excreted by ten hyperkinetic children
Appleman,
off medication was similar to that ex-
creted by an age-matched group of
normal controls. In addition, the mean
valUes for both the hyperactive patients
and control children were in the sam-
range as our aduLlt normal controls.
After pro'onged physical activity (foot-
ball), there was no significant differ-
eince between the pre- and postexercise
levels of Urinatry cyclic AMP in seven
normall suibjects. In contrast to our
findings, one study suggested that ex-
crcise maly elevate urinary cyclic AMP
levels ( 17). Robison et al. also recently
reported normiial levels of cerebral spinal
fluid cyclic AMP in manic patients:
however, they did not study patients at
the time of the swvitch into mania and
thuLs would have missed a transient
mlarked peak at that time (18). In our
stuLdies, cyclic AMP excretion is inde-
penldent of age anid sex (3).We have re-
viewed the known factors intluencing
cyclic AMP excrction (3, 5), but the
relatixe proportion of cyclic AMP com-
inig from extrarenal sources remainis to
be cletermined. Thus, at this time, one
canniot state wvhether the cyclic AMP
response is mediated centrally or is a
reflection of peripheral metabolism. De-
creases in the excretion of urinary cy-
clic AMP have been demonstrated in
patients with pseudohypoparathyroidism.
Thtus, changes in calcium metabolism
may be associated with alterations in cy-
clic AMP metabolism (3). Althou,gh the
changes iin urinary cyclic AMP nmax
be secondary to catecholamine or cal-
ciulm chaniges, our evidence docuLmlents
atn alteration in an important process
that accomiipanies and, in at least one
instance, preceded gross behavioral
changes. This further suLggests the im-
portance of biochemical changes in the
manic-depressive illness. It is of inter-
est to consider the possibility that the
concept of cyclic AMP as a trigger
mech'anism for metabolic processes may
be relevant to the switch process from
depression to mania.
MICHAEL I. PAUL
Neuropsychiatric Institute, Ceiter
for the Health Sciences, University
of California, Los Anugeles 90024
HINRICH CRAMER
Laboratory of Cheiinical Pharmnacolog,
Natio!nal Heart anid Luiil Institlute,
Bethesa., Maryland 20014
WILLIAM E. BUNNEY, JR.
L,abora-'ory of Clinical Scienice,
National Institute of Menttal Healtlh,
Bethesda, Maryland 20014
22 JANTJARY 1971
References and Notes 9. GC. Biaooker, 1. J. Tlhomas, NI. NI.
1. W. E. (E.
Bisochemistrru 7, 4177 (1968).
W. Sutherland, G. A. Robison, R. W.
Bunney, Jr., G. F. Boi-ge, D. L. 11)
Murphy, F. K. Goodwin, paper presented at
the anntual meeting of the American Psychi-
BLutcher, C irclllation 37.
279 (1968).
11. R. C. Haynes, S. B. Koritz. F. G. Peron,
atric Assoc., Bal Harbour. Fla., 1969; W. E. J. Biol. Chemn. 234, 1421 (1959).
Btunney. Jr., D. L. NMurphy, F. K. Goodwin, 12. R. V. Farese, L. G. Linarelli, W. H. Glin-
G. F. Borge, Lancet 1970-I, 1022 (1970). mattn, B. R. Ditzion, M. 1. Patul, G. L. Pauk,
2. H. Cramer and W. Kuhlo, Acta Aeutrol. En doc,ini0ologt,
85. 867 (1969).
3.
Ps c/hiat. Belg. 67, 658 (1967).
M. I. Paul, B. R. Ditzion, G. L. PauLk, D. S.
13. P. S. Schonhofer, I. F. Skidatlore,
B. RZ. Ditzioni, G. L. PauLk, G. Krishna, B. B.
NM.
I. Patil,
3a,nowskv. Ainer. J. Psvtchiat. 126. 1493 BI-odie, unpublished data.
(1970); iM. I. Pauil, B. R. Ditzion, D. S. 14. R. Strom-Olsen and H. W. Weil-Mallherbe,
JanoxAskv. Lanicet 1970-I, 88 (1970). J. Mesit. Sci. 104. 696 (1958): A. Bergsman.
4. Y. H. Abdtdllah and K. Hamadah, lancet A ita Pst/vshitt. Neirol. Sc asitl. Stippl. 33,
1970-I 378 (1971)). S133 (1959); N. Shinfuku, 0. Mlichio, K.
5. HI. t. PaUl, H. C ramner, F. K. Goodswin, ibidl., Nl.>.ao. Yoniago Acta Med. 5, 1(09 (19'll);
p. 996; Arc/h. Gen1. Pstchiat., in prcss. R. B. Sloane, W. Htughes. NI. I.. HaLuSt, C'ani.
Ps
6. NI. 1. PauLl. B. R. Ditzion, G. L. PauLk. tc/hiat.
15. J. J.
Ass. J. 11, 6 (1966).
P/iar, uacologv 3, 148 (1971)). Schildkrauit, E. K. Gordotn, J. DLurell,
J. PsYc/ciat. Res. 3, 213 (1965).
7. G. 1. Gessa, J. Forn, A. Tagliamonte, G.
Krishnia, in Role of Cyclic AM5/P in NAelronal 16. W. E. Btinney. Jr., and J. NI. Davis, Arch.
Funizctioni, E. Costa and P. Green-ard. Eds. Gen. Psv chiat. 13, 483 (1965).
(Raven, New York, in pi-ess). 17. D. Eccleston. R. Loose, 1. A. Pttllar, R. F.
8. W. E. Bunney and D. A. Hambturg. Arc/h. Stigdcn, Lancet 1970-lII 612 (1971)).
18. G. R. Robison et al., ibid., p. 1028.
Ceti. PtYchiat. 9, 280 (1963); A. Beigel, D.
NluLrphy, W. E. BuLnlney, uLnpublished data. 23 July 1970; i-eviscd 14 October 197(1
Speech Perception in Infants
Abstract. Discriminiationi of synith1etic speechll .SOlunid was stldie(i in 1- anid 4-
mizonith-old infanlts. The speechl soiunds varied i alolig anl (colustic diuICInsionI pie-
viously shown'n to cute plionieniiic distiinctions amiiong the voiced an?6d voiceless stop
consonants ini adullts. Disciiin?tilability wtas ienas ured by ani increase ii conIdi-
tioiiedl responise rcate to a second speech souiCd a/f/er hiabitutatioln to the first speechi
sound. Recovery fromiz habituation tas greater for a givcn acoustic dif/ereiice
whi en the twvo stimulili were frcomii difJecient adlult ph1ontein ic cate(,tories thlcan wivhen
they were fi-oini the samze caltegory. The discontinuity in discriniiiiacltioni at the
regionr of the adirlt pllollenilic boundary t'as teikeii as ev,idence for categorical
)e-cetption.
In this study of speech perception, investigation of this nature (2) revealed
it was founid that 1- and 4-month-old that the perceptioin of this cue was very
infants were able to discriminate the nearly categorical in the sense that
acoustic cue underlying the adult pho- listeners couldl discriminate continuous
nemic distinction between the voi-cd variations in the relative onset of the
and voiceless stop consonants /b/ and first formlant very little better than they
/ Mp.loreover, and more important, could identify the sound paltterns abso-
there was a tendency in these subjects lutely. ThaLt is, listeners could rcadily
toward categorical perception: discrim- discriminate between the voiced and
ination of the same physical difference voiceless stop consonants, just as they
was reliably better across the adult would differentially label them, but they
phonemic boundary than within the were virtually unable to hear intra-
adult phonemic category.
phonemic differences, despite the faAt
Earlier research using synthetic that the acouLstic varitation xxvas the same
speech sounds with adult subjects un- in both conditions. The most measur-
covered a sufficient cue for the per- able indication of this categorical per-
ceived distinction in English between ception was the occurrence of a high
the voiced and voiceless forms of the peak of discriminability at the bouLndary
stop consonants, /b-p/, /d-t/, and Ig- between the voiced and voiceless stops,
k/. occurring in absolute initial position
and a nearly chance level of discrim-
(1). The cue, which is illustratcal in in,ability among stimuli that represented
the spectrograms displayed in Fig. 1, is acoustic variations of the same pho-
the onset of the first l'ormnant relative to Such
neme. categorical perception is
the second and third formaints. It is pos- not found with nonspeech sounds that
sible to construct a series of stimuli that vary continuously along physical con-
vary continuously in the relative onset tinua such as frequency or intensity.
time of the first formant, and to investi-
Typically, listeners are able to discrim-
gate listeners' ability to identify and inate many more stinmuli than they are
discriminate these sound patterns. An able to identify absolutely, and the dis-
303
 the special processing to which sounds
of speech are subjected and thus to be
characteristic of perception in the
@2 - F -2|
speech or linguistic mode (4).
Because the voicing dimension in the
F-1
stop consonants is universal, or very
nearly so, it may be thought to be
+ 10 msec
reasonably close to the biological basis
of speech and hence of special interest
3
Not all languages studied make use
of the three modal positions. English,
for example, uses only two locations, a
short lag in voicing and a relatively long
lag in voicing. Prevoicing or long voic-
ing lead, found in Thai, for example,
is omitted. Of interest, however, is the
fact that all languages use the middle
location, short voicing lag, which, given

F-3 to students of language development. certain other necessary articulatory

32
F-2
+ 100 msec
Fig. 1. Spectrograms of syntheltic speech
showing two conditions of voice (onset time
(VOT): slight voicing lag in the upper
figure and long voicing lag in the lower
figure. The symbols F-1, F-2, an(d F-3 rep-
resent the first three formants, tihat is, the
relatively intense bands of ener
speech spectrum. [Courtesy of L. Lisker
and A. S. Abramson]
criminability functions do not normally
show the same high peaks and low
troughs found in the case of the voic-
ing distinction (3). The stronl g and un-
usual tendency for the stop co onsonants
to be perceived in a categoricail manner
has been assumed to be the result of
75
20 D
0
U) 60
CL
U1)
o
c 45 _
C._
Though the distinctions made along the
voicing dimension are not phonetically
the same in all languages, it has been
found in the cross-language research of
Lisker and Abramson (5) that the
usages are not arbitrary, but rather
very much constrained. In studies of
the production of the voicing distinc-
tion in 11 diverse languages, then in-
vestigators found that, with only inor
exceptions, the various tokens fell at
three values along a single continuum.
The continuum, called voice onset time
(VOT), is defined as the time between
the release burst and the onset of
laryngeal pulsing or voicing. Had the
location of the phonetic distinctions
been arbitrary, then different languages
might well have divided the VOT con-
tinuum in many different ways, con-
strained only by the necessity to space
the different modal values of VOT suffi-
ciently far apart as to avoid confusion.
20 S O~~~~~~~~~~~
I
1-
I
I cedure developed by Siqueland (7).
I
Ie*.
I
I each infant, the presentation and inten-
events, corresponds to the English
voiced stop /b/, and one or both of
the remaining modal values. The acous-
tic consequences for two modes of pro-
duction are shown in Fig. 1; these cor-
respond to short and long voicing lags,
/b/ and /p/, respectively.
Given the strong evidence for uni-
versal-and presumably biologically de-
termined-modes of production for the
voicing distinction, we should suppose
that there might exist complementary
processes of perception (6). Hence, if
we are to find evidence marking the
beginnings of speech perception in a
linguistic mode, it would appear rea-
sonable to initiate our search with in-
vestigations of speech sounds differing
along the voicing continuum. What was
done experimentally, in essence, was to
compare the discriminability of two
synthetic speech sounds separated by a
fixed difference in VOT under two con-
ditions: in the first condition the two
stimuli to be discriminated lay on op-
posite sides of the adult phonemic
boundary, whereas in the second con-
dition the two stimuli were from the
same phonemic category.
The experimental methodology was a
modification of the reinforcement pro-
After obtaining a baseline rate of high-
amplitude, nonnutritive sucking for

O sity of an auditory stimulus was made

U)
%.-
contingent upon the infant's rate of
0
high-amplitude sucking. The nipple on
30 I which the child sucked was connected
.0

E 0
I
_ ,~~~~~~~~~~~~~~~IX
to a positive pressure transducer that
c

C
I provided polygraphic recordings of all
I
responses and a digital record of cri-
15 _ B 54 3 2 1 2 3 4~~~~ terional high-amplitude sucking re-
I sponses. Criterional responses activated
a power supply that increased the in-
I tensity of the auditory feedback. A
I I I ; ! , . I
rate of two responses per sec-
I I

B 5 4 3 2 1 1 2 34
sucking
B5 43 2 1 1 2 34 ond maintained the stimulus at maxi-
Time (min) mum intensity, about 75 db (13 db
Fig. 2. Mean number of sucking responses for the 4-month-old infants, as a function over the background intensity of
of time and experimental condition. The dashed line indicates the occurrence of the 62 db).
stimulus shift, or in the case of the control group the time at which the shift would The presentation of an auditory stim-
have occurred. The letter B stands for the baseline rate. Time is measured with ulus in this manner typically results in
reference to the moment of stimulus shift and indicates the 5 minutes prior to and an increase in the rate of sucking com-
the 4 minutes after shift.
304
SCIENCE, VOL. 171
pared with the baseline rate. With con-
tinued presentation of the initial stim-
ulus, a decrement in the response rate
occurs, presumably as a consequence of
the lessening of the reinforcing prop-
erties of the initial stimulus. When it 4-
was apparent that attenuation of the
reinforcing properties of the initial stim- 0
ZOD stimulation was made. The control
group served to counter any argument
that the increment in response rate as-
sociated with a change in stimulation
was artifactual in that the infants
tended to respond in a cyclical manner.
Eight infants from each age level were

randomly assigned to conditions 20D

ulus had occurred, as indicated by a 0.~~~~~~~~~~
and 20S, and ten infants from each age
decrement in the conditioned sucking
rate of at least 20 percent for two con-
secutive minutes compared with the im-
mediately preceding minute, a second
CsL
0~~~~~~0
level were assigned to the control con-
dition.
Figure 2 shows the minute-by-min-
ute response rates for the 4-month-old
auditory stimulus was presented with- W~ subjects for each of the training condi-
out interruption and again contingent 4 inonth% tions separately. The results for the
upon sucking. The second stimulus was -5~ ~ ~ ~ ~ younger infants show very nearly the
maintained for 4 minutes after which identical overall pattern of results seen
the experiment was terminated. Control with the older infants. In all conditions
subjects were treated in a similar man- as Fig. 3. The mean change in response rate at both age levels, there were reliable
a function of experimental treatments,
ner, except that after the initial de- shown separately for the 1- and 4-month- conditioning effects: the response rate
crease in response rate, that is, after old infants. (See text for details.) in the third minute prior to shift was
habituation, no change was made in significantly greater than the baseline
the auditory stimulus. Either an increase rate of responding (P < .01). As was
in response rate associated with a The main experiment was begun expected from the nature of the pro-
change in stimulation or a decrease of after several preliminary studies estab- cedure, there were also reliable habitu-
smaller magnitude than that shown by lished that both age groups were re- ation effects for all subjects. The mean
the control subjects is taken as infer- sponsive to synthetic speech sounds as response rate for the final 2 minutes
ential evidence that the infants per- measured by a reliable increase in the prior to shift was significantly lower
ceived the two stimuli as different. rate of sucking with the response-con- than the response rate for the third min-
The stimuli were synthetic speech tingent presentation of the first stimulus ute before shift (P < .01). As is ap-
sounds prepared by means of a parallel (P < .01). Furthermore, these studies parent from inspection of Fig. 1, the
resonance synthesizer at the Haskins showed that stimuli separated by dif- recovery data for the 4-month-old in-
Laboratories by Lisker and Abramson. ferences in VOT of 100, 60, and 20 fants were differentiated by the nature
There were three variations of the bi- msec were discriminable when the stim- of the shift. When the mean response
labial voiced stop /b/ and three varia- uli were from different adult phonemic rate during the 2 minutes after shift
tions of its voiceless counterpart /p/. categories; that is, there was reliable was compared with the response rate
The variations between all stimuli were recovery of the rate of sucking with a for the 2 minutes prior to shift, condi-
in VOT, which for the English stops change in stimulation after habituation tion 20D showed a significant incre-
/b/ and /p/ can be realized acoustically (P < .05). The finding that a VOT dif- ment (P < .05), whereas condition 20S
by varying the onset of the first formant ference of 20 msec was discriminable showed a nonsignificant decrement in
relative to the second and third form- permitted within-phonemic-category dis- responding (P > .05). In the control
ants and by having the second and criminations of VOT with relatively condition, there was a fairly substantial
third formants excited by a noise source realistic variations of both phonemes. decrement in responding during the first
during the interval when the first form- In the main experiment, there were 2 minutes of what corresponded to the
ant is not present. Identification func- three variations in VOT differences at shift period in the experimental condi-
tions from adult listeners (8) have in- each of two age levels. In the first con- tions. However, the effect failed to
dicated that when the onset of the first dition, 20D, the difference in VOT be- reach the .05 level of significance, but
formant leads or follows the onset of tween the two stimuli to be discrim- there was a reliable decrement when the
the second and third formants by less inated was 20 msec and the two stimuli mean response rate for the entire 4
than 25 msec perception is almost in- were from different adult phonemic minutes after shift was compared with
variably /b/. When voicing follows the categories. The two stimuli used in con- the initial 2 minutes of habituation
release burst by more than 25 msec the dition 20D had VOT values of +20 and (P < .02). The shift data for the
perception is /p/. Actually the sounds +40 msec. In the second condition, younger infants were quite similar. The
are perceived as /ba/ or /pa/, since the 20S, the VOT difference was again 20
patterns contain three steady-state form- msec, but now the two stimuli were only appreciable difference was that in
condition 20S there was a nonsignifi-
ants appropriate for a vowel of the type from the same phonemic category. In cant increment in the response rate
/a/. The six stimuli had VOT values of this condition the stimuli had VOT during the first 2 minutes of shift.
-20, 0, +20, +40, +60, and +80 values of -20 and 0 msec or +60 and In Fig. 3 the recovery data are sum-
msec. The negative sign indicates that +80 msec. The third condition, 0, was marized for both age groups. The mean
voicing occurs before the release burst. a control condition in which each sub- change in response rate (that is, the
The subjects were 1- and 4-month-old ject was randomly assigned one of the mean response rate for the initial 2
infants, and within each age level half six stimuli and treated in the same man- minutes of shift minus the mean re-
of the subjects were males and half ner as the experimental subjects, except
were females. sponse rate during the final 2 minutes
that after habituation no change in before shift) is displayed as a function
22 JANUARY 1971
305
of experimental treatments and age. 8. L. Lisker and A. S. Abramson, Proc. Int.
Congr. Phonet. Sci. 6th (1970), p. 563.
facilities of the Haskins Laboratories. We also
thank Drs. A. M. Liberman, L. 0. Mattlngly,
Analyses of these data revealed that the 9. Supported by grants HD 03386 and HID 04146 A. S. Abramson, and L. Lisker for their critical
magnitude of recovery for the 20D con- from the National Institute of Child Health comments. Portions of this study were pre
and Human Development. P.J. and J.V. were sented before the Eastern Psychological As-
dition was reliably greater than that for supported by the NSF Undergraduate Partici- sociation, Atlantic City (April 1970).
the 20S condition (P < .01). In addi- pation Program (GY 5872). We thank Dr. F.
S. Cooper for generously making available the 14 September 1970 .
tion, the 20D condition showed a great-
er rate of responding than did the con-
trol condition (P < .01), while the dif-
ference between the 20S and control Randomization and the Draft Lottery
conditions failed to attain the .05 level
of significance. Abstract. Fifty "random permutations" were prepared for use by the Selective
In summary, the results strongly indi- Service System as a basis for a two-stage randomization that preceded the lottery
cate that infants as young as 1 month drawing on 1 July 1970. This report identifies the permutations used. It also
of age are not only responsive to speech gives the orders in which calendar dates and numbers were put into and drawn
sounds and able to make fine discrim- from two drums and the correlations between them.
inations but are also perceiving speech
sounds along the voicing continuum in The Selective Service System asked dars" and 25 "random permutations"
a manner approximating categorical the National Bureau of Standards of the numbers 1 to 365 for use in
perception, the manner in which adults (NBS) to prepare 25 "random calen- connection with preparations for the
perceive these same sounds. Another draft lottery conducted on 1 July 1970,
way of stating this effect is that infants which determined the order in which
are able to sort acoustic variations of Table 1. Coefficients of correlation (r) and men born in 1951 would be called for
adult phonemes into categories with their significance levels (P) for pairs of induction into military service (1).
relatively limited exposure to speech, permutations used in and produced by the In order that all details of the prep-
draft lottery procedures. I, Calendar dates
as well as with virtually no experience and ranks in the order loaded into capsules; arations be reproducible, the permuta-
in producing these same sounds and permutations 53 and 43, respectively. II, tions were drawn from published tables
Calendar dates (ranks) in the order of entry
certainly with little, if any, differential into drum, computed from permutation 53 [tables 8 and 9 in (2), pp. 152-229].
reinforcement for this form of behav- (43) and permutation 46. III, Calendar dates Table 8 contains 38 permutations of
ior. The implication of these findings is (ranks) in the order drawn from the drum; the numbers 1 to 500; by omitting
permutation shown in Fig. 2 (Fig. 3).
that the means by which the catego- numbers greater than 365, 38 permu-
rical perception of speech, that is, per- Calendar dates Ranks tations of the numbers 1 to 365 were
Pair
ception in a linguistic mode, is ac- r P r P* formed. Table 9 contains 20 permuta-
complished may well be part of the I-II .0469 .37 .0706 .18 tions of the numbers 1 to 1000; by
biological makeup of the organism and, I-Ill -.0493 .35 -.0133 .80 considering numbers from 1 to 365 and
moreover, that these means must be- lI-Ill .0387 .46 -.0573 .27 those from 501 to 865, 40 permutations
operative at an unexpectedly early age. * Probability that correlation between two ran-
of the numbers 1 to 365 were formed.
dom permutations exceeds (in magnitude) the
PETER D. EIMAS observed value. An assortment of tests for randomness
EINAR R. SIQUELAND
PETER JUSCZYK
JAMES VIGORITO I0 NOVEMBER 3n 30 tIOVEMRER 1 1/30 33'9
Department of Psychology, 2. AUGIUST I 8 I A AUGUST a/ l8 230
Brown University, 3. APRIL 8 8 APR I L 14 / A 98
4 JtNF 14 1'I J lINE .6/14 1 65
Providence, Rhode Island 02912 22 OCTOB3ER 10o/22 295
5. OCTOBER 22
JAtJUARY 9 A JANUARY 1/8 8
References and Notes
1 ,
DECEMBER 12 1 2 DECEMBER l2Xl2 3'16
1. A. M. Liberman, P. C. Delattre, F. S. Cooper, AUGUIST 27 27 AUJGUST 8/27 239
Language and Speech 1, 153 (1938); A. M. 9/2n 263
Liberman, F. Ingemann, L. Lisker, P. C. Delat- 9. SEPTEMRER 20 2nr SEPTEMlER
tre, F. S. Cooper, S. Acoust. Soc. Amer. 31, ln SEPTEMBER 8 SEPTEMBFR 9/e 251
1490 (1959). It should be emphasized that the
cues underlying the voicing distinction as dis-
1I0. DECEMBER 31
OCTOBER 13
31
13
DECEMBER
oCTroBER
1 2/3 1
1 0/ 1 3
365
286
cussed in the present report apply only to 12.o
sound segments in absolute initial position. 13 . FEBRUJARY 2 2 FEBRuARY 2/2 33
2. A. M. Liberman, K. S. Harris, H. S. Hoffman 15 MAY 5/ I s 1 35
H. Lane, J. Exp. Psychol. 61, 370 (1961). MAY 15
3. P. D. Eimas, Language and Speech 6, 206 Is. OCTORER 20 20 nCTOBER 10/2n 293
(1963); G. A. Miller, Psychol. Rev. 63, 81 16o OCTORER 4 14 ncTOBER I 0/'4 277
(1956); R. S. Woodworth and H. Schlosberg, 1 6 APR I L '4/ 1 6 1 06
Experimental Psychology (Holt, New York, I 7 . APRIL 16
1954). 15. NOVEMBER 3 3 NO V EM B ER 1 1/3 3n7
4. A. M. Liberman, F. S. Cooper, D. P. Shank- JlJNE 20 2n j IJ NE 4/20 1 71
weiler, M. Studdert-Kennedy, Psychol. Rev. 74, 1 9.
431 (1967); M. Studdert-Kennedy, A. M. Liber- A'JGUST 2 3 23 AUGuST 8/23 235
man, K. S. Harris, F. S. Cooper, ibid. 77,
234 (1970); M. Studdert-Kennedy and, D.
21n. MAY 3 3 M AY 5/3 123
282
Shankweiler, J. Acoust. Soc. Amer., in press. 22. OCToBER 9 9 oCTO8ER 1 0/9
5. L. Lisker and A. S. Abramson, Word 20, 23. ftAY 2 2 MAV 5/2 122
384 (1964). 24 . JUNE I l JtJNE 6/ Il 52
6. P. Lieberman, Linguistic Inquiry 1, 307 (1970).
7. E. R. Siqueland, address presented before the Fig. 1. Part of a random calendar ("calendar number 53") prepared by NBS, that
29th International Congress of Psychology, was used for loading dates into capsules, showing the redundant format used for
London, England (August 1969); - and
C. A. DeLucia, Science 165, 1144 (1969). printing. [Derived from pages 204-205 of Moses and Oakford (2)]
306 SCIENCE, VOL. 171