Louis Herman and Ronald Schusterman used large-scale arm movments, signs (hand-shaped) and computer-generated sounds with bottlenose dolphins, Puka, Kea, Ake, Phoenix (Tursiops truncates). The animals had to associate specific (artificial) whistles with objects and/or actions. Dolphins were asked also to use (press on) dedicated paddle [40,41]. Scientists used command gestures to teach Californian sea lion (Zalophus californianus), Bertie and Rocky, names and characteristics of objects [42-44].

In the 70s Irene Pepperberg started her work with Alex, an African grey parrot [45]. She used the Model/Rival technique in order to English words to the bird. In this method, two persons interact in front of the bird and take successively the role of the trainer—asking for the name or characteristic of an object and the student—that represents the model while answering correctly. In this situation, the student represents also a rival for the bird because he gets the trainer’ attention and receive the reward (object or food) if the label was correct-corresponding to the question. Humans exchange their role and sometimes ask directly to the bird for the correct answer [14]. The Human student also produces sometimes incorrect answer so that the bird can see that not all the sounds have the same meanings and that the trainer is expecting for a specific (referential) label. Several birds were trained by Irene Pepperberg and currently Griffin is still learning more words and concepts [46]. Other laboratory decided to work on language-trained parrot, in France [47,48], Czech Republic [48] or Canada [49]. Different methods have tested in which several parameters were changed such as the number or the nature of the trainers [50], the social interaction [51,52], the referentiallity [48], the joint attention [53], the use of video [54, 55] or sound [48] records, etc. The results vary according the birds (sex, age, personality, cognitive ability, vocal competencies, motivation) and their social environment [46,48]. Most of the bird learned rapidly new words but some of them had difficulties to associate to an object or a characteristic.

More recently a Portuguese team starts to train a dog to use a keyboard in order to ask for food, a walk or to be petted [56]. A female mongrel dog named Sofia, was submitted to a training program in which she acquired the discrimination of lexigrams associated to specific desires and the use of a keyboard to produce requests. The authors make sure about the motivational state of the animal (intention) and the communicative nature of the interactions (perspective taking; [57]). The authors report that the dog learn quickly to use the device when experimenter could see the information. They observed some modifications in the behavioral display before and after keyboard pressing suggesting a certain arousal for specific needs [56].

In general, all the individuals (whatever species they belong to) have (had) a good understanding of human language (English, French, Portuguese, etc.) [48,56,58].

Apes signed and used plastic magnet and lexigram to ask for toys, food, etc., but also for walk and interactions. In few occasions they described their environment [59,60]. Alex learned and use English labels to answer the questions but also to ask for something. Sometimes he even asks questions about objects characteristics or name [46]. Dolphins learned to mimics specific sounds, used to designed specific objects. [41]. The Sea lion was able to associate the different gestures with specific meaning such as object’ name, actions or even object characteristic [43]. Loulis (chimpanzee) learned to sign for several different categories such as people name, objects, actions, location. He also displays “No” or “Want” signs [61]. Apes and parrots learn different categories such as “color of”, “name of”, “size of”, etc. [37].

Sofia’s performances at the keyboard was based on the association of a specific sign with a specific desire and that it was part of goal directed sequences similar to those normally displayed by dogs. The dog does not use it when alone, presses in view of the experimenter [57] and alternates gazing between the experimenter and the object.

The apes species, dolphins and grey parrot show evidence of both discrete combinatory and category-based rules. The dolphins and Kanzi show knowledge of argument structure (although for Kanzi, only in his comprehension of English). The capacities that underlie these three properties of syntax, then, do not appear uniquely human [62]. Referentiality and syntax have been also demonstrated with other apes. Sherman and Austin (two chimpanzees using lexigrams) show some understanding of the different category “food” vs “tool”, and classify the objects, their pictures or even their corresponding lexigrams. [37] Lana learned to chain lexigrams into strings that were similar to human sentences. Heidi Lyn showed that the vocabulary errors produced during tenyear data collecting period for Kanzi and Pabanisha revealed that apes are able spontaneously to create a complex, hierarchical, web of representations when exposed to a symbol system (lexigrams) [63]. The Californian sea lion show some understanding of semantic [42]. Sofia has been able to extract the information contained in complex messages and to integrate it in directed performance. In a recent study Sofia responds to verbal requests composed of two independent terms, one referring to an object and the other to an action to be performed relative to the object [64]. Nevertheless, when looking at word comprehension in other dogs, we can see that Rico (border collie) discriminate dozens of words and showed a “fast-mapping’’ performance similar to children’s, that is, the capacity of attributing, by exclusion, a new word to an object never seen before [22] and Chaser, another border collie [65], comprehends object names as verbal referents (more than a thousand words). Of course dogs never learned to say any of the words they learned (understand). In this respect, their limited vocal production and extensive comprehension are similar to those of humantrained sea lions [42], and dolphins [66]. Apes are also limited regarding their vocal ability but they have a quite good understanding of human words and are able to map new words onto new object (fast mapping) in ostensive context like dogs [67]. Alex also displayed fast-mapping [68].

Most of the individuals (apes and parrots) have been observed while babbling. Some individuals also recombine phonemes or sign or even create new ones (e.g. Washoe created a sign for “bib”) [69,70]. These are some example of combination: “ring-ball” combination of bottlenose dolphins [71] while playing with both toys at the same time, Washoe sign “water bird” for swan or “cry hurt food” for radish; Lana press the lexigrams “Coke which is orange” for Fanta [72], Koko sign “White tiger” for zebra; Alex: “green nut” for pumpkin seeds, “banerry” for apple [46].

Animals have been also observed “talking” to themselves [38,73]. This occurred during the learning and consolidation phase but not only. Indeed, parrots enjoy these moments (resting period). They can also “talk” to describe their environment or express what they have to do (e.g. Washoe signed “Quiet” when she sneaked into a forbidden room [38].

Dolphins can label things and report presence or absent of such object by pressing on specific paddles [17].

Alex was trained on absence of a difference or similarity “what same?/what different?”, as well as apes. Thus individuals were able to display a zero-like concept—which is the expression of an absence of something (e.g. no difference or none green block) [46]. Not only Sarah mastered the concept of same and different (e.g. apple different banana) but also was able to judge the relationship between relationships, a higher order of conceptualization (e.g. pairs of apple similar pairs of bananas but different an apple and a banana). Sarah learned the concept of “name of”, “color of”, she also mastered conditional “if-then” [39]. Lana learned to use “more” and “less”. Alex also was able to compare quantity based on Arabic numerals or sets of objects or to compare the size of two objects. Sea lions were able to mastered comparative and absolute size comprehension in the Sea lion [43] (See the categorization section above for birds examples).

Kanzi, Panbanisha and Panpanzee used precursor of morality while pressing the lexigrams “good” and “bad” in appropriate contexts. This suggests a symbolic language processing in the different species but heavily influenced by the value judgments of their human caregivers [74].

The ability to categorize elements of the environment, i.e. to classify objects according to proprieties they share, is a fundamental aspect of information processing. Conceptual categorization implies two criteria: a rapid generalization over class members of items and a classification of items not necessarily similar perceptually (e.g. “food” vs “tool” items [75,76]). Alex, was able to categorize items according to their color, shape or matter. He was capable of giving the similar and/or different characteristics of the items presented [77,78]. He was also able to identify the number of items according to two modalities [79,80]. Alex’s categorization was noteworthy because he expressed this classification by verbalizing labels. However, Alex received a long training (months to years for various items and/or concepts) before being able to label and categorize correctly the items presented. In a recent study two African grey parrots learning French labels spontaneously categorized items with vocal imitations [48]. At the time of the experiment, the birds knew a few referential labels belonging to two different classes (food and objects) and imitated several labels that did not refer to particular items. Although we were trying to teach them new words, we observed that both subjects tried to say some other labels that they knew in order to obtain the items presented. The unexpected result was that, although never rewarded for this behavior since they did not use the labels corresponding to the items shown to them, both individual spontaneously emitted more “food labels” in front of food items and the male produced more “object labels” in front of object items [47]. Note that labels of a given category (i.e. either food or object labels) do not share common acoustic or visual features. Thus, the parrots spontaneously attributed a class to their vocalizations according to the edibility of the items presented (food or object items). This ability is similar to that described in macaques able to distinguish food vs non-food items [81]. Such an untrained categorization is rarely described in animals, but it could be related to a mechanism commonly observed in human children learning to talk. During the acquisition of the meaning of words, children generally adopt two different strategies: over-extension and under-extension [82]. The latter consists of using different labels for the same item. The over-extension strategy consists of attributing one label to different items. In children, over-extension usually is a process associated with perceptual categorization. Overextension gradually disappears when children add features to a word such as its functional role [83]. In the case of parrots, it seems that they used something like the over-extension strategy, by using a few food labels to ask for all food items and a few object labels for all object items. This type of error in the mechanism of information treatment was also observed in language-trained apes [38,63]. Pepperberg, without having formerly tested categorization with her parrots, reports that when Alex erred on, for example, a color label for objects, he most often provided another color label [46,77].

Furthermore, Alex spontaneously established equivalency relations between Arabic numerals and the corresponding sets of items [84]. Alex learned to quantify sets of objects with vocal labels [79], choose the set corres ponding to the vocal number label [80], utter the corres ponding number label in response to the presentation of the number symbol and to select the number corresponding to the vocal number label [85]. Griffin is currently tested on visual perception. He already knows some colors, shapes and numbers. Irene Pepperberg is currently addressing the following question: Are you still seeing a square even when one of its corners is overlapped by a masking object? Yes, it seems that the bird perceive the object as complete structure. Maybe grey parrots are able to represent mentally the “normal” image of a square. Alex had been tested on the Müller-Lyer illusion [86] and the results suggest that like humans, birds are sensitive to optical illusion.

A bottlenose dolphin discriminates visual stimuli differing in numerosity [87]. Apes have been tested on their numerical competencies and their ability to compare quantities using items or symbolic representations (object or lexigrams). Symbolic representations help chimpanzees to succeed in a reversed reinforcement contingency task. They maximize reward choosing the smaller Arabic number (and thus receiving the larger quantity of reward) but failed to do so when facing the food item [88].

Kanzi learned through observation (of the adult) how to use lexigram. Loulis (Washoe was his foster mother) was not cross-fostered and learn through interaction with the other chimpanzee to sign. Loulis mastered sign language through watching, babbling and then doing the sign properly [89]. The researchers also observed maternal handguidance as Washoe molded Loulis’s hand into the sign for “food”, or place “drink” on his lips. Several times Washoe sign the name of object in front of Loulis and before using them such as “brush”, “hat”, etc. Washoe also learn new sign from her conspecific such as “blanket” for instance [38].

The same for birds who learn from each other new vocalization and thus new words and then associate the label with the correct object (or characteristic). Irene Pepperberg reports also that Alex vocalized several times “say better” or even giving the (right) answer when another birds was questioned [46].

Using lexigrams, words or plastic object is also a good way to assess the memory of an individual across the time. Thus Lana, Koko, Kanzi et al. remember the name of object they did not see for several months [37,90].

Irene Pepperberg, taught to employ the vocalization “want” before an object label so as to discriminate functional “labeling” from “requesting” [91]. Thus When Alex was asking “What toy?” even if he was not motivated to play with it he could have ask for a nut after having gave the right answer. Indeed, Shango (African grey parrot) who was at the beginning of French word acquisition, stopped rapidly to say “semoule” (“cornmeal”) because he did not like the taste of the corresponding food [48]. Alex was requesting for a reward (want nut) but also to stop the session (“wanna go back/ wanna go chair”), or vocalized when he disagree saying “No” [46]. Cabanac, who trained Aristote (African grey parrot) with the Model/Rival technique, observed that his bird was able to combine labels that express preference and pleasure. Indeed, the bird was taught with different labels and when the experimenter was interacting with the birds he pronounced specifics words such as “bon X” (“good X”) when given the object X or doing some actions (such as preening) the birds just ask for. Latter on the author observed that while he was giving a grape to Aristote, the bird spontaneously pronounce “raisin bon” (“good grappe”), a label combination he never heard before [49].

In the same way, apes using lexigram or sign language requested for food, toy, tickles, etc using the appropriate signal and also request for “more” or “less” of it. Washoe sign “More” to get more tickling after that the Gardners introduced “More” into a game of pulling [38].

Four African grey parrots have been tested on their ability to obtain an item suspended from a string such that multiple, repeated, coordinated beak-foot actions were required for success. Only the birds with little training in referential English requests (e.g. “want X”) succeeded. The two others (Alex & Griffin) failed to obtain the reward as they did not try and engaged in repeated requesting “want nut” [92]. Even if their intentions were not to manipulate humans’ behavior at least the birds understand the usefulness of the vocalization and its use within a heterospecific communicative system.

The African grey parrots tested in the French laboratory developed specific asking calls for food [93] but they never used it during the test sessions and on the contrary labeled differently the items according to their category (food vs toy) [48].

Apes signed, pressed lexigram or combined object in order to request for food or toy or whatever. In captivity, Chimpanzees have been observed to beg extending an open hand and vocalizing. Loulis for instance produced the sign “Gimme” to request for a preferred food [89].