Behavioral data validation

Over 963 hours of data were recorded on the 10 adult male subjects over the course of the year of this study. To assess potential variation in behavioral data from the different male subjects, frequency data were examined according to time of day and in relation to activity (across time of day) according to the 10 males (Table 3.4 and Table 3.5). Male variability in terms of total activity budget was also examined for biases in the data (Table 3.6).

Additionally, individual male data were adjusted randomly to control for the amount of data taken on each male to find whether biases exist in data collection from each male (see below).

There was no significant difference between the 10 males and the distribution of data collected over the course of the day (ANOVA, F-ratio = 0.96, df = 9, 30, p-value = 0.49). Males were sampled equally across time of day (Table 3.4). However, significance exists in

terms of total data collected over the course of the day (ANOVA, F-ratio = 20.2, df = 3, 36, p-value < 0.0001). Data were taken mostly from 0900 hours to 2000 hours, with the morning hours having less data (16%) recorded (Table 3.4). This may have implication for the results in terms of activity budget. Pruetz and Bertolani (in press) found that activity was influenced by the time of day; the Fongoli chimpanzee fed more in the morning, while resting more often in the middle of the day. Thus, feeding in this study may be underrepresented due to lack of data in the morning hours. To find if this discrepancy exists, analyses on activity across time of day were conducted.

Table 3.4: Proportion of frequency of observations on 10 adult male focal subjects according to time of day from August 2006 to July 2007 at Fongoli (instantaneous sampling).

Time 0600-0859 0900-1159 1200-1459 1500-2000 Bandit 20.5 30.9 24.6 24.0 100 Bilbo 16.4 34.1 29.6 19.9 100 Diouf 15.3 35.5 28.8 20.4 100 Foudouko 17.7 31.0 30.7 20.6 100 Karamoko 16.3 25.6 23.3 34.9 100 KL 16.7 30.6 28.7 24.1 100 Lupin 16.1 32.3 28.6 23.0 100 Mamadou 15.0 31.3 29.7 24.0 100 Siberut 16.6 34.0 25.2 24.2 100 Yopogon 11.7 33.6 32.6 22.1 100 Total Data 16.1 32.0 28.4 23.4

Table 3.5: Proportion of data organized by activity across time of day using pooled data from 10 males (instantaneous focal sampling).

0600-0859 0900-1159 1200-1459 1500-2000 Other Feed/forage 44.2 18.6 9.4 33.1 Termite Fish 5.5 4.5 4.6 15.9 Rest 25.9 48.6 64.3 33.1 Travel 13.4 5.5 2.6 9.2 Social 7.6 18.7 16.4 7.5 Other behaviors 3.3 4.2 2.6 1.2 100 100 100 100

Data on activity budget were pooled from all males (Leger & Didrichsons 1994, Maddala 1971). There was no significant variation in activity across time of day (ANOVA, F-ratio = 0.34, df = 3, 20, p-value 0.80) (Table 3.5). When examining the activities according to time of day using matched pairs Student’s t-test, no significant differences exist between any of the three-hour time slots throughout the day (df = 5): 0600-0859 hours and 0900-1159 hours t-ratio = 1.37, p-value = 0.22; 0600-0859 hours and 1200-1459 hours t-ratio = 0.78, p- value = 0.47; 0600-0859 hours and 1500-2000 hours t-ratio = 1.87, p-value = 0.12; 0900- 1159 hours and 1200-1459 hours t-ratio = -0.76, p-value = 0.47; 0900-1159 hours and 1500- 2000 hours t-ratio = -0.90, p-value = 0.41; 1200-1459 hours and 1500-2000 hours t-ratio = - 0.37, p-value = 0.73. For example, feeding may occur proportionally more in the morning (44%) and evening (33%) hours, however significant variation between these times and the middle of the day does not exist.

Pooling the data from all 10 males may be a possible weakness of this study. Thus, individual behavioral data across males were examined for variation. Individual’s activity budgets were similar between males (ANOVA, F-ratio = 0.14, df = 9, 50, p-value = 0.99) (Table 3.6). However, in examining individual activity categories, Siberut was an outlier in that he rested more than all other males (Table 3.6). If Siberut was taken out of the data set, little difference is displayed in the proportion of activity budget with and without including Siberut (Table. 3.7) and no significant difference exists between frequencies (Student’s t-test on frequency, t-ratio = 1.84, df = 5, p-value =0.12). Thus, the pooling of data for male subjects does not have an influence on data presented in this dissertation.

To control for the possibility that certain males had a preference for termite fishing (or other activity) and these individuals were consequently observed the most, a target of 10

hours of data collection per male each month was set. In examining the data on individuals (Table A7), the male with the least amount of collected data was Karamoko, while the most data was collected on K.L. (Figure 3.6). To determine if the amount of data collected affected the results, I randomly adjusted all frequencies for each male to equal the total data collected on Karamoko. Using the adjusted data (Table 3.8), the activity budget did not differ from the original data (ANOVA, F-ratio = 0.14, df = 1, 14, p-value = 0.71).

Table 3.6: Proportion of individual male activity budget at Fongoli from August 2006 through July 2007 (instantaneous sampling).

Feed/forage Termite fish Rest Travel Social Other Total

Bilbo 21.3 10.7 47.4 6.6 12.4 1.7 100.0 Bandit 25.2 6.4 45.7 6.6 14.2 1.9 100.0 Diouf 23.5 7.6 42.0 8.5 15.0 3.3 100.0 Foudouko 21.6 5.2 44.1 8.8 19.0 1.4 100.0 Karamoko 25.3 4.7 42.3 7.4 16.1 4.2 100.0 KL 24.7 9.8 43.3 5.4 12.4 4.3 100.0 Lupin 27.8 11.2 38.5 6.3 13.1 3.1 100.0 Mamadou 26.4 6.4 46.2 6.9 11.3 2.7 100.0 Siberut 16.6 6.8 60.6 6.1 7.5 2.4 100.0 Yopogon 22.0 4.8 44.6 6.4 18.4 3.9 100.0 Total average of male average 23.4 7.4 45.5 6.9 13.9 2.9

Table 3.7: Pooled proportion of daytime activity budget from 9 males and from 10 males.

Feed/forage Termite fish Rest Travel Social Other

Data from 9 males* 24 8 44 7 14 3

Data from 10 males 24 7 45 7 14 3

* Siberut extracted from the pooled data

Table 3.8: Original activity budget data and data adjusted for bias assessed from instantaneous focal male pooled data (N=10).

Original activity budget (%) Adjusted activity budget (%) Difference

Feed/forage other 23.57 23.94 -0.37 Termite fish 7.50 8.17 -0.67 Rest 45.46 48.59 -3.13 Travel 6.83 5.01 1.82 Other Social 13.39 11.37 2.02 Other 2.77 2.47 0.30 Aggression 0.41 0.40 0.01 Drink 0.08 0.07 0.01

Figure 3.6: Proportion of instantaneous focal data taken from each of the ten adult males at Fongoli from August 2006 to July 2007.

Table 3.9: Original data and data adjusted for bias on monthly proportion of termite fishing (TF) assessed from instantaneous focal male data (N=10).

Termite fishing proportion (%) Adjusted TF proportion (%) Difference

August 0.12 0.14 -0.02 September 0.00 0 0.00 October 0.00 0 0.00 November 0.47 0.27 0.20 December 0.00 0 0.00 January 1.06 0.96 0.10 February 3.17 1.23 1.94 March 5.87 5.6 0.27 April 24.06 22.13 1.93 May 16.78 17.49 -0.71 June 25.82 28.83 -3.01 July 22.65 23.36 -0.71

Finally, sampling error may be suggested, since termite fishing was one of the main objectives, and it could be argued that I was biased in my data collection. However many thorough steps were taken to avoid such biases. The termite fishing data obtained from 10

different males did not differ from the adjusted and original data (one-way ANOVA, f-ratio = 0.87, df = 1, 18, p-value = 0.36). To examine specific activity, the adjusted data on monthly termite fishing (Table 3.9) did not differ from the original (one-way ANOVA, f-ratio = 0.08, df = 1, 22, p-value = 0.78), and it positively correlated with the original data (Spearman’s rank correlation ρ = 0.88, p-value <0.001). Therefore, the differences in amount of data taken from the ten males do not significantly affect the results.

The only data that stands out in terms of potential bias in the results of this dissertation concerns those data collected across time of day. Data collected in the early morning were significantly lower than the data collected over the course of the day from 0900-2000 hours. Thus, the data might be skewed in terms of certain activities. Pruetz and Bertolani (in press) found that time of day had a significant influence on activity. Travel and feeding were conducted more often in the morning hours, while resting and grooming were conducted more often during midday (Pruetz & Bertolani, in press). Although, this research did not find such significant differences in daytime activity patterns, but the proportional trend is similar to Pruetz and Bertolani (in press). The differences between the two studies might be due to the year data were collected and the amount and method of data collected. Pruetz and Bertolani collected both focal and scan sampling data from March 2005 through March 2006, obtaining 2500 hours, more than twice as much as the data collected here. Results of this research are based only on focal animal sampling from 963 hours. Resting more during midday, when data were most often recorded in this study, would possibly skew the total proportion of resting among males, thus some behaviors might be underestimated. No significance was found between behaviors across time of day. Consequently, pooling of the data across time of day would not have an influence on the results of this study.

Additionally, individual behaviors between males did not significantly differ and pooling of data from all 10 males did not affect the overall activity budget and data found in this study.