Dispatches Comparative Cognition: Action Imitation Using Episodic Memory Jonathon D. Crystal Department of Psychological & Brain Sciences, Indiana University, 1101 E. 10th Street, Bloomington, IN 47405-7007, USA Correspondence: [email protected]
Humans encounter a myriad of actions or events and later recall some of these events using episodic memory. New research suggests that dogs can imitate recently encountered actions using episodic memory. Imagine that I arrange for an unusual action to occur outside my building before a visitor arrives for a meeting at my office. If a juggler stands in front of the building, when my guest arrives for a visit, I might ask her to imitate the actions she observed earlier at the entrance to the building. If my guest noticed the juggler’s unusual actions, she might mimic the juggler’s arm movements to show her knowledge. What type of memory subserves successful imitation? Students of human memory refer to this type of memory as episodic memory, a memory system that stores personally experienced events. When I ask for imitation, presumably my visitor retrieves a memory of the earlier episode and proceeds to mimic the remembered actions. In a report in this issue of Current Biology, Fugazza et al.  have now shown that dogs can imitate distinctive actions recently performed by a human trainer, and they do so using episodic memory. Dogs Can Be Trained to Imitate Actions The dogs in the Fugazza et al.  study were trained to imitate an action upon a ‘‘do it’’ command. In this approach, the trainer rewards the dog for repeating a demonstrated action. Although this training occurs at the dog’s home with a variety of trained actions, the experimenter used novel actions (each demonstrated only once) in the experiment — the specific actions/ imitation in the experiment had never been previously rewarded. Further, the autthors used an elegant, so-called two-action method . In this method, two versions of a target action are identified; for example, touch an umbrella with a hand/paw (Figure 1) versus touch the umbrella with one’s nose. One action is demonstrated to each animal, but some animals received
one version whereas other animals received the other version. The imitation is considered successful only when the target action (but not the similar action) occurs. Previous work showed not only that dogs can imitate upon the ‘‘do it’’ command, but that they remember the target action over long delays [3,4]. This earlier work established that dogs remember previously seen actions, but that was not necessarily explained by episodic memory. After each dog was trained to imitate at home, it was brought to the laboratory for a more carefully arranged sequence of action demonstrations mixed with other distracting activities. To arrange for a distracting activity, the dog was also trained to lie down upon arrival on a blue carpet. Importantly, during training to lie down, the dog was never asked to imitate when on the carpet. The training to lie down on the carpet continued until the dog spontaneously lay down without a prompt from the trainer. Next, the experimenter provided the dog with an opportunity to observe an action (Figure 1A) followed by a delay, but instead of requesting imitation with a ‘‘do it’’ command, the dog was led to the carpet. At this point, the dog lies down spontaneously, suggesting that it expected a lie-down request rather than a ‘‘do it’’ command. When the trainer subsequently said ‘‘do it’’, the dog performed the corresponding action imitation (Figure 1B). Why Is Action Imitation Such a Powerful Approach? Efforts to model episodic memory in animals have been beset by controversies. Some of the controversies stem from the fact that students of human memory have been concerned with subjective experiences that are thought to accompany episodic memory
R1226 Current Biology 26, R1226–R1245, December 5, 2016 ª 2016 Elsevier Ltd.
retrieval in people . Finding behavioral expression of putative subjective experiences is problematic in animals. Other controversies stem from differing views about suitable criteria to establish episodic memory in animals [6–10]. Tom Zentall and colleagues [11–13] identified a key problem in experiments designed to document episodic memory in animals. He noted that, when animals are trained to learn some rules (or contingencies), they likely develop expectations about the upcoming memory test. Expectations about upcoming tests present a problem for animal models of episodic memory. The central hypothesis of an animal model of episodic memory is that, at the moment of a memory assessment, the animal remembers back in time and retrieves a memory of the earlier event or episode [9,10]. An important alternative explanation exists whenever an animal can anticipate the upcoming test. According to this view, the animal carries forward the information needed to answer the upcoming test. Critically, an animal that carries forward the information needed to answer the question successfully answers the question, but it does not need to retrieve a memory of the episode at the moment of the memory assessment. Clearly, this alternative is a threat to the central hypothesis that the animal uses episodic memory. An Unexpected Question Fugazza et al.  provide two compelling lines of evidence that, in their study, the ‘‘do it’’ test of memory was unexpected. First, they relied on a well-established observation from developmental science, namely that young infants fixate at unexpected or impossible events longer than at expected or possible events . This approach has been used
the ability to answer the unexpected question was selectively eliminated whereas the ability to answer an expected question was not impaired . Because episodic memory is hippocampal dependent , these data support the conclusion that the rats used episodic memory to answer the unexpected question. Perhaps the elusive confirmation of incidental encoding in dogs will come from the combination of action imitation and neuroimaging .
Figure 1. Action imitation in dogs is based on episodic memory. A person demonstrates a distinctive action (A) and after a delay, when the owner says ‘‘do it’’, a dog imitates the corresponding action (B). (Photos courtesy of Claudia Fugazza.)
successfully with non-humans, including dogs . So the authors compared the duration of looking at the owner after unexpected or expected ‘‘do it’’ commands. In their new experiment, the authors argued that the ‘‘do it’’ command was unexpected, because the dogs spontaneously laid down and did not receive ‘‘do it’’ commands in the previous stage of training. By contrast, in earlier experiments [3,4], the dogs were always trained with a demonstration reliably followed by a ‘‘do it’’ command; in this situation, when a demonstrated action occurs, the dog likely expected a ‘‘do it’’ command to occur later. Confirming these viewpoints, the dogs looked to their owners when the ‘‘do it’’ command was given unexpectedly, and the average duration of these looks was longer than in earlier experiments with expected memory assessments [3,4]. Notably, the authors compared the duration of looking at the owner after equivalent retention-interval delays (one minute and one hour) and after equivalent distraction from lying down (with a short, one minute delay). A second line of evidence that the memory assessment was unexpected comes from a comparison of accuracy after expected and unexpected memory assessments. Fugazza et al.  documented a dissociation of memory accuracy. The likelihood of successful imitation declined rapidly when the ‘‘do it’’ command was unexpected (declining after one minute and one hour delays, relative to expected tests without a delay). Notably, when the memory assessment was expected, the likelihood of a successful imitation did not decline (even after a long delay).
Incidental Encoding Although the evidence is compelling that the memory assessment was unexpected, it is difficult to be certain that the encoding was incidental given the action–imitation training history. Because action demonstrations were reliably followed by a ‘‘do it’’ request, action demonstrations may have been explicitly encoded, at this stage, in anticipation of the memory assessment. To discourage explicit encoding, the dogs were subsequently retrained with a different trained action (to spontaneously lie down when directed to the carpet). To highlight an approach to incidental encoding, I will describe an experiment from my lab on this problem which used an eight-arm radial maze . Rats foraged on five arms, depleting food as they searched locations. On other occasions, using three other arms in the shape of a T, the rats learned to ‘‘report’’ about the presence or absence of food by making left/right turns. On a critical test, the animals were permitted to forage and then prompted, for the first time after foraging, to report whether they had just eaten food or not. After discovering food, the rats made the corresponding turn to indicate that they remembered the food whereas they turned in the opposite direction when they foraged but did not find any food . Rats naturally and spontaneously — without the need for training — forage by avoiding revisits to recently visited locations [17,18]. Foraging likely provided the opportunity for incidental encoding, because explicitly encoding the presence of food during foraging would have produced a preponderance of turns in the corresponding direction, which would preclude the observed high level of spatialmemory accuracy. Moreover, when the hippocampus was temporarily inactivated,
REFERENCES 1. Fugazza, C., Poga´ny, A´., and Miklo´si, A´. (2016). Recall of others’ actions after incidental encoding reveals episodic-like memory in dogs. Curr. Biol. 26, 3209–3213. 2. Akins, C.K., and Zentall, T.R. (1996). Imitative learning in male Japanese quail (Coturnix japonica) using the two-action method. J. Comp. Psychol. 110, 316. 3. Fugazza, C., and Miklo´si, A´. (2014). Deferred imitation and declarative memory in domestic dogs. Anim. Cogn. 17, 237–247. 4. Fugazza, C., Poga´ny, A´., and Miklo´si, A´. (2016). Do as I. Did! Long-term memory of imitative actions in dogs (Canis familiaris). Anim. Cogn. 19, 263–269. 5. Tulving, E. (1985). Memory and consciousness. Can. Psychol. 26, 1–12. 6. Clayton, N.S., Bussey, T.J., and Dickinson, A. (2003). Can animals recall the past and plan for the future? Nat. Rev. Neurosci. 4, 685–691. 7. Suddendorf, T., and Busby, J. (2003). Mental time travel in animals? Trends Cogn. Sci. 7, 391–396. 8. Corballis, M.C. (2013). Mental time travel: a case for evolutionary continuity. Trends Cogn. Sci. 17, 5–6. 9. Crystal, J.D. (2013). Remembering the past and planning for the future in rats. Behav. Process. 93, 39–49. 10. Crystal, J.D. (2016). Animal models of source memory. J. Exp. Anal. Behav. 105, 56–67. 11. Zentall, T.R., Clement, T.S., Bhatt, R.S., and Allen, J. (2001). Episodic-like memory in pigeons. Psychonom. Bull. Rev. 8, 685–690. 12. Singer, R.A., and Zentall, T.R. (2007). Pigeons learn to answer the question ‘where did you just peck?’ and can report peck location when unexpectedly asked. Learn. Behav. 35, 184–189. 13. Zentall, T.R., Singer, R.A., and Stagner, J.P. (2008). Episodic-like memory: Pigeons can report location pecked when unexpectedly asked. Behav. Process 79, 93–98. 14. Wang, S.-h., Baillargeon, R., and Brueckner, L. (2004). Young infants’ reasoning about hidden objects: evidence from violationof-expectation tasks with test trials only. Cognition 93, 167–198.
Current Biology 26, R1226–R1245, December 5, 2016 R1227
Dispatches 15. Pattison, K.F., Laude, J.R., and Zentall, T.R. (2013). The case of the magic bones: Dogs’ memory of the physical properties of objects. Learn. Motiv. 44, 252–257.
17. Gaffan, E.A., and Davies, J. (1981). The role of exploration in win-shift and win-stay performance on a radial maze. Learn. Motiv. 12, 282–299.
19. Tulving, E., and Markowitsch, H.J. (1998). Episodic and declarative memory: Role of the hippocampus. Hippocampus 8, 198–204.
16. Zhou, W., Hohmann, A.G., and Crystal, J.D. (2012). Rats answer an unexpected question after incidental encoding. Curr. Biol. 22, 1149– 1153.
18. Timberlake, W., and White, W. (1990). Winning isn’t everything: Rats need only food deprivation and not food reward to efficiently traverse a radial arm maze. Learn. Motiv. 21, 153–163.
20. Andics, A., Ga´bor, A., Ga´csi, M., Farago´, T., Szabo´, D., and Miklo´si, A´. (2016). Neural mechanisms for lexical processing in dogs. Science 353, 1030–1032.
Plant Evo–Devo: How Tip Growth Evolved Stefan A. Rensing Faculty of Biology, University of Marburg, Marburg, Germany Correspondence: [email protected]
Apical elongation of polarized plant cells (tip growth) occurs in root hairs of flowering plants and in rhizoids of bryophytes. A new report shows that the formation of these cells relies on genes already present in the first land plants. Learning how structures and functions in flowering plants evolved is often difficult without a frame of reference. The use of plant lineages that diverged early from what was to become the flowering plant lineage has greatly aided the discipline of plant evolutionary developmental (evo-devo) studies over the past decade. In particular, the model moss Physcomitrella patens and, more recently, the liverwort Marchantia polymorpha are employed for such studies. Apical growth of filamentous cells (tip growth) is well documented in root hairs and pollen tubes in flowering plants. In a new study, published in this issue of Current Biology, Liam Dolan and colleagues  show that the genes
controlling the cytoskeleton and surface formation of such polarized cells are an ancestral feature of the first land plants. It had previously been shown that the formation of rhizoids (analogous to root hairs in providing nutrition and anchorage) of P. patens is controlled by genes orthologous to those controlling Arabidopsis thaliana root hair development . Moreover, genes of the basic helix-loop-helix transcription factor (bHLH-TF) family in question control the development of genes with rooting functions in the monocot Oryza sativa and the legume Lotus japonicus as well, and thus represent an ancestral regulatory module . This module originally evolved
LATER... I made it to land, but how do I get nutrients and propagate? let’s try protrusions!
Instead of making a thallus after germination, what if I form filaments instead, to spread and regenerate after stress?
EVEN LATER... Roots are so thick...let’s make root hairs using those genes for protrusion I had when I was a gametophyte.
in the haploid gametophytic generation and was co-opted during plant evolution to control similar structures in the diploid sporophyte. In the moss P. patens and the liverwort M. polymorpha, the bHLH-TF sub-family ROOTHAIR DEFECTIVE SIX-LIKE (RSL) controls the protrusion of uni- and multicellular structures (such as gemmae, vegetative propagules) from single epidermal cells . Tip-growing rhizoids are also seen in charophyte algae, the sister lineage of land plants. The evolution of such a tip-growth module might have been an important feature that allowed adaptive morphological innovations in the earliest land plants  (Figure 1).
Motile gametes’ need for water is inconvenient. Let’s use those tip growth genes to make pollen tubes!
Figure 1. Land plant evolution. Cartoon of some of the evolutionary scenarios discussed here. In the left panel, the evolution of cellular protrusions for anchorage, nutrition and propagation in the first land plant is depicted, while the middle panel shows the hypothetical secondary evolution of tip growth featuring cell division in mosses. The panel to the right shows an early seed plant, living approximately in the Middle Devonian (380–400 million years ago), evolving root hairs and pollen tubes. Accurate placement of these features is difficult due to absence in the early fossil record. Artwork by Debbie Maizels, Zoobotanica Scientific Illustration.
R1228 Current Biology 26, R1226–R1245, December 5, 2016 ª 2016 Elsevier Ltd.