Regulation of nursing in chimpanzees

Regulation of Nursing in Chimpanzees

HERMAN DIENSKE WILLEM VAN VREESWIJK

Primate Center TNO 151 Lange Kleiweg, 2288 GJ Rijswijk

The Netherlands

The regulation of nursing was studied in captive chimpanzees from birth to 6 months of age. It was asked whether regulation was predictable or timing was irregular. A search for unimodal frequency distributions resulted in a distinction among nursing bouts, nursing episodes (bouts with brief interruptions) and nursing pauses. The frequency distributions of these types were either normal with very large standard deviations or exponential (i.e., randomly terminated). This implies a very irregular timing. Longer nursing episodes were followed by somewhat longer pauses; pauses with daytime sleep (that were relatively long) were followed by longer nursing. However, these regulatory effects were only slight. Most of this loose regulation of nursing was due to the infant, as the mothers did not put the infant on the breast and usually were permissive. Comparisons with humans suggest a similarly loose organization as well as frequent feeding in societies that provide relatively unlimited access to the breast. The regular 4hr , meal-like schedule in industrial countries seems mainly to be due to human mothers and their advisors.

Human adults tend to eat at rather fixed times of the day. In industrialized countries, many mothers nurse their babies also after rather constant time intervals. In other cultures, however, the timing of breast feeding is more relaxed. In view of this diversity, it is interesting to know how nursing is regulated in chimpanzees. Such knowledge obviously does not provide norms for humans, but interspecies comparisons may indicate which human habits are novel.

Timing of nursing may be described by durations of feeding and pauses. If nursing occurs very regularly, these durations vary only slightly around a mean value. Regulation also implies that large meals are followed by longer pauses than small meals and vice versa. Absence of such a deterministic regulation means that initiation and termination of nursing is independent of the durations of nursing and pauses in between.

In chimpanzees, the durations of nursing and the pauses in between have been described as highly variable (Van Lawick-Goodall, 1967; Nicolson, 1977; Plooilj, 1984). Brief nursing bouts were very common; bouts as long as 7-12 min have

~~

Reprint requests should be sent to Herman Dienske, Primate Center TNO, P.O. Box 5815, 2280 HV Rijswijk, The Netherlands.

Received for publication May 28, 1985 Revised for publication June 16, 1986

Developmental Psychobiology 20(1):71-83 (1987) 8 1987 by John Wiley & Sons, Inc. CCC OO12-1630/87/01OO7 1- 13504.00

72 DIENSKE AND VAN VREESWIJK

only occasionally been observed. The mean nursing bout length was between 2 and 3 rnin during the first week or month and about 2 rnin at later ages (Van Lawick-Goodall, 1967; Plooij, 1984). However, nursing is often briefly inter- rupted. Nicolson (1977) disregarded interruptions shorter than 1 rnin and found bout lengths that were about 1 rnin greater.

The time intervals between nursing bouts were also highly variable in duration. Brief intervals were common; long pauses were scarce and lasted about 90 rnin maximally (Van Lawick-Goodall, 1967; Plooij, 1984).

The enormous variability of nursing bouts and intervals complicates a judi- cious description of the timing of nursing. It will be necessary to establish whether there is a distinction between brief interruptions and longer pauses. In addition, the influence of the frequent spells of daytime sleeping on nursing regulation must be accounted for. Finally, it is not known whether nursing regulation in chimpanzees includes correlations between durations of subsequent nursing bouts and pauses.

This paper describes a detailed analysis of these aspects of nursing regulation by captive chimpanzees during the first half year of life. Comparisons will be made with the various styles of breast feeding that are found among human communi- ties.

Methods

Subjects and Behavior Recording The behavior of chimpanzee mother-infant pairs (n = 9) was observed while

they remained alone in a familiar, small cage. At other times, the animals were together with a few other mother-infant pairs in a large cage. Observations were made from birth up to about 4 (n = 3) or 6 months (n = 3); 3 other pairs were studied from 4 to 6 months of age. The intervals between observations gradually increased with age from 1 to 4 weeks. The total number of observation periods was 74. The observation time was 1 hr 20 min and took place between 12.00 and 13.20 hours, a quiet period in our chimpanzee building. Data were collected, among others, on body contact, rooting, nursing, body movements, and eyes being closed. The moments of onset and termination of these categories were recorded with a keyboard system; the accuracy was 1 sec. Simultaneous occur- rence of different categories was recorded as such.

Models for the Quantification of Timing The timing of nursing can essentially be described with the durations of nurs-

ing bouts and the intervals between them. Two simple models are available for this description.

1. One model is a deterministic cycle, with normally distributed durations of nursing bouts and nursing pauses. Such a cycle can be represented quantitatively with 2 means and 2 standard deviations.

2. The other model comprises a fully stochastic alternation between 2 states: nursing and pauses. In a deterministic cycle, termination probability steadily in- creases. The word “stochastic,” however, implies that termination is equally likely at any moment after the start of a bout. In other words, the termination rate of a state is constant: every time unit the same proportion of the not yet termi-

REGULATION OF NURSING IN CHIMPANZEES 73

nated bouts of a state ends. For an easy judgment of the cofistancy of a termination rate over the time elapsed since the onsets of bouts, it is usual to plot a log survivorship curve. If the stochastic model applies, this is a straight decreasing line of which the constant slope corresponds with a constant termination rate. The frequency distribution D(t) is exponential: D(r) = N . R exp(-Rt), where N is the sample size, R the termination rate, and t time. Tests for exponentiality are described in Dienske, Metz, Van Luxemburg, and De Jonge (1980). An exponential distribution has a termination rate which is equal to the inverse of the mean bout length. Disparities from linearity will be specified by a coefficient of Linearity Deviation: L.D. = -2(N In M - C In xi)/N + 1.154, where M is the mean bout length and xi the length of bout i. Negative values of the L.D. correspond to a decreasing termination rate, that is a decreasing slope of the log survivor (see Dienske et al., 1980, p. 89). Further details are described by Metz (1974), Dienske and Metz (1977), and Haccou, Dienske and Meelis (1984).

Our procedure will be to test whether nursing bouts and pauses differ significantly from normal or exponential distributions. There are 4 possibilities. (1) If a frequency distribution significantly differs from exponentiality but not from normality, the sample size is reasonably large, and the graphical representation con- firms nomality, it is concluded that the type of interval is deterministically regulated. That is, the likelihood to terminate the interval is greatest near the mean duration. (2) If a frequency distribution significantly differs from normality but not from exponentiality, the sample size is reasonably large, and the log survivorship curve is approximately linear, it is concluded that there is no deterministic regulation. That is, termination is equally likely at each moment after the onset of the type of interval. (3) If the difference with both exponentiality and normality is not significant, and the sample size is small, no firm conclusion can be drawn. (4) If the difference with both types of distributions is significant, inspection of the histogram and log survivor of the data'may suggest that the collection of time intervals consists of a few components. The identification of the components may be based on behavioral criteria. The procedure is to search for certain events or acts that characterize one component and not the other. If a candidate characteristic is found, the set of bouts is split on the basis of presence or absence, whereafter it is tested whether the subsets do not significantly differ from normality or exponentiality. If this is successful, the identifying events may be used to interpret and name the subsets. Splitting data sets results in reduction of sample size. Conclusions on the nature of the resulting distributions are only meaningful if the sample size is large and the fit is good. If significant differences persist after splitting, other types of distributions may be considered.

Results

Time of Nursing The frequency distribution and log survivor of bouts during which the infants

were on the nipple are given in Figure 1. Test results are given in Table I . The frequency distribution of on-nipple bouts clearly was not normal. The large number of short bouts and the gradual decrease of frequencies with bout length suggest an exponential distribution. However, the distribution significantly differed from exponentiality as well (Table 1). The log survivor demonstrates an excess of bouts shorter than about 30 sec. These bouts were identified as on-nipple bouts without


-150

E -100 :

P c

. so

. - a

looo]l'aII data

- n 5 100 c -

1 0 5 10

bout length lmin l

Fig. 1. Log survivorship curves and frequency distributions (histogram) of on-nipple bouts with sucking (i.e., nursing) and without sucking. See Table 1 for statistics.

sucking. So the set of on-nipple bouts was split into those without and with sucking; these two splitting products are also represented separately in Figure 1 and Table 1. On-nipple bouts with sucking did not differ from exponentiality. This is not due to reduced sample size as followed from repeated random sampling of

TABLE 1 . Timing of Nursing in Chimpanzees.

Attributes of the Time Intervals Used for the Description of the

Exponential Distribution Normal Distribution

Mean Linearity Test Mean f SD Test N (sec) Deviation of Fit (set) of Fit

On-nipple All bouts No sucking With sucking = nursing

Nursing gaps All Interruptions Pauses

Nursing episodes Eyes closed

All Sleep bouts

Awake From rooting to nipple

All Finds nipple quickly (<30 sec) Finds after trying (30 to 300 sec) Gives up (>300 sec)

312 42

270

20 1 106 95

140

700 128 63

193 145

29

19

9 104

64

154

909 979

5

124

1530

-0.5 -0.3 -0.1

-1.9 -0.3 +0.7 +0.3

-2.1 +0.2 +o.o

-4.9 -0.1

+0.1

-0.1

*** n.s. n.s.

*** n.s. ***

*** *** *** n.s.

* 154 t 123 * 1474 t 940 n.s. ***

*** n.s. n.s.

*** n.s.

*** ** **

*** ***

n.s. n.s.

n.s. n.s.

* p < 0.05, ** p < 0.01, *** p < 0.001. The Linearity Deviation is zero if a distribution is exponential and the log survivor hence is a

straight line (constant slope). A positive (negative) value indicates an increasing (decreasing) slope of the log survivorship curve (see Dienske et al., 1980). The Kolmogorov-Smirnov one-sample test was used to test for normality.


’ O o o l n i n t c r r u p t i o n s I l l 0

0 30 K O 90

bout length (min)

Fig. 2. Log survivorship curve and frequency distribution of the time intervals between 2 nursing bouts. By means of behavioral criteria, brief interruptions were distinguished from pauses; separate log survivors and frequency distributions of interruptions and pauses are also illustrated. In addition, a normal distribution with the mean and S.D. of the pauses is given.

270 bouts from 312 bouts (p always less than 0.0005). The lack of significances for on-nipple without sucking, however, may be due to a reduced sample size.

On-nipple without sucking was both short (mean = 9 sec) and infrequent (only 13% of all on-nipple bouts). Sucking milk, therefore, was the rule if the infant was on the nipple. Nipple bouts with sucking will be termed “nursing.” As these were approximately exponentially distributed, nursing in chimpanzees had not the character of a meal with a rather fixed duration.

The frequency distribution of the time intervals between nursing is given in Figure 2. This distribution differed from normality as well as exponentiality (Table 1).

A short component was identified with behavioral criteria. From the observations and inspection of the behavior records, it followed that infants often briefly released the nipple and resumed nursing without doing anything else in particular. It also happened that the infant lost the nipple because of movements of the mother; it sometimes vocalized in response. As soon as the mother moved the infant near the nipple or took a position that allowed access to it, the infant resumed nursing. The set consisting of these 2 types of nursing interruptions was about exponentially distributed with a mean duration of 1 min (Table 1).

The remaining nursing pauses, however, fitted a normal distribution with a mean duration of half an hour (Table 1). The normality of this distribution implies some regularity in nursing. However, the standard deviation of nursing pauses was quite large, namely a quarter of an hour.

The fit for interruptions and pauses is not due to reduced sample size. This follows from Figure 2 as well as drawing random samples of n = 106 or 95 from the mixed data (n = 201).

A substantial proportion of the nursing bouts was followed by a brief interruption and not by a long pause. It is reasonable to regard nursing bouts separated by a brief interruption as one nursing episode. The interruption time was not included in these episodes, so that the time spent in obtaining milk corresponds to the duration of the nursing episodes. The resulting frequency distribution differed significantly from exponentiality as well as normality (Fig. 3 and Table I ) . There


0 5 10

bout length ( m m l

Fig. 3. Log survivor, frequency distribution, and fitted normal distribution of nursing episodes, i.e., nursing disregarding brief interruptions.

were too few short bouts for an exponential distribution to fit; the distribution was too much skewed to the left in comparison with a normal distribution. Since this suggests a heterogeneous mixture of data, the frequency distributions were considered for each month of age. For each of the first 4 months, termination rates were only slightly increasing (L.D. +0.2, +0.2, +0.2 and +0.4, N between 19 and 29, p > 0.26). These distributions differed neither from exponentiality nor from normality. In contrast, for months 5 and 6, the distributions differed from exponentiality (L.D. +0.7 and +0.8, N = 20 and 19; p < 0.03) and were closer to normal (0.20 > p > 0.15). In conclusion, termination of nursing episodes rather gradually shifted from a stochastic to a relatively deterministic process. However, at ages over 4 months, the variability of episodes was still large (105 ? 56 sec, mean ? S.D.) .

To summarize, nursing bouts were approximately exponentially distributed, nursing episodes (disregarding brief interruptions) shifted with age from exponential to normal and nursing pauses were about normally distributed with a very large standard deviation. These aspects of the timing of nursing demonstrate only a minor degree of regulation.

Another aspect of regulation is the relationship between the duration of nursing episodes and the pauses between them. As hunger probably appears sooner after the infant has taken a relatively small volume of milk, one expects a positive correlation between the durations of nursing episodes and nursing pauses. How- ever, it seems likely that daytime sleeping considerably influences digestion of milk as well as the development of hunger. For this reason, the timing of sleep is examined before the regulation of nursing is analysed further.

Daytime Sleep The recorded behavior indicative of sleep was closed eyes and the absence of

limb movements. The log survivorship curve of eyes-closed demonstrates an almost continuously decreasing slope (Fig. 4, Table 1). This type of log survivor may result from a mixture of time intervals that are sampled from a large number


- Y n 5 1 0 0 - c - ,

I

B : E f L

1 0 -

E

1 -

1000 ,

:,eyer closed

50

b “ Y

U a

E 25

0 0 60 90

bOu1 lenglh (mln)

Fig. 4. Log survivors of eyes-closed and sleeping. Sleeping is defined as eyes-closed bouts with immobility disregarding brief interruptions. The frequency distribution of sleep bouts is also given.

of exponential distributions with diverse termination rates. This suggests that it may be difficult to find identifiers that can be used to split the set of eyes-closed bouts into a small number of components. We therefore used a different technique to define sleep bouts.

Inspection of actograms, in which bars were plotted at the times that eyes- closed and movements were recorded (examples are given below and in Dienske, 1984) revealed similarities with the behavioral states of human babies. For this reason we used the moving frame method described by Prechtl and O’Brien (1982) to identify sleep bouts. An eyes-closed bout was regarded to be sleep if limb movements lasted shorter than 2 sec and had gaps longer than 5 sec. Subsequent eyes-closed bouts were linked if the intervening eyes-open bout lasted less than a minute and only contained limb movements shorter than 2 sec with gaps longer than 5 sec. These arbitrary time limits were chosen as such because the number of doubtful cases was then minimal. The frequency distribution of sleep bouts thus defined did not differ from exponentiality (Fig. 4, Table 1). Initially, we thought that eyes-closed bouts should have a substantial minimal duration in order to be regarded as sleep. However, then the log survivor was horizontal for the duration of that limit and subsequently decreased linearly. This was judged to be too artificial and the minimal duration was chosen as short as possible, namely 2 min. If a shorter value was used, the distribution showed an excess of short bouts. It must be realized that this limit was only chosen in order to find a simple frequency distribution and not because we “believe” that distributions “should” be exponential. It will be clear that the procedure used for finding sleeping bouts is more arbitrary than the utilization of an identifying activity.

The intermittent nonsleeping bouts also appeared to fit an exponential distribution (Table 1). Nonsleeping will be simply called waking, although some of such bouts consisted of drowsiness as was apparent from brief and recurrent closing of the eyes and relative immobility. The mean durations of completely observed sleeping and waking bouts each were about 15 min. This is an underestimation, as incompletely observed bouts of sleeping as well as waking each were on average longer than the completely observed bouts.


TABLE 2 . and Nursing Pauses.

The Influence of Daytime Sleep on the Durations of Nursing Episodes

Preceding r Nurse r Following Nursing (nurse- Pause (pause- Nursing

(min) pause) (min) nurse) (min) ~~~ ~ ~

Nursing pause

Nursing pause

Difference between data for

with sleep (n = 33) 2.9 +0.30 35.9 +0.22 3.4

without sleep (n = 58) 2.3 +0.21 23.8 -0.15 2.1

pauses with and without sleep *** ** n.s.

** p < 0.01. *** p < 0.001.

Correlations between Durations of Nursing Episodes and Pauses Nursing pauses with sleep were, as one would expect, longer than those

without sleep (Table 2). Pauses with and without sleep tended to be longer if the preceding nursing episode had lasted longer ( r = +0.30, respectively +0.21). However, this positive relationship was only significant if data for all pauses were pooled ( r = +0.28; n = 91; p < 0.05). After a pause with sleeping, nursing lasted significantly longer (3.4 min) than if the pause did not include sleep (2.1 min). Correlations between pauses and the following nursing episodes were not significant in either case (Table 2). These results demonstrate a weak but significant regulation of nursing: Nursing pauses were somewhat longer after long nursing episodes and after long pauses with sleep the infant suckled a little longer.

Simultaneity of Nursing and Sleep Was there an association between the timing of nursing and daytime sleep?

Relevant data are schematically represented in Fig. 5. In most cases, nursing took place randomly between 2 sleeping bouts: the

intervals from sleep to nursing (n = 44) and vice versa (n = 50) both were about exponentially distributed. Mean durations of these intervals were almost equal: 9 and 11 min. This implies that nursing was as much associated with the end as with the next onset of sleep.

In a smaller number of cases, nursing occurred close to either the termination or the onset of sleep (see Fig. 5) . Only 4 of the nursing bouts entirely coincided with sleep; these were situated near the onset or end of sleep. Partial overlap was

1 2 - 1 6-

deep - 1 1 5- sleep m

111 V I 50 -

10 mi"

Fig. 5. Schematic representation of the position of nursing with respect to daytime sleep. Bouts and intervals are given proportional to their mean durations. Figures are observed frequencies. More than one bout of nursing could occur between 2 sleeping bouts.


60 90 30

bout length I m i n l

Fig. 6. Log survivor of the time interval between searching for the nipple (rooting) and nipple contact. Components are specified in Table 1 .

more common. In 16 cases, the interval between nursing and sleep was very short (distinguished from the longer intervals by inspection of the log survivor).

Rooting and Nipple Access

Rooting, the characteristic sideways movements of the head oriented toward the nipple, was common. The success of rooting can be derived from the time between rooting and the start of on-nipple. The log survivor of this latency is given in Figure 6. Three about linear components were found (see Table 1): (1) almost immediate nipple access, i.e., within 30 sec; (2) some persistent trying that was successful only after some time, namely 30 to 300 sec; (3) giving up, as nipple contact follows after more than 300 sec; The distributions of the 3 components did not differ from exponentiality. In this case, no identifying event could be detected. The time limits of 30 and 300 sec were found by moving these until the deviation from exponentiality was minimal. The distributions of the 2 longer components neither differed from normality, but the fit was worse (p < 0.15); moving time limits did not improve the fit. If it is assumed that the 2 longer components were exponentially distributed at values less than 30 and 300 sec, respectively, their relative frequencies would be somewhat greater than those given in Table 1. The hypothetical frequencies were estimated by extending the log survivors linearly to the left. Under this assumption, 65% of the cases of rooting were almost immediately followed by nipple contact, 22% after some time and 13% not at all.

Synopsis of Types of Time Intervals

Examples of the various kinds of time intervals distinguished above are given in Figure 7. This is an actogram derived from an actual behavior record. As no record contained all the relevant features, some were artificially added. These additions include the interruption in the 1st nursing bout and the interruptions and movements during the 3rd sleep bout.

Figure 7 illustrates that a nursing bout requires sucking and that a brief interruption of nursing separates 2 bouts in 1 episode. It also demonstrates that brief


rooting on-nipple sucking nursing bout nursing episode nursing pause

creep

eyes closed sleep awake

move

I , I

I. I

m I

I. I

I I

- I I I I-. I I ,

I 4 I , I I . .I - , I . II I-- - - - - - I 1

and scattered eyes-closed bouts are not regarded to be sleep. Finally, it shows that brief opening of the eyes or incidental short movements are not regarded as interruptions of sleep.

Changes with Age Nursing and sleep decreased during the first half-year of life. The length of

nursing bouts decreased slightly but significantly with age (r = -0.16, n = 270, p < 0.001); the decrease in the duration of nursing episodes was greater ( r = -0.34, n = 140, p < 0.001). Nursing pauses became somewhat longer, but the correlation with age is not significant. The percentage of the observation time spent in nursing is determined by durations of bouts and pauses. It is obvious therefore that this percentage more strongly changed with age (r = -0.39, n = 74) than did bouts or pauses. All of these changes during the first half-year of life were approximately linear. Nursing bouts decreased from 2.1 to 1.3 min and nursing episodes from 3.6 to 1.8 min. Pauses increased from 24 to 33 min. Total nursing duration changed from 1 to 0.5% of the observation time. Nursing bout frequency remained practi- cally constant ( r = -0.08) at 2.3 t imesh .

Daytime sleep also decreased with age. Sleep bout durations tended to decrease (r = -0.14, n = 128, n.s.) and wake bouts to increase ( r = +0.24, n = 63, n.s.). The percentage of time spent in sleeping, which results from these 2 bouts as well as from incompletely observed sleep and wake bouts, clearly decreased ( r = -0.46, n = 74, p < 0.001). This decrease was not linear. During the 1st month, the mean sleeping time was 50%. For the 2nd month, this value was only 13%. Sleep subsequently decreased to 7% of the time at 6 months of age. Observation periods without sleep were more common in older infants.

Rooting frequency decreased linearly and significantly with age from 5.2 to 2.1 t i m e s h (r = -0.23, n = 74, p < 0.05). The main age effect on the time between rooting and on-nipple was that the intermediate component (finding the nipple after some attempting) was almost absent after 2 months of age. This is probably connected with the fact that the chimpanzee infants were then capable of locomotion.

The analysis given above was applied to pooled data from 9 individuals over an age range of 6 months. If either individual differences or changes with age were considerable, it would not be possible to detect frequency distributions that do not


significantly differ from normality or exponentiality. Individual differences and changes with age mostly were not large enough to produce mixtures of frequency distributions that differed from exponentiality or normality. The only exception concerned nursing episodes, that, as described, were intermediate between an exponential and normal distribution during the first 4 months, but apparently normal thereafter. This change contributed to the decrease in nursing episode duration with age.

Discussion A quantitative description of the regulation of nursing in chimpanzees is pre-

sented. The timing of nursing differed little from a random process, as nearly all of the time intervals involved were randomly terminated and correlations between adjacent nursing bouts and pauses were minor. Only a few weakly deterministic (i.e., regulatory) aspects were found. Nursing pauses were normally distributed, but the very large standard deviation implies little predictability. Nursing episodes (i.e., brief interruptions disregarded) were terminated approximately at random during the first 4 months of age; later, nursing episodes were normally distributed with a large standard deviation. Longer nursing episodes were followed by longer pauses, and nursing was longer in duration if the preceding pause was prolonged by daytime sleep. These significant effects were, however, slight. It is concluded that the regulation of nursing in chimpanzees had little similarity with a regular meal-type pattern.

Data from studies of free-living chimpanzees are compatible with this conclusion inasmuch as comparable data were given (Van Lawick-Goodall, 1967; Plooij, 1984). The same applies to the findings by Nicolson (1977) on captive chimpanzees. Horvat, Coe, and Levine (1980) and Miller and Nadler (1981) quantified nursing as one-zero scores per minute, which is not sufficiently accurate for comparison. This is also the case with data on gorillas given by Hedeen (1980) and Hoff, Nadler, and Maple (1981). Mufiiz (1983) found for one orang-utan a mean nursing bout length of 67 sec; 5-4% of the time was spent in sucking. This might suggest that orang-utans nurse shorter and more frequent than chimpanzees, but the data were limited.

Although our findings on mean nursing and pause durations are similar to those from field studies, some of the more detailed results on regulation may be at variance with data from the natural situation. One limitation of our data is that they were collected only between 12.00 and 13.20 hours. Furthermore, durations of longer bout types such as sleep, wake, and nurse pauses were underestimated because the observation time (80 min) tended to leave longer durations more often unfinished so that these could not be included.

The role the chimpanzee mothers played in the timing of nursing was very passive. Chimpanzee mothers typically did not put the infant on the nipple; if the infant was near the nipple it initiated sucking on its own. This happened almost immediately in 65% of the cases of rooting. However, in the Arnhem Zoo we observed that a mother put her infant back on the nipple after a jerky movement by her had terminated nipple contact. Such a maternal initiative of nipple contact is apparently rare. Many nursing bouts started briefly after a change in the posture of a mother. If the infant was unable to reach the nipple, it sometimes vocalized. In only 13% of the cases was searching for the nipple unsuccessful. See Van Lawick-Goodall (1967) and Dienske (1984) for further details.


The maternal influence on the regulation of breast feeding by humans, in contrast, is often strong. The timing of feeding varies from being very regular and meal-like to relatively unlimited access to the nipple.

In industrial countries, mothers are usually advised to feed their babies every four hours. If a baby cries earlier, it is assumed not to be hungry, if it sleeps at the end of the 4-hr period, it is awakened (Wright, Fawcett, & Crow, 1980).

A less rigid approach i s “self-demand” feeding. This regime also resulted in an approximately 4-hr feeding rhythm (Morath, 1974; Salariya, Easton, & Cater, 1978; Verronen, Visakorpi, Lammi, Saarikoski, & Tamminen, 1980; Wright et al., 1980). When this schedule is in use, feeding episodes last about 22 2 7 min (mean 2 S.D.; Thoman, Leiderman, & Olsen, 1972; Thoman, 1975; Carlsson, Fager- berg, Horneman, Hwang, Larsson, Rodholm, Schaller, Danielson, & Gundewall, 1978; Rybski, Almli, Gisel, Powers, & Maurer, 1984). These episodes consist of 1-6 sucking bursts (Hwang, 1978; Wright et al., 1980). Slightly more than half of the feeding episode is spent in actual sucking (Thoman et al., 1972; Thoman, 1975; Hwang, 1978; Alberts, Kalverboer, & Hopkins, 1983; Rybski et al., 1984). The number of episodes is 5 to 7 per 24 hr.

It is questionable how much of this “demand” can be ascribed to the baby himself. As the baby must usually be taken from a crib, it is the care giver who decides whether the baby’s signal is a “demand.” Such a decision may easily be influenced by the opinion that a 4-hr schedule is the norm.

La Leche League mothers restrict access to the breast as little as possible. For infants younger than 3 months, the number of suckling episodes per 24 hr was 15 2 8.5; range 2 to 35 (Cable, & Rothenberger, 1984). These data include nights. Brazelton, Robey, & Collier (1969) observed in Mexican Indians a nursing frequency of 1.3 2 0.6 times per hour and a mean duration of 6.2 2 1.5 min during daytime. Among the !Kung hunter-gatherers in Africa, nursing was even much more frequent (Konner & Worthman, 1980). Sucking took place in 4 bouts per hour; the mean bout length was only 2 min. The actograms given by these authors suggest an absence of nursing episodes with several sucking bursts, separated by long pauses. Sucking bouts seemed to be randomly scattered over the observation periods.

These profound differences among subcultures or cultures are probably due to the degree to which the baby is permitted to regulate nipple access. It seems likely that a !Kung mother leaves much of the regulation of nursing to her baby. This results in a pattern that is strikingly similar to the loose organization we found in the chimpanzee. In cultures that tend to control many aspects of life, the balance is strongly on the side of the caregiver. The differences among cultures and subcultures suggest that the rigid control on nursing in most industrialized societies is much more arbitrary than is usually realized.

Acknowledgments We are grateful to L. G. Ribbens for assistance, C. Goosen for comments on

an earlier draft, D. de Keizer for typing the manuscript, and E. J. van der Reyden for preparation of the figures.

References Alberts, E., Kalverboer, A. F., and Hopkins, B. (1983). Mother-infant dialogue in the first days of life:

an observational study during breast-feeding. J . Child Psychol. Psychiatr., 24: 145-161.


Brazelton, T. B., Robey, J. S., and Collier, G. A. (1969). Infant development in the Zinacanteco Indians of Southern Mexico. Pediatrics, 44: 274-293.

Cable, T. A., and Rothenberger, L. A. (1984). Breast-feeding behavioral patterns among L a Leche League mothers: a descriptive survey. Pediatrics, 73: 830-835.

Carlsson, S. G., Fagerberg, H., Horneman, G., Hwang, C-P, Larsson, K., R(idholm, M., Schaller, J.. Danielsson, B., and Gundewall, C. (1978). Effects of amount of contact between mother and child on the mother’s nursing behavior. Dev. Psychobiol., 11: 143-150.

Dienske, H., and Metz, J. A. J. (1977). Mother-infant body contact in macaques: a time interval analysis. Biol. Behau., 2: 3-37.

Dienske, H., Metz, J. A. J., Luxemburg, P. J. C. M. van, and Jonge, G. de (1980). Mother-infant body contact in macaques 11: further steps towards a representation as a continuous time Markov chain. Biol. Behau., 5: 61-94.

Dienske, H. (1984). Early development of motor capacities, sleep characteristics and social interactions in the rhesus monkey, chimpanzee and man. In H. F. R. Prechtl (ed.), Continuity of Neural Functions in Pre- and Postnatal Life. London, Spastics International Medical Publications, Pp.

Haccou, P., Dienske, H., and Meelis, E. (1983). Analysis of time-inhomogeneity in Markov chains applied to mother-infant interactions of rhesus monkeys. Anim. Behav., 31: 927-945.

Hedeen, S. E. (1980). Mother-infant interactions of a captive lowland gorilla. Ohio J . Sci., 80: 137- 139.

Hoff, M. P., Nadler, R. D., and Maple, T. L. (1981). Development of infant independence in a captive group of lowland gorillas. Dev. Psychobiol., 14: 251-265.

Horvat, J. R., Coe, C. L., and Levine, S. (1980). Infant development and maternal behavior in captive chimpanzees. In R. W. Bell and W. P. Smotherman (eds.), Maternal Influences and Early Behauior. Lancaster, MTP Press, Pp. 285-309.

Hwang. C.-P. (1978). Mother-infant interaction; effects of sex of infant on feeding behavior. Early Hum. Dev., 2: 341-349.

Konner, M., and Worthman, C. (1980). Nursing frequency, gonadal function, and birth spacing among !Kung hunter-gatherers. Science, 207: 788-791.

Lawick-Goodall, J. van (1%7). Mother-offspring relationships in free-ranging chimpanzees. In D. Moms (ed.), Primate Ethology. London, Weidenfeld and Nicolson, Pp. 287-346.

Metz. J. A. J . (1974). Stochastic models for the temporal fine structure ofbehaviour sequences. In D. J. McFarland (ed.), Motivational Control Systems Analysis. London, Academic, Pp. 5-86.

Miller, L. C., and Nadler. R. D. (1981). Mother-infant relations and infant development in captive chimpanzees and orang-utans. Int. J. Primatol., 2: 247-261.

Morath, M. (1974). The four-hour feeding rhythm of the baby as a free running endogenously regulated rhythm. Inr. J . Chronobiol., 2: 39-45.

Mufiiz, R. S. (1983). Methodology used for the study of early mother-infant behavior in Bornean urang-utan (Pongo p . pygmaeus Linnaeus, 1760). Misc. Zool., 7: 213-219.

Nicolson, N. A. (1977). A comparison of early behavioral development in wild and captive chimpanzees. In S. Chevalier-Skolnikoff and F. E. Poirer (4s . ) Primate Bio-Social Development. New York, Garland, Pp. 529-560.

Plooij, F. X. (1984). The Behavioral Development of Free-Living Chimpanzee Babies and Infants. Monographs on Infancy, L. P. Lipsitt (ed.), Norwood, New Jersey, Ablex.

Prechtl, H. F. R., and O’Brien, M. J. (1982) Behavioural states of the full-term newborn. The emer- gence of a concept. In P. Stratton (ed.), Psychobiology ofthe Newborn. New York, Wiley. Pp. 53-73.

Rybski. D. A,, Almli, C. R., Gisel, E. G., Powers, J., and Maurer, M. (1984). Sucking behaviors of normal 3-day-old female neonates during a 24hr period. Devel. Psychobiol., 17: 79-86.

Salariya, E. M., Easton, P. M., and Cater, J. I. (1978). Durations of breast-feeding after early initiation and frequent feeding. Lancet, 2: 1141-1143.

Thoman, E. B. (1975). Development of synchrony in mother-infant interaction in feeding and other situations. Fed. Proc., 34: 1587-1592.

Thoman, E. B., Leiderman, P. H., and Olson, J. P. (1972). Neonate-mother interaction during breast- feeding. Dev. Psychol., 6: 110-118.

Verronen, P., Visakorpi, J. K., Lammi. A., Saarikoski, S., and Tamminen, T. (1980). Promotion of breast feeding: effect on neonates of change of feeding routine at a maternity unit. Acto Paediarr. Scand., 69: 279-282.

Wright, P., Fawcett. J., and Crow, R. (1980). The development of differences in the feeding behavior of bottle and breast fed human infants from birth to two months. Behav. Processes, 5: 1-20.

126- 143.

Documents

Regulation of nursing in chimpanzees