The studies Penn et al. critique to discount nonhuman animal relational competencies are heavily weighted toward primates and birds, plus a few additional citations on bees, fish, a sea lion, and dolphins. Cognitive differences among nonhuman species are largely ignored, as if all were cut from the same mental cloth. Here, we focus on several findings on cognitive skills of the large-brained bottlenose dolphin (hereafter “dolphin”) that suggest a capability for reasoning about higher-order relations through the spontaneous combination or concatenation of previously generalized concepts. Dolphin brain architecture is divergent from other large terrestrial mammals, such as the apes (Marino et al. Reference Marino, Connor, Ewan Fordyce, Herman, Hof, Lefebvre, Lusseau, McCowan, Nimchinsky, Pack, Rendell, Reidenberg, Reiss, Uhen, Vander Gurcht and Whitehead2007), giving pause to Penn et al.'s statement that “the substantive difference between human and nonhuman brains will be found in the prefrontal cortices” (sect. 11.1, para. 3). That structure (though not its functions) is lacking in the dolphin brain, but as a whole, the highly encephalized brain allows for levels of cognitive skills convergent with many of those demonstrated among apes (Herman Reference Herman and Herman1980; Reference Herman, Hurley and Nudds2006; Marino Reference Marino2002). Refuting Penn et al.'s claims for functional discontinuity, the following examples illustrate dolphin relational competencies, as exhibited through laboratory studies of four dolphins, Akeakamai (“Ake”), Phoenix, Hiapo, and Elele.

Example 1

Herman et al. (Reference Herman, Richards and Wolz1984) showed that Ake could learn to understand instructions conveyed gesturally within an artificial language system. A sequence of three gestures framed grammatically as locative (L)+object (O)+action (A) required Ake to take the signified action to the signified object that was at the signified location. A second three-gesture frame specified as object1 (O1)+object2 (O2)+relational term (R) required her to construct the signified role-specific relation between the two objects (e.g., by transporting O2 to O1). Without further training, she immediately understood (on the first trial), all of the following four- and five-gesture frames: O1+L+O2+R; L+O1+O2+R; and L _i+O1+L _j +O2+R, where L _i and L _j may be the same or different locations (Herman Reference Herman, Schusterman, Thomas and Wood1986; Herman et al. Reference Herman, Richards and Wolz1984). Thus, she understood spontaneously the concatenation of these strings of symbols into a single instruction, and in so doing inferred the higher-order relation resulting from conjoined first-order relations. Additionally, when we substituted sequences of two deictic gestures for the symbolic gestural references to those objects, by pointing (P) briefly at one distal object and then at a second distal object, followed by a relational term, that is, P1+P2+R, Ake spontaneously incorporated the inverse grammar used with symbolic gestures, O1+O2+R, into her interpretation of the deictic gestural sequence, by taking the object pointed to second to the object pointed to first (Herman et al. Reference Herman, Abichandani, Elhajj, Herman, Sanchez and Pack1999). The abstract grammatical rule was transferred not merely to other object exemplars, or to extensions of the existing grammatical rules within a closed symbolic communication system, but to a completely different indexical communication system, in a manner refuting “particular specificity.”

Furthermore, Ake, when presented with long anomalous strings of symbols that violated the familiar syntactic structure of the artificial language, spontaneously extracted subsets of items that comprised a legitimate grammatical rule, even combining nonadjacent items when necessary, and she carried out the instruction contained within that embedded subset (Herman et al. Reference Herman, Kuczaj and Holder1993a; Holder et al. Reference Holder, Herman, Kuczaj, Roitblat, Herman and Nachtigall1993). Additionally, when given proper syntactic strings that violated a semantic rule, such as a request to transport an immovable object, she rejected the string, either offering no response at all or occasionally carrying out a substitution response by transporting a movable object. These responses to anomalous sequences show a profound understanding of implicit relations within the grammar of the imposed language, and an ability for reinterpretation as necessary of the relations between symbols and the real world.

Example 2

Ake could report accurately whether a symbolically referenced object was present or not in her tank, by pressing a Yes or No paddle, respectively (Herman & Forestell Reference Herman and Forestell1985). Later, on the first occasion that she was given an O1+O2+R instruction in that context, with O1, the destination object, absent, she spontaneously (and thereafter) carried O2 to the No paddle (Herman et al. Reference Herman, Pack, Morrel-Samuels, Roitblat, Herman and Nachtigall1993b). This, in effect, was a reinterpretation of the perceptual world and a conditional reassignment of the function of the No paddle as a destination object.

Example 3

Our work with all four dolphins shows that dolphins are generalized mimics, capable of faithfully imitating on command arbitrary artificial sounds (Richards et al. Reference Richards, Wolz and Herman1984), as well as copying the behaviors of others (humans, other dolphins, or themselves), either viewed live or on a television screen (Herman Reference Herman, Nehaniv and Dautenhahn2002b; Mercado et al. Reference Mercado, Murray, Uyeyama, Pack and Herman1998; Reference Mercado, Uyeyama, Pack and Herman1999). Behavioral imitation is a form of “sameness” comprehension in that it requires relating one's body image (see Herman et al. Reference Herman, Matus, Herman, Ivancic and Pack2001) to the perceived image of another, and may require inferring analogies where imitation crosses species boundaries, such as the dolphin representing the human leg by the dolphin's tail.

Example 4

As described in Herman (Reference Herman, Nehaniv and Dautenhahn2002b) the four dolphins were taught a gestural sign, tandem. When each of a pair of dolphins was given the tandem sign followed by a sign for a particular behavior, such as back dive, they joined together and carried out that behavior in exquisitely close synchrony. Each dolphin was also taught a sign create, which required it to perform any behavior of its own choice. Then, when a pair was given the two-item sequence tandem+create, they joined together and in close synchrony performed the same self-selected behavior. On a later formal test of Elele and Hiapo's responses to tandem+create, 79 different highly synchronized behaviors were recorded with 23 of them novel (i.e., they were not under control of established gestures). The tandem responses were very closely timed, and although careful video analysis could detect some slight asynchrony in timing in some cases, there was no consistent “leadership” by one dolphin or the other. These results reveal close collaboration, as well as the marrying of two abstract concepts, tandem, a social collaboration, and create, a self-determined behavioral innovation, into a higher-order abstract relationship. This collaborative capability likely finds expression in the wild, for example, in the fluid first-and second-order alliances formed in collaborative efforts by male dolphins to secure female consorts (Connor et al. Reference Connor, Wells, Mann, Read, Mann, Connor, Tyack and Whitehead2000); interestingly, Penn et al. deny that nonhumans can participate in collaborative activities.

Penn et al. make a top-down claim for genetic pre-specification in humans alone of a module for higher-order cognition. However, bottom-up theories may offer better paths to understanding nonhuman animal cognitive potential – for example, the non-generativist image schemas (Clausner & Croft Reference Clausner and Croft1999; Lakoff & Turner Reference Lakoff and Turner1989) that are claimed to create conceptual representations through cumulative sensorimotor experiences. The relational capabilities of the dolphins illustrated (as well as others described in Herman Reference Herman, Hurley and Nudds2006) we believe derive in part from the cumulative cognitively challenging experiences provided within their long-term immersion in a varied and stimulus-rich educational environment (in contrast to many of the animal studies cited by Penn et al.). These experiences allow for the emergence of intellectual competencies that might not be realized otherwise (Herman Reference Herman, Bekoff, Allen and Burghardt2002a).