality coding

93

Brain Research, 253 (1982) 93-104 Elsevier Biomedical Press

Amino Acids as Taste Stimuli. II. Quality Coding THOMAS C. PRITCHARD* and THOMAS R. SCOTT

Department of Psychology and Institute for Neuroscience, University of Delaware, Newark, DE 19711 (U.S.A.) (Accepted June 8th, 1982)

Key words: amino acids - - chorda tympani - - conditioned taste aversion - - single unit recording - - taste primaries

Two experiments were performed in rats to evaluate the relative taste qualities of 12 L-amino acids, each at a concentration which evoked half the maximum response for that chemical. The first study involved recording the activity of 40 individual chorda tympani fibers to the stimulus series. Only 34 ~ of the evoked responses resembled the short latency phasic-tonic activity which characterizes gustatory responses to inorganic salts and acids. 32 ~ had latencies exceeding 1 s; another 27 ~ consisted of only a phasic burst lasting less than 1 s. The remaining 7 ~ were inhibitory. Both long latency and purely phasic activity were stimulus selective: 61 ~ of the former were in response to Gly or Pro while 69~o of the latter were evoked by Cys-HCI, Lys-HCI or His. Response inhibition was not associated with either specific fibers or stimuli. Thus amino acids, which to humans represent a class of perceptually complex stimuli, show a corresponding complexity of evoked neural properties in the rat. The second study employed a conditioned taste aversion paradigm to assess the qualitative similarity of each amino acid to the others and to the 4 prototypical taste stimuli: NaCI, HCI, quinine-HCl and sucrose. Some amino acids showed strong generalization to a single gustatory prototype (Pro and Gly to sucrose; Cys-HCI to HC1); others generalized well to multiple prototypes (e.g. Arg to sucrose and NaCI). Several showed poor generalization to all 4 prototypical tastes, calling into question the assumption that these 4 totally encompass the gustatory domain. INTRODUCTION W h o l e nerve recordings a n d 24 h t w o - b o t t l e preferences p r o v i d e 'whole system' d a t a on taste. These concern thresholds, i n t e n s i t y - r e s p o n s e functions a n d neural time courses, b u t do n o t specifically address issues o f coding, d o n e at the n e u r o n a l level, o r relative taste qualities. T o study the neural coding o f a m i n o acids an electrophysiologicai study on single axons o f the rat c h o r d a t y m p a n i (CT) nerve was conducted. In addition, a b e h a v i o r a l c o n c o m i t a n t to these neural measures was o b t a i n e d t h r o u g h use o f a c o n d i t i o n e d taste aversion ( C T A ) p a r a d i g m . A n i m a l s develop a C T A when a novel taste is p a i r e d with gastrointestinal malaise, even t h o u g h the taste might be considered appetitive a priori. The c a p a c i t y o f a C T A to generalize to o t h e r stimuli on the basis o f taste quality has been r e p e a t e d l y d e m o n s t r a t e d . N a c h m a n z3 f o u n d t h a t rats with C T A s to LiCI w o u l d show s t r o n g generalization to N a C I ( a n o t h e r

'salty' salt), b u t m u c h less to the bitter salts KC1 a n d NH4CI. In a m o r e comprehensive study involving hamsters N o w l i s et al. 26 showed that c o n d i t i o n e d aversions generalize to varying degrees to each o f the 4 p u t a t i v e p r i m a r y taste qualities: sweet, salty, sour a n d bitter. Smith et al. 33 used C T A generaliz a t i o n gradients in hamsters to desclibe the relative similarity a m o n g a range o f taste qualities. The tendency for a c o n d i t i o n e d aversion to generalize to similar tasting chemicals represents the basis for using a C T A p a r a d i g m to m a k e a qualitative assessment o f the rat's g u s t a t o r y experience. The degree o f generalization between two tastes is a behavioral a n a l o g o f the correlation coefficient between the neural patterns they evoke in afferent neurons. A n e u r o b e h a v i o r a l c o m p a r i s o n is then available: a m i n o acids which evoke similar profiles a m o n g a r a n d o m sample o f single units should also elicit similar b e h a v i o r a l reactions in this C T A p a r a digm.

* Present address for all correspondence: Laboratory of Physiological Psychology, The Rockefeller University, 1230 York Avenue, New York, NY 10021, U.S.A. 0006-8993/82/0000-0000/$02.75 © 1982 Elsevier Biomedical Press

94 EXPERIMENT I

UNIT 1-580

Materials and Methods

Illlllltl

m.IHIl

IIIL

_

III1i111111i111111!111 IIII

Subjects and stimuli Forty single axons were functionally isolated from the CTs of 32 adult, male Sprague-Dawley rats by dissecting small fascicles free from the larger nerve trunk. Stimuli were the same 12 L-amino acids used by Pritchard and Scott ~s, each at its half-maximum concentration (Table I). All stimuli except Arg were presented in random order. Arg was tested last because extensive rinsing was required to restore the spontaneous activity level of the neurons. A 0.l M NaC1 comparison stimulus was applied at the beginning and end of each series as well as after the fourth and eighth stimulus presentations to provide a measure of response stability. All other surgical, stimulation, and recording procedures were identical to those used by Pritchard and Scott 2s. Response criteria Previous studies have typically used response criteria based upon some statistical index of dispersion around the mean of the first 3 or 5 s of evoked activity 7,13,29,33. This type of criterion is most appropriate for short latency neural responses consisting of both phasic and tonic periods. The time courses found in the present study often did not conform to this pattern: some responses were purely phasic while others showed uncommonly long latencies. For example, unit 1-580 (Fig. 1) showed a phasictonic response to NaC1 but a purely phasic burst to TABLE 1

.~

__ - lit

NaCl

~

-

-

'

L- LYSINE Hal

L2 SEC~

-

-"

'2 SEC,~

PIMIIill

ll lll!llllr

I +l

IIII Illllll II

L-PROLINE 5 SEC.

GLYCINE 5 SEC.

1

Fig. 1. Single fiber responses representative of the range of time courses observed in Experiment 1. Lys-HC1 and a long latency tonic response to Pro and Gly. It was difficult to apply a single response criterion to such neurons so a more complex system of response evaluation was devised. Several criteria, each based upon the statistical probability of dispersion around the mean spontaneous activity for 1, 5, or 15 s were employed (Table II). Response latency was determined by applying the criterion for a phasic response (mean spontaneous discharge plus two standard deviations) to successive I-s periods of evoked activity.

Stimuli and concentrations

0.15 M 0.3 M 0.006 M 1.0 M 0.001 M 0.03 M 0.01 M 0.03 M 0.06 M 1.15 M 0.1 M 0.003 M 0.1 M

L-alanine L-arginine L-cysteine hydrochloride glycine L-histidine L-isoleucine L-leucine L-lysine hydrochloride L-methionine L-proline L-threonine L-tryptophan sodium chloride

(Ala) (Arg) (Cys-HCI) (Gly) (His) (lie) (Leu) (Lys-HCI) (Met) (Pro) (Thr) (Trp) (NaC1)

Data analysis The neural responses were passed through a window discriminator before being processed by a PDP11/03 computer which provided poststimulus-time (PST) and interspike interval (ISI) histograms, as well as total spike count during each 100 ms of stimulation. The pattern of activity evoked by each amino acid across all 40 neurons was determined and the Pearson product-moment correlation coefficient between each pair of patterns was calculated.

95 TABLE II Response criteria

Criteria for categorizing responses as excitatory or inhibitory. Also included are the percentages of false positives expected and the number actually obtained, as calculated by applying each criterion to periods on spontaneous activity Type of response

Criterion for excitatory response

Criterion for inhibitoryresponse

Additional requirements and comments

Expected rate Actual rate of of false positives false positives

Phasic (lst s only)

1 s mean spontaneous rate + 2.00 S.D.

--

First sec total must exceed 50 % of 5 s total

2.3 %

5.1

Minimum excitatory response was 3 spikes/5 s

10%

14.7%

Only neurons with a spont, rate /> 4 spikes per 5 s analyzed

10 %

7.1%

Only applicable to Gly and Pro

--

Tonic (lst 5 s)

5 s mean spontaneous rate ÷ 1.28 S.D. 5 s mean spontaneous rate --1.28 S.D.

Tonic (lst 15 s)

15 s mean sponta-neous rate ÷ 1.00 S.D.

These correlations served as the basis for generating a m u l t i d i m e n s i o n a l stimulus space (see below). Results Effectiveness T h e 40 n e u r o n s isolated in this study were rem a r k a b l y consistent in their responses to the 12 a m i n o acids a n d NaC1. Arg, Pro, Lys-HC1, Cys-

HC1, G l y a n d NaC1 all activated m o r e t h a n 70 % o f the cells while His, Met, Trp, lie a n d Leu each e v o k e d responses f r o m fewer t h a n 20 %. This r a n k ing o f stimulus effectiveness c o r r e l a t e d + 0.90 with the o r d e r o b t a i n e d b y r a n k i n g each a m i n o acid a c c o r d i n g to its whole nerve response at the halfm a x i m u m concentration. However, while the r a n k ing o f effectiveness was preserved f r o m whole nerve to single n e u r o n levels, response m a g n i t u d e was not. Chemicals which were least effective in whole nerve recordings were u n d e r - r e p r e s e n t e d in the single n e u r o n sample while the m o s t effective stimuli were over-represented. This conclusion follows f r o m a c o m p a r i s o n o f e v o k e d responses f r o m the whole C T a n d f r o m single neurons, shown in Fig. 2. A direct m a t h e m a t i c a l c o m p a r i s o n is strained b y the fact that w h o l e nerve a n d single unit recordings rely u p o n different features o f the neural activity. Still, it is a p p a r e n t t h a t the 5 m o s t effective chemicals for the

whole C T yielded even greater increases f r o m the s u m m e d unit responses while 7 less effective a m i n o acids fared even m o r e p o o r l y a m o n g the unit sample The most extreme cases were those o f Ile a n d Met, for which the s u m m e d single n e u r o n activity across

/~ I000

/ /

E_

CO Z n-

°ARG~ / e LYS HCI

CYS HCle ~

LLI%t.

NoCI

//.-~Ly ,///• PR0 I00

E~

,,,z LLI Z

lO

THRe

/

/

TRP

/

_1

F'~ 0 t"-

/

/

_1

/ mLEU / , .'LE "

. . . . . .

"' 50 MULTIUNIT

"HIS

.MET I

i

I00 RESPONSE

(ARBITRARY

n I i nn i I

500 I000 MAGNITUDE

UNITS)

Fig. 2. Comparison of single and multiunit response magnitudes. Stimuli plotted below the equal response line were under-represented in the single unit sample; those above the equal response line were over-represented. The net inhibitory responses evoked by Ile (--42) and Met (--48) were not plotted on the ordinate or included in the analysis for the bestfit line.

96 6/~ ~

ALA ARG

,

_

~

0L

~

illustrated in Fig. l constituted 59.1 ~,~ (123 of 208) of the total n u m b e r o f responses obtained by 480 (40 × 12)amino acid-neuron interactions. An additional 6.7°,~ (n == 14) were inhibitory responses. Thus approximately two-thirds o f the responses to these amino acids were not o f the short latency, phasictonic variety typically seen in the taste system when inorganic salts and acids are used as stimuli. Response latency in the rat CT may be as short as 25-30 ms when the tongue is stimulated by electrolytes; longer latencies are associated with sugars, but even those rarely exceed 500 ms. Thus it is extraordinary that 32.2 % (n =: 67) o f the response latencies in this experiment exceeded I s and 25.0'.',,'i (n :-- 52) surpassed 2 s. Response latency histograms

~

2O

CYS HCl

GLY

'0l'2.r . . . . . .

d HIS

.-~

~

_

^

~

o

W

hi CO --I

LEU LYS HCI 4° 20 I 0 ~

12.

MET

"'" ~ ~

[~

d

o

I--

PRO

TRP

NaCI

0 /'-"-~

o

I

2

5

I

2

3 4

';h

-"/~'~

4

I0

II

12 13 14'15'

5

6

7

8

9

5

6

7

8

9 10 11 12 13 14 15

20O Io0

U3 b3 0 Z

T I M E (Sec.)

Fig. 3. Superimposition of the single unit PST histogram (cross-hatched area) and the average, multiunit time course of each stimulus. The multiunit time course for each amino acid was adjusted in size relative to the response of 0.1 M NaCI (see Pritchard and Scottes). The PST histograms were scaled to approximate the size of the multiunit response for the corresponding stimulus. PST histogram bin width = 100 ms. all 40 fibers represented a net inhibition. N o n e o f the whole nerve records provided any evidence of inhibition for lie, Met or any other amino acid. This is shown in Fig. 3, which presents the summed unit data superimposed on whole nerve records for each stimulus. While the correspondence o f single and multiunit time courses varies considerably a m o n g the different stimuli, the inhibitory PST histograms o f Ile and Met clearly provide the poorest correspondence with their respective multiunit records. T i m e course

The purely phasic or long latency tonic responses

ARG

!~

i

LYS HCI

o

35~

OLL

NaCl rn

HCl

Z

,

.

5

.

.

.

MET

.

!L o

[EU

5[~'~

HIS

12'3'45

123

POSTSTIMULUS

:L 6

,

45

.

12

ONSET TIME

THR

34'5'

(Sec.)

Fig. 4. Response latency histograms for the stimuli in Experiment 1. Onset time was designated as the second during which evoked activity exceeded the mean spontaneous rate by two standard deviations.

97 reliability of the temporal disparities seen in the present experiment imply that this is particularly true for this range of amino acids.

__J

GLY

SPONT= 0 . 4 / 5

SEC.

mJ

GLY

SPONT= 1 . 6 3 / 5 SEC. J

I

5 SEC.

UNIT 5 - 7 2 5 J

GLy

SPONT = 0 . 4 3 / 5

SEC.

i

i

15 SEC.

Fig. 5. Three long latency responses recorded from individual chorda tympani fibers.

for all amino acids and NaC1 are shown in Fig. 4. Long latency responses (latency > 1 s) were specifically associated with Pro and Gly, which accounted for 61.2 ~ (41 of 67) of the cases. A majority of the neurons (51.1~) which responded to these two amino acids did so with latencies greater than 1 s. Fig. 5 illustrates the responses of two neurons which had latencies of at least 5 s and one in which the response was not initiated for nearly 30 s. Stimulus applications were replicated several times and even extremely long latencies proved to be quite stable, typically deviating by less than 10~ across trials. Sodium chloride and Lys-HC1 were the only stimuli which evoked responses in less than 1 s in all cases. The temporal pattern of the evoked discharge, whether phasic or tonic, was primarily a function of the stimulus rather than the fiber. Cys-HCI, His and Lys-HC1 elicited purely phasic responses (no tonic component) from 41 ~, 50 ~ and 63 ~ respectively of the responding fibers. This contrasts with Pro, Gly and Ala which evoked purely phasic responses from only 3 ~, 11 ~ and 12 ~ respectively. Temporal characteristics have been shown to be of some significance in the quality code for chemicals8, z5 and the

Inhibition Analysis of response inibition in the rat CT is largely precluded by the low spontaneous activity of these cells. The mean spontaneous rate of the 40 neurons reported here was 1.09 4- 2.19 spikes/s. Only 13 cells had rates above 0.8 spikes/s, the level below which statistical inhibition was not possible. Of these 13, 8 produced a total of 14 inhibitory responses, 3 of which are shown in Fig. 6. Multidimensional analysis The pattern of responses evoked by each amino acid across all 40 fibers was used to characterize the neural code for that chemical. The intercorrelation of each pattern with all others yielded a matrix, the cells of which contain the correlation coefficients indicating the degree of similarity between any two stimuli. The matrix was then used to generate a multidimensional space within which the distance between any two stimuli is inversely proportional to the correlation coefficient between the patterns of activity they evoke 15. The resulting space is shown

UNIT 4-1189

X SPONT.- 19.2/5 SEC. J

UNIT

MET

6-314

x SPONT = 13,26/5 SEC. i

TRP

JNIT 5-930

X SPONI~ • 5.88/5 SEC. J

MET

I

2 Sec

4

Fig. 6. Three inhibitory responses recorded from individual chorda tympani fibers.

98 NoCI

r I/,W

NoCI

/ / /,,.<.'.I//;,~,'/ /~

/ ; l ,I / / r " , °" / / I x / / / / i I L V ! / / / 1,,I VI ;'~¢ /~/ /

L_II I . If _/L / I I I I

I LV I I V I I /~1/ I A I

B

Fig. 7. Multidimensional taste space generated from chorda tympani single fiber analysis. As a 3-dimensional solution, this space accounted for 93 % of the data variance.

in 3 dimensions in Fig. 7*. The dimensions are labeled only A, B and C because no physicochemical properties clearly define the axes. In an ideal preparation, the 3 applications of NaCI would be at identical coordinates (the final replication of NaC1 proved to be unreliable since it followed Arg in each case, and so was not included in this analysis). The rather tight cluster in this space indicates stability of the preparation throughout the recording session. Arg at 0.3 M is the only amino acid to show any similarity to NaCI. This corresponds to psychophysical descriptions of its having a distinct salty component at midrange concentrations 34. Cys-HC1 and Lys-HCI are grouped together, probably because of the common HC1 radical. However, this is not simply a pH effect since Cys-HC1 had a pH of 1.80 while that of Lys-HC1 was 5.95. Neither does their proximity reflect hedonic similarity for rats, for Cys-HCI was strongly avoided while Lys-HCI, though generally aversive, was nearly neutral at this concentration 2s. On the basis of their behavioral and neural similarities, Gly and Pro should have been tightly bound in the space. The modest spread between them was probably due to our inability to include in

this analysis the very long latency responses which they alone shared. The remaining 7 amino acids were less effective in activating CT fibers. Within this group efficiency ranged from Leu which evoked responses from two of the 40 fibers, to Thr which activated 9. Perhaps because of this shared property, these stimuli clustered in one corner of the space. EXPERIMENT 2

Materials and Methods Subjects and stimuli Subjects were 192 adult, male Wistar and Sprague-Dawley derived rats housed individually and allowed free access to food but not water. They were divided into 12 groups of 16 each and assigned to either experimental (n = 160) or control (n z 32) conditions. Stimuli were the 12 amino acids used in Experiment 1, all at the same concentration except for Arg, whose intensity was reduced to 0.1 M because subjects refused to consume 0.3 M under the deprivation conditions. Additional stimuli were 0.1 M NaCI, 1.0 M sucrose, 0.03 M HC1, and 0.001 M quinine-HCl (QHC1). Training The drinking schedule for each subject was manipulated by restricting daily access to water to a 15min session at 11.00 h and a 1-h session at 16.00 h. Acclimatization to the schedule was complete in 7 days. The 1-h drinking period was continued throughout to ensure adequate hydration and a stable level of fluid deprivation across days. After this adaption period each of the 16 rats in one experimental group was given access to one of the 16 CSs during the 15-min drinking session. Each subject then received an intraperitoneal injection of LiCI (127 mg/kg), the UCS, to induce gastrointestinal malaise. Control rats drank distilled water before the LiC1 injection. Following a 48-h recovery period each of the 16 test stimuli, including the CS, was presented in random order, one per day, to

* A 3-dimensional solution was chosen because it accounted for 93 % of the data variance, and additional solutions added only minimally to this percentage.

99 each subject during the 15-min drinking session. Each chemical was used as both CS and test stimulus to maximize the number of taste quality comparisons (both primary, e.g. sucrose, and complex, e.g. Arg) and thereby expand the descriptive power of the experimental paradigm. The volume of each test stimulus consumed was recorded to the nearest 0.5 ml. The original CS-UCS pairing was repeated for each subject after the fifth and tenth test days so as to maintain a robust taste aversion.

I00 90

A

80 70

PRO

50 40 30

'

,'

~t I iI

20

co I..u n,." n n,

a---a

'

W\, ,....,,,'1 it,, ,,,.,, ,,

60

z 0

,'",

i i

GLY . - - ,

10 0 -I0

03

Data analysis The strength of the CTA to any stimulus was expressed as a percent suppression according to the formula: suppression ~-

ml test stimulus (Exp'l) ml test stimulus (control)

× 100

(.9 Z

tOO

B

9O

Z 80

NaCl

o---o

70

IZ LU n,, LIJ

i t I

•

60

j

50

•

I

40 30 2O

High suppression scores represent strong generalization from the CS to the test stimulus, implying similar taste qualities. However, a more comprehensive measure of similarity between two chemicals could be gained by a comparison of suppression scores across the full array of CSs. This is illustrated in Fig. 8A and B for Gly, Pro and NaCI from the present experiment. Gly and Pro exhibit highly correlated patterns across the 16 test stimuli (r = +0.81) suggesting a similar taste quality. Gly and NaC1, however, show poorly correlated suppression patterns (r ---- +0.02) characteristic of stimuli having quite dissimilar tastes. Pearson product-moment correlations were calculated for each stimulus pair and these were used to position each amino acid within a multidimensional space.

Results Suppression and salience Percent suppression scores indicated that in each case a stronger aversion was formed to the CS than to any other stimulus. While this result was expected, the range of self-suppression scores for the CSs was wide (range = 1 7 - 9 8 ~ ; mean = 6 8 ~ ) , showing that each CS was not an equally effective cue. Previous investigators, noting this effect, have described effective CSs as being salient 20 and have tried to specify the characteristics which determine this attribute. CS intensity, quantity consumed, and ex-

I0

"o

0

"I

-I0

Ac. I c~s I .~s I L~01 .~'. I .~. I .o'~, I s'0 I ARG

GLY

ILE

TEST

LYS

PRO

TRP

HCI

QHCi

STIMULUS

Fig. 8. Spectrum of drinking suppression scores for two highly correlated stimuli (A, Gly and Pro; r = 0.81) and two uncorrelated stimuli (B, Gly and NaCI; r = 0.02). High correlations imply high qualitative similarity; low or negative correlations imply dissimilarity.

posure period are all directly related to the strength of an aversion lm, and these factors may relate to palatability 14,35 or the degree of neophobia a rat might show ~4. However, intense CSs or prolonged exposure to them may merely serve to clearly distinguish them from more mundane experiences, thus enhancing their saliencO s. In the present experiment, the stimulus characteristic which related most strongly to CS salience was hedonic intensity. This was defined as the absolute difference in taste preference from hedonic neutrality (50 ~ ) based on two-bottle preference data 28. The correlation between self-suppression and hedonic intensity, depicted in Fig. 9, was +0.87 (t = 5.58; P < 0.001). Moreover, the partial correlation coefficient between percent suppression and hedonic intensity (with CS concentration held constant) was nearly identical (r = +0.89), indicating that the

100 effect of hedonic intensity upon CTA acquisition is not merely a concentration effect. c

Multidimensional analysis

NaCl

Since multidimensional scaling techniques organize a data set such that spatial proximity represents degree of similarity, the space generated from suppression scores (Fig. 10) should relate to the electrophysiological results of Experiment 1, the behavioral data of the preceding paper 2s and human psych•physical studies. The psych•physical literature describes Gly and Pro as sweet ')2,3°-a',3v, and indeed both are within the same cluster as sucrose. Electrophysiological and behavioral results emphasize the similarity of Gly and Pro. Both have high neural and behavioral thresholds. At moderate concentrations they are the two most preferred of the dozen amino acids tested, but each becomes sharply more aversive above 0.6 M. Integrated CT activity shows moderate transient responses with a sustained or even increasing tonic component, and this was verified at the single neuron level by the large number of long latency responses to both. Psych•physical studies indicate that at low concentrations Arg has a sweet component 22, but at I00 90

ALA•

B°i 7£

x q,bc~THeR B£

5C

50 -~0 • e RP HIS r0

~o HEDONIC

~o

~:o

,~o

INTENSITY

(ABSOLUTE CHANGE EROM 5 0 %

PREFERENCE)

Fig. 9. Percent suppression of drinking as a function of hedonic intensity. Hedonic intensity was defined as the absolute difference in taste preference from hedonic neutrality (50 ~ ) based upon two-bottle preference data 2s.

,o. r / //

I VI

I/LPt-'U I I I

/

la . [ I Ill

/l

I .t 7A

//t//7'lt/7 Y/ / /

Fig. 10. Spatial representation of gustatory similarity among the stimuli included in Experiment 2. This taste space, constructed in 3 dimensions, accounted for 88% of the data variance.

higher intensities saltiness and bitterness become pronouncedal, 35. In the neural space of Experiment 1, Arg at 0.3 M was the closest amino acid to NaCl, but in the behavioral space 0.1 M Arg fell near the sucrose cluster. Correspondingly, the rat's preference for Arg falls from 63 ~ at 0.1 M to 24 ~,~ at 0.3 M28. Cys-HCI is situated near HCI, its position probably determined by its acidic nature (pH -- 1.8), Ala and Met are adjacent in the CTA space but this is contrary to other findings. At half-maximum concentrations Ala was preferred in the two-bottle preference test (75 ~o acceptance); Met was rejected (9 ~o). Psych•physical tests indicate that Ala is sweet while Met is sulfurous, obnoxious and bitter '~°. In addition, Met has a distinct olfactory component, which Ala lacks. In the CTA study both stimuli proved to be salient CSs (Ala: 8 8 ~ self-suppression; Met: 96 ~), indicating that the behavioral and psych•physical distinctions mentioned above should have been manifested in dissimilar generalization gradients and separate placements in the stimulus space. The remaining 6 amino acids (His, lie, Leu, LysHC1, Thr, Trp) cluster in the lower right corner of the space. All but Lys-HCI were only moderately effective in Experiment 1, and their congregation in a common corner of the CTA space suggests that

101 they are poorly defined behaviorally as well. LysHC1 evoked the largest summated CT response of any amino acid used, yet proved to be of low salience. It was hedonically neutral at its halfmaximum concentration 28, showed poor self-suppression in this CTA experiment (42.4~), and correspondingly moved away from Cys-HCI, its neural neighbor, into the lower right corner of the CTA space. A significant neural response and behavioral aversion had been established for Lys-HCI at a lower concentration than that used in the CTA study. Thus a clear distinction is drawn between taste recognition and the salience of that taste to the rat's recall. DISCUSSION

The data from Experiment 1 permit a comparison between responses from a single neuron sample and those from the whole nerve 28. They also generally corroborate the varied time courses seen in whole CT records. The protocol of Experiment 2 allows a separate evaluation of each amino acid, first in the role of a CS and then as a test stimulus. From this come implications for the use of reciprocal suppression scores in the CTA paradigm. Multidimensional scaling techniques applied to these results also allow a determination of whether these perceptually complex chemicals can be accommodated by the scheme of four primary tastes. These issues will be discussed in turn. Single neuron vs whole C T responses

Chemicals which were least effective in the whole nerve recordings were under-represented in the single neuron sample while the most effective stimuli were over-represented. One factor contributing to this disparity is the interaction between chemical sensitivity and fiber diameter in the CT. Boudreau and colleagues a-5 have presented data based upon single neuron recordings from the geniculate ganglia of cats, dogs and rats indicating that chemical sensitivity is not randomly distributed across the neural population. Instead, large diameter axons are primarily sensitive to inorganic acids while medium diameter fibers tend to be salt-sensitive and the smallest to be quinine-sensitive. These data imply that CT anatomy presents an impediment to ran-

dom and representative electrophysiological sampling; they also provide the basis for the disparity of response magnitudes between single neuron and whole CT recordings. While electrophysiological recordings in general are biased toward large neurons, whole nerve responses may be less so than single neuron samples since smaller axons remain within the ken of the electrode. The disparity in response magnitudes may result from the application of two differentially biased sampling techniques to a population of neurons distributed, in their chemical sensitivity, in a non-random fashion. T i m e course

Another way of evaluating the correspondence between whole CT and single neuron responses is to compare time courses (Fig. 3). For several stimuli, the two forms of data are in accord: Lys-HCl, His and Cys-HCI show distinct transients and rapid decay; Pro increases over time in both records due to the numerous long latency responses it evokes. Comparisons for Trp and Arg, however, indicate an overrepresentation of phasic-responding single cells while for Ala and Thr there appears to be an u~der-representation. These discrepancies are assumed to result from a sampling bias among single neurons, and their significance may be considerable. Both the phasic portion 16 and the temporal sequence 8 of the response have been shown to be related to quality coding. Nearly one-third of the responses evoked by this amino acid series were of latencies greater than 1 s, with Pro and Gly accounting for a majority of these. Latencies of 2 s have been reported previously, but only as the product &exceptional experimental conditions, such as removal of the tongue's epithelial layer s or recording from the lingual nerve central to the departure point of CT2, el. Ii~ both cases successful stimulation required concentrated solutions which quickly desensitized the tongue. Several lines of evidence suggest that latent responses observed in this study were not due to chemical stimulation of free nerve endings, as was the case in the studies described above. First, despite being concentrated, neither Gly (1.0 M) nor Pro (1.15 M) is an irritant which might permeate the tongue's surface to reach free nerve endings. Both Arg (pH = 11.2)and Cys-HCI (pH = 1.8), on the

102 other hand, are quite caustic, yet produced few latent responses. Secondly, fibers showing long-latency respgnses had normal chemical sensitivity, and in all cases responded to at least one other amino acid at the half-maximum concentration.

Reciprocity of conditioned aversions In most CTA studies the subjects are retested with the CS to determine the degree of drinking suppression, and so the effect of the experimental manipulation. Thus it has not been possible to determine whether the effective stimulus properties, such as novelty, hedonic intensity and palatability, relate to the CS or the test stimulus since they were the same. The assumption has been that effective cue properties relate to a chemical's role as a CS rather than as a test stimulus even though experimental protocols could not separate the two. This assumption follows from the belief that CTA learning is a variation of traditional learning paradigms where the CS properties relate directly to strength and speed of conditioning~,19, z7. However, this emphasis on the CS cue value ignores the role of the test stimulus as well as possible interactions between specific CSs and test stimuli. In the present experiment each chemical was both a CS and a test stimulus, making it possible to evaluate their respective roles in the process of CTA generalization. It was found that the degree of generalization for most CS-test stimulus pairs was dependent not only uport perceived similarity but also on which chemical was the CS. This is exemplified by the pairing of Ala and Arg, two stimuli with excellent self-suppression scores (88 }o~ and 96'~ respectively), indicating considerable salience for both. With Ala as CS the generalization to Arg was 64 o/ suppression, but generalization from Arg to Ala was only 22 %. Across all stimulus pairs, the correlation for reciprocal generalization was +0.36. This lack of symmetry implies that the roles of CS and test stimulus are quite different in a generalization paradigm and calls into question the appropriateness of combining suppression scores based upon the assumption of reciprocity of generalization gradients. It also suggests that different characteristics of a given stimulus are pertinent for the roles of CS and test stimulus. One stimulus property clearly related to generalization but differing from CSs to test stimuli was

hedonic value. The most effective test stimuli were sucrose, Gly, Pro and Arg: chemicals which showed similar patterns of drinking suppression across the array of stimuli. In contrast to these appetitive tastes, QHCI, lie, HC1 and Trp, all aversive, ranked among the least effective test stimuli. The same pattern was not evident for CSs, in fact sucrose, Gly and Pro were among the least effective CSs in this study. Despite these anomalies, there are no data from the present experiment which question the evidence of Nachman 23, Nowlis et al. ~6 and Smith et al. 33 that CTA generalization follows a gradient of gustatory similarity.

Taste primaries Perhaps the most protracted debate within gustation has centered about the issue of primary taste qualities. In the nineteenth century, the question was the number and type of primaries, but today the very existence of primaries is disputed ~6. Still, most researchers support the notion of primaries as Henning 17 described it in his taste tetrahedron. The corners of the tetrabedron are occupied by the primary qualities salty, sweet, sour, bitter, and transitional tastes lie along a continuum between primaries or on a tetrahedral face bounded by three primaries. NaCI, sucrose, HC1 and QHC1 are the prototypical primary stimuli. Each of the putative primary stimuli was included in this experiment and in the multidimensional space shown in Fig. 10. They should form a perimeter which encloses the other chemicals. This was not the case. Many amino acids, despite good self-suppression, showed poor generalization to all primaries, and so were positioned outside the tetrahedron formed by connecting the loci of these 4 stimuli. There are several interpretations for the failure of Henning's model in this study. The first is that stimuli which did not conform to the tetrahedron were tasteless, or that non-gustatory characteristics of these complex stimuli influenced construction of the space. The location of pungent Met and several pHrelated stimuli at the edge & t h e space supports this view. However, other occupants of the same corner have no apparent olfactory or trigeminal components (viz. Ala, Leu, Thr). Moreover, the strong selfsuppression shown by several of these chemicals implies that they impart a recognizable taste, for

103 non-gustatory cues alone would not suffice to generate a strong conditioned aversion with one C S - U C S pairing. The sensory effectiveness o f these stimuli is further established by the preferences shown by rats in the preceding study 2s. Five o f the 9 amino acids which do n o t fall within the tetrahedron elicit preference scores deviating > 20 ~ from neutrality at the concentrations used here. A second possibility is that the segregation o f amino acids to a corner o f the space reflects the influence o f an additional primary. Japanese researchers include a fifth primary, umami (meat-like), in their descriptions o f taste quality 22. While the prototypical umami taste is represented by an amino acid - - m o n o s o d i u m L-glutamate - - a m o n g the stimuli in the present study it is Met and Ala which impart this taste. These are the two most behaviorally salient stimuli which do not c o n f o r m to a tetrahedral arrangement. The final interpretation o f the inadequacy o f Henning's model is that taste is n o t founded on a system o f primary qualities. Erickson10-12, a6 has

proposed that taste quality varies continuously along one or more stimulus dimensions. Thus, in a multidimensional space, NaC1, sucrose, HCI and Q H C I should diverge f r o m one another because o f their obvious gustatory dissimilarities, but they would not be required to f o r m a perimeter encompassing all other chemicals. W h e n a ' p r i m a r y ' is situated at the edge o f a space it merely indicates that this stimulus represents an extreme position along one or more stimulus dimensions. While psychophysical reports often fall into four categories, this could result f r o m the constraints placed u p o n subjects by the dearth o f gustatory descriptors available to them. I n the label-free language o f licking, these restrictions disappear.

REFERENCES

aversions with variations in intensity of the CS and US in two strains of rats, Psychonom. Sci., 22 (1971) 303-304. 10 Erickson, R. P., Sensory neural patterns in gustation. In Y. Zotterman (Ed.), Olfaction and Taste, Pergamon Press, Oxford, 1963, pp. 205-213. 11 Erickson, R. P., Neural coding of taste quality. In M. Kare and O. Maller (Eds.), The Chemical Senses and Nutrition, Johns Hopkins Press, Baltimore, MD, 1967, pp. 313-327. 12 Erickson, R. P., Doetsch, G. S. and Marshall, D. A., The gustatory neural response function, J. gen. Physiol., 49 (1965) 247-263. 13 Frank, M., The classification of mammalian afferent taste nerve fibers, Chem. Senses Flay., 1 (1974) 53-60. 14 Green, K. F. and Churchill, P. A., An effect of flavors on strength of conditioned aversions, Psychonom. Sci., 21 (1970) 19-20. 15 Gutman, L., A general nonmetric technique for finding the smallest coordinate space for a configuration of points, Psychometrika, 33 (1968) 469-507. 16 Halpern, B. P. and Tapper, D. N., Taste stimuli: quality coding time, Science, 171 (1971) 1256-1258. 17 Henning, H., The quality continuum of taste, Z. PsychoL, 74 (1916) 203-219. 18 Kalat, J. W., Taste salience depends on novelty, not concentration, in taste-aversion learning in the rat, J. comp. physioL Psychol., 86 (1974) 47-50. 19 Kalat, J. W., Status of 'learned safety' or 'learned noncorrelation' as a mechanism in taste aversion learning. In L. M. Barker, M. R. Best and M. Domjan (Eds.), Learning Mechanisms in Food Selection, Baylor Univer-

1 Barker, L. M., CS duration, amount and concentration effect in conditioning taste aversion, Learning Motiv., 7 (1976) 265-273. 2 Beidler, L. M., Comparison of gustatory receptors, olfactory receptors, and free nerve endings, Cold Spr. Harb. Symp. quant. Biol., 30 (1965) 191-200. 3 Boudreau, J. C., Bradley, B. E., Beirer, P. R., Kruger, S. and Tsuchitani, C., Single unit recordings from the geniculate ganglia of the facial nerve of the cat, Exp. Brain Res., 13 (1971) 461-488. 4 Boudreau, J. C. and White, T., Flavor chemistry of the carnivore taste systems, Amer. Chem. Soc. Syrup. Series, 67 (1978) 102-128. 5 Boudreau, J. C., Oravec, J., Moang, N. K. and White, T., Neurophysiology of the geniculate ganglion amino acid sensitive (sweet-bitter) system. Paper presented at the Second Annual Meeting of the Association for Chemoreception Sciences, 1980. 6 Church, R. M., Response suppression. In B. A. Campbell and R. M. Church (Eds.), Punishment and Aversive Behavior, Appleton-Century-Crofts, New York, 1969, pp. 111-156. 7 Contreras, R. and Frank, M., Sodium deprivation alters neural responses to gustatory stimuli, J. gen. Physiol., 73 (1979) 569-594. 8 Covey, E. and Erickson, R. P., Temporal neural coding in gustation. Paper presented at the Second Annual Meeting of the Association for Chemoreeeption Sciences,

1980. 9 Dragoin, W. B., Conditioning and extinction of taste

ACKNOWLEDGEMENTS Supported by a research grant f r o m the Delaware Institute for Medical Education and Research and by G r a n t NS 10405 f r o m the National Institutes o f Health.

104 sity Press, Waco, TX, 1977, pp. 273-293. 20 Kalat, J. W. and Rozin, P., 'Salience': a factor which can override temporal contiguity in taste aversion learning, J. eomp. physiol. Psychol., 71 (1970) 192-197. 21 Kawamura, Y., Okamoto, J. and Funakoshi, M., A role of oral afferents in aversion to taste solutions, Physiol. Behav., 3 (1968) 537-542. 22 Kirimura, J., Shimizu, A., Kimizuka, A., Ninomiya, T. and Katsuya, N., The contribution of peptides and amino acids to the taste of foodstuffs, J. Agrie. Food Chem., 17 (1969) 689-695. 23 Nachman, M., Learned aversion to the taste of lithium chloride and generalization to other salts, J. comp. physiol. Psychol., 56 (1963) 343 349. 24 Nachman, M., Rauschenberger, J. and Ashe, J. H., Studies of learned aversions using nongustatory stimuli. In L. M. Barker, M. R, Best and M. Domjan (Eds.), Learning Mechanisms in Food Selection, Baylor University Press, Waco, TX, 1977, pp. 385-417. 25 Nagai, T. and Ueda, K,, Stochastic properties of gustatory impulse discharges in rat chorda tympani fibers, J. Neurophysiol., 45 (1981) 574-592. 26 Nowlis, G. H., Frank, M. E. and Pfaffmann, C., Specificity of acquired aversions to taste qualities in hamsters and rats, J. comp. physiol. Psychol., 94 (1980) 932 -942. 27 Passey, G. E., The influence of intensity of unconditioned stimulus upon acquisition of a conditioned response, J. exp. Psychol., 38 (1948) 420--428.

28 Pritchard, T. C. and Scott, T. R., Amino acids as taste stimuli. I. Neural and behavioral attributes, Brain Research, 253 (1982) 81-92. 29 Sato, M., Ogawa, H. and Yamashita, S., Response properties of macaque monkey chorda tympani fibers, J. gen. Physiol., 66 (1975) 781-810. 30 Schiffman, S. S. and Dackis, C., Taste of nutrients: Amino acids, vitamins and fatty acids, Percept. Psychophys., 17 (1975) 140-146. 31 Schiffman, S. S., Sennewald, K. and Gagnon, J., Comparison of taste qualities and thresholds of D- and Lamino acids, Physiol. Behav., 27 (1981) 51 59. 32 Shallenberger, R. S. and Acree, T. E., Molecular theory and sweet taste, Nature (Lond.), 216 (1967) 480-482. 33 Smith, D. V., Travers, J. and van Buskirk, R., Brainstem correlations of gustatory similarity in the hamster, Brain Res. Bull., 4 (1979) 359-372. 34 Stone, H., Gustatory responses to L-amino acids in man. In T. Hayashi (Ed.), OIfaction and Taste, Vol 11, Pergamon Press, Oxford, 1967, pp. 289--306. 35 Sutker, L. W., The effect of initial taste preference on subsequent radiation induced aversion conditioning to saccharin solution, Psyehonom. Sei., 25 (1971) 1-2. 36 Woolston, D. C. and Erickson, R. P., Concept of neuron types in gustation in the rat, J. Neurophysiol., 42 (1979) 1390 1409. 37 Yoshida, M. and Saito, S,, Multidimensional scaling of the taste of amino acids, Jap. Psychol. Res., I l (1969) 149-166.