Well, I warned readers in Part I that the number work was the equivalent of a book chapter and even though I’m reporting it in several entries, I’m still summarizing the various experiments to get everything presented in a reasonable space. My point is that number studies are complex, and anyone who says anything different simply does not understand numerical concepts.
I left off last time with Alex having demonstrated some simple form of a zero-like concept. At that point, we really didn’t know exactly what he understood, and were in the process of designing some new experiments—but we were also beginning number work with Griffin.
Given that we had done visual, simultaneous numbers with Alex, we wanted to see if it would be easier or more difficult for Griffin to learn auditory, sequential numbers—that is, go back to some of the original number work on grey parrots that was done by Koehler (1943) decades earlier, where he compared the two types of learning.
Alex Counts Finger Snaps
We had worked out a computer system that played different lengths of click sounds, with different inter-click intervals so that only the number of clicks would be relevant, not the amount of sound. One day, however, we were just messing around with Griffin and were using finger snaps, asking him “How many?” We snapped twice, asked him the question—and he refused to answer. We tried again; again he refused to answer. But Alex, right next to us, piped up “Four!” I told him to be quiet, that I had clicked “two.” I tried with Griffin again, and once more Griffin refused to respond. But now Alex said “sih” (his word for six. I realized he was adding! Given that testing auditory addition would be extremely difficult for a number of reasons (I won’t bore you with the details!), we decided to replicate the addition studies of my colleague, Sally Boysen (Boysen & Berntson, 1989).
In that study, Sally’s chimpanzee, Sheba, walked around the lab, uncovering various boxes with different numbers of fruits (apples, oranges, etc.) in each, then went to a number board and tapped on the Arabic number that represented the sum in all the boxes. We couldn’t let Alex do quite the same thing, so we hid different numbers of treats under two different cups, showing him what was under each cup separately, and then, with both quantities covered, asking him “How many total?” (see figure).
Alex Adds Up the Cups
To ensure that he couldn’t use mass or contour as a cue, the task involved summing sets that were made of differently sized and/or irregularly shaped items. We also realized that we could use null-set trials (i.e., leaving both cups empty) to further test. his understanding of zero (Pepperberg, 2006). During questioning, only cups were visible. To respond correctly, Alex had to remember quantities under each cup, perform some combinatorial process, then vocally label the total.
For sets not involving two empty cups, Alex scored ~85% on the first trials; identical-sized items did not improve accuracy (Pepperberg, 2006). Interestingly, he consistently erroneously labeled 5+0 sets as “sih” (six) when visibility was 2-3 seconds, but accurately labeled 4+1 and 2+3 (Pepperberg, 2006). Given the chance, he corrected the few mistakes he made on all sets except 5+0. Only if given 6-10 seconds to view the addends did he correctly (100%) label 5+0 as “five.”
Additional time did not alter accuracy for other sets. Such data suggested he was actually counting for 5: Only above 4 did he, like humans who can accurately estimate sets only up to about 4 when under time pressure, need additional time so as to count/label the set exactly (Pepperberg, 2006). Overall, his data are comparable to children’s (Mix et al., 2002) and, because he added to six, are beyond those published on apes (Boysen & Hallberg, 2000).
His responses for the sets of totally empty cups were also very interesting. Remember, we had never trained him on zero (Pepperberg & Gordon, 2005); we just wanted to see what else about the zero-like concept he had figured out on his own. On the first four trials, Alex simply looked at the tray and said nothing. He would sometimes try to lift the cups himself, and I would then show him again, by lifting the cups one at a time, that there was nothing present. On the fifth, sixth, and seventh trials, he said “one.” On the last trial, he again refused to answer. His answers might seem odd, but think about them for a moment: Numerical competence is based on the assumption that something exists to enumerate, whatever the process involved. My asking Alex to enumerate something that did not exist clearly presented a challenge—it was different from asking him to comment on the attribute of a specific, if absent, set—which is what he had done previously.
Alex’s two different responses to absence were both intriguing. His failure to respond on five trials suggests he recognized something was different from the other trials; that is, even if he did not understand what was expected, he knew his standard number answers would not be correct. He did not, as he has done when bored with a task (e.g., Pepperberg, 1992; Pepperberg & Gordon, 2005), give strings of wrong answers or request treats, or to return to his cage. He acted more like an autistic child (D. Sherman, personal communication, January 17, 2005) who simply stare at the questioner when asked “How many X?” and there is nothing to count. His response of “one” in the fifth, sixth, and seventh trials suggests a comparison to that of Matsuzawa’s number-trained chimpanzee Ai, who confused “one” with “zero” (Biro & Matsuzawa, 2001). Although Alex was never trained on ordinality and had learned numbers in a random order (see Pepperberg, 1999), he, like Ai, seemed to grasp that “none” and “one” represented the lower end of the number spectrum. It thus seemed that he had some, but not a full, understanding of a zero-like concept. Thus, for Alex—who, again, had never been trained on zero—the attribute of a set—its quantity—could be missing from a collection and be labeled “none,” but the missing object itself could not be denoted as “none.” The distinction is subtle but can explain his responses.
Alex’s numerical summation abilities didn’t end there. Subsequently, we found he could extend his understanding to quantities hidden under three cups, and perform equally well if Arabic numerals rather than sets of objects were hidden (Pepperberg, 2012).
Overall, these data provided several new insights into Alex’s abilities. First, as noted at the beginning of this article, I did not initiate the study on my own: Alex himself triggered our interest in addition. Second, the task was more complex than simply labeling a visible set: Alex had to remember the addends and combine them to form a representation of the total, which he then had to identify. Third, he demonstrated that his self-generated concept of zero was not exactly the same as that of adult humans, but did match that of young children and possibly apes. Finally, his behavior provided the first evidence of a counting-like strategy. But still more is yet to come!