1
Table S1. (Whiten & Mesoudi) Summaries of published diffusion experiments
Publication / Species studied / Experimental design / Findings, authors’ conclusions / Commentary / critical appraisalMenzel et al. 1972 / chimpanzees,
Pan troglodytes / No trained model. Seventeen consecutive trios of juvenile chimpanzees, with one chimpanzee replaced by a naïve one each ‘generation’, were exposed to two novel objects. No non-observing controls were included, but the initial, naïve trio provided baseline data. / Early trios showed avoidance of the strange objects, but as habituation proceeded, it became the norm for the later trios. This occurred within trio 4 for a swing, within trio 8 for a more alarming robotic device. Authors concluded “a culture-like process was at work in the data”. / This appears to be the first, pioneer study to apply the diffusion-with-replacement method. The core conclusions of the authors seem entirely justified: a tradition of active engagement with the objects replaced the avoidance of early generations. Authors explicitly acknowledged they had not investigated underlying mechanisms.
Curio et al. 1978a,b / blackbirds,
Turdus merula / An initial model showed alarm behaviour to a stuffed predator (owl) while an observer simultaneously observed only a non-predator (honey eater). The observer later became a model for a further observer, and so on along a transmission chain restricted to the honey-eater only. Each observer was first exposed to this stimulus in a baseline test. / Alarm responses significantly above baseline rates were passed through six transmission events without decrement. In a separate, dyadic model-observer experiment, elevated alarm responses were also shown in relation to a plastic bottle, but the effect was less than for the honey-eater stimulus. The data were interpreted as supporting a “cultural transmission hypothesis” for the function of mobbing among kin. / This appears to be the first, pioneer experiment to exploit a linear, rather than replacement, transmission chain design. Results clearly demonstrated transmission along a series of model-observer pairs.
Cambefort 1981 / baboons,
P ursinus; vervets,Cercopithecus aethiops / Novel cues coupled to hidden food items were introduced to wild groups. Initial discoverers were designated ‘teachers’; later discoverers were described as the ‘taught’. / Spread of cue recognition was described. / Apparently the first ‘open group’ diffusion experiment, but numerous details of methods were not described, so that evidence for observational rather than individual learning cannot be ascertained.
Sumita et al. 1985 / chimpanzees, Pan troglodytes / Three chimpanzees that learned to crack nuts were re-introduced to their group containing nine others. No baseline for latter. / One three-year-old acquired nut-cracking, other adults and young did not. Authors concluded stimulus enhancement plus trial and error accounted for the one success. / Lack of any baseline data for the three-year-old means inference of social learning is weak.
Lefebvre 1986 / pigeons,
Columbia livia / Models trained to peck through paper covers to obtain grain were introduced into marked, feral, urban flocks and a captive colony. A separate urban flock acted as a control group. / Once models began to pierce, some naïve urban flock members began to do so within seconds. The behaviour spread to 24 naïve birds during 30 presentation sessions over 55 days, although to only one other in the captive group (others scrounged). By contrast, piercing did not appear until presentation 14 in the naïve control group. Diffusion through social learning was demonstrated, including factors accelerating or limiting it. / In separate experiments, Palameta & Lefebvre (1985) dissected several aspects of the social learning mechanisms pigeons employ in acquiring the piercing technique. These include stimulus enhancement, but this was not ‘blind’ insofar as observing a bird piercing, then feeding, was significantly more effective than observing piercing alone.
Hannah & McGrew 1987 / chimpanzees, Pan troglodytes verus / One of three captive chimpanzees released on an island in Liberia began cracking nuts. Spread of this behaviour was recorded. / Whereas 10 residents had never done so, three began to crack the same day as the first one began cracking and in total eight were cracking nuts within one month. / Chimpanzees were captured as infants in the wild too young to nutcrack but may have witnessed cracking in the wild. The authors concluded “the sequence of events suggests that they acquired the nut-cracking technique by observation, or at the very least, had their memories prompted by her actions” (p. 43). This conclusion appears appropriate.
Laland & Plotkin 1990 / rats,
Rattus norvegicus / Experiment 2 in this paper was a transmission chain study of digging up carrot pieces, contrasting control animals who saw no model with those in chains composed of 8 consecutive model-observer pairs in one of three conditions: (i) each observer became model for the next; (ii) same but observer also had individual learning time as well, before becoming a model; (iii) no model was trained – rats had to discover carrot themselves. / Rats in conditions (i)-(iii) gained more carrot pieces than controls, a social learning advantage maintained through all 8 ‘generations’. The effectiveness of rats in conditions (i) and (ii) dropped a little to stabilise after about five transmission episodes whereas in condition (iii) discovery of food led to a rise in items recovered that also stabilised at similar levels to (i) and (ii) by step five. / This was the first transmission chain experiment on foraging behaviour and the first to include careful between-group controls, as well as the different combinations of social and individual learning opportunities listed here. It provided clear demonstrations of social transmission across the 8 generations tested.
Laland & Plotkin 1992 / rats,
Rattus norvegicus / Experiment 2 in this paper was a transmission chain study of digging up carrot pieces, with 3 conditions: (i) as in the prior 1990 study, 8 consecutive model-observer pairs, but now with a 24-hour delay between transmission; (ii) the same but with additional individual experience in this period; (iii) controls observing no model. / Group (ii) gained significantly more carrot pieces than (i), and both of these significantly more than the controls (iii). However the magnitude of differences was not great and in the final session the success of rats in condition (i) was the same as controls. / This study showed that social learning can maintain superiority in a particular form of foraging over repeated transmission episodes separated by as long as 24 hours, particularly if there is also time for individual learning to improve performance within this period.
Paquette 1992 / chimpanzees, Pan troglodytes / Four subjects were given the opportunity to use tools to probe for honey, with non-probers potentially able to learn from probers; no trained model, no non-observing controls. / The probing technique spread among all four chimpanzees. There was circumstantial evidence that the 3rd and 4th observed the success of the 2nd before succeeding. / Absence of a control condition means that it is not possible to reject with confidence the hypothesis that effects observed were due to increasing individual experience over time.
de Waal & Johanowicz 1993 / rhesus, Macaca mullatta; stumptailed macaques,
Macacaarctoides. / Juvenile rhesus and stumptailed macaques were reared together for 5 months: reconciliation behaviour is common after fights in stumptails, who were denoted tutors to the rhesus, among whom reconciliation is rare. Control rhesus were housed with conspecifics. / Rhesus reared with stumptails were three times more likely to show reconciliation behaviour than controls. The authors conclude that social transmission of ‘social attitude’ involving a ‘more relaxed dominance style’ was transmitted. / The evidence for social transmission is compelling, although whether the greater tendency to reconcile diffused through the rhesus or was directly learned from stumptails is not known. This is an unusual ‘open diffusion’ design with approximately as many tutors as learners.
We have classed this in row 5 in Table 2 but it could be seen as appropriate for row depending on how the conspecific control condition is viewed.
Laland & Plotkin 1993 / rats,
Rattus norvegicus / Transmission chain study of diet preference, with each of 2 batches of 8 transmission chains initiated by a rat with a preference for one or other of two flavours; matched control individuals received no information from others. / In two experiments (Expts 2 & 3) preference was transmitted along chains of 8 rats differentially according to the founder diet, but fidelity was also influenced by inherent preferences and the kinds of information source available, including gustatory cues in the breath and excretory products near the food site. / This appears to be the first example of the most complex and powerful design we favour (row 7 in Table 2), employing two alternative seeded behaviours plus a control group. In addition, this is the first and perhaps only study to date to tease out learning mechanisms within a transmission procedure.
Matsuzawa 1994; Biro et al. 2003 / chimpanzees, Pan troglodytes / Two species of nut (coula, panda) cracked at other sites in Guinea were introduced at Bossou where these nuts are absent and chimpanzees crack only oil-palm nuts. / One chimpanzee (a suspected immigrant) readily cracked the coula nuts and others observed this, sometimes intently. Coula nut cracking spread from 11% adults in 1993 to 67% in 2002. A mixture of evidence including perseverance of youngsters in cracking without reward leads the authors to infer that social learning is involved. / The circumstantial evidence is compelling as far as it goes, but given the lack of any control conditions means it seems impossible to rule out a scenario in which increasing individual experience explains the rising frequency of coula cracking over the years.
Galef & Allen 1995 / rats,
Long-Evans Rattus norvegicus / Founder groups of 4 rats were trained to eat only one of a pair of differently-flavoured foods. In later sessions, one rat was replaced by a new, naïve one (the ‘replacement method’) and the new group exposed to the same choices of food types. / Experiments 1-3 showed that if exploration time is limited, stable traditions of preference can be maintained through 14 consecutive quads, amounting to 4 entire replacement generations. Experiments 4-5 showed that olfactory and stimulus enhancement cues were not necessary: rat-rat contact was sufficient for transmission to be sustained. / The study was explicitly (and correctly) presented as offering a “new model system” for studying cultural transmission (i.e. two alternative behaviour patterns seeded, coupled with the replacement method to simulate plausible social context). Surprisingly the approach was not adopted for a decade (see Whiten et al. 2005, Horner et al. 2006, Bonnie et al. 2007 for use of two seeded alternative behaviour patterns).
Langen 1996 / white-throated magpie jays,
Calocitta formosa / Birds trapped in the wild were trained whilst in captivity to open just one of up to three different kinds of doors to gain food from a foraging box, then released back to their groups. The box was then presented in their home range. Controls saw no models. / Only two of 14 models would perform in front of naïve birds. Only for one of these models was identity of observers known. Opening doors spread preferentially among observers of models, who ignored one of the alternative doors but later showed a preference for the third over that the model had opened. / Provides salutary information on reluctance of captive-trained birds to later act as models. Nevertheless, provided robust evidence of spread of door-opening once models were observed, with generalisation to other doors requiring a somewhat different technique.
Tonooka et al. 1997 / chimpanzees, Pan troglodytes / Juice was provided in a container from which it could potentially be extracted using natural tool materials available in the compound. No trained model, no non-observing controls. / Spontaneous spread of using a tool to extract juice was recorded. On 9th day, use of one among the 15 materials ever tried – the leafy stem of a particular species - became the stable pattern. Authors inferred this was a socially learned tradition. / The interpretation is plausible taking the data set as a whole. However, absence of a control condition, or a condition in which preference for an alternative tool-type was seeded, weakens the inference that social learning was involved rather than progressive individual discovery of the inherently optimal tool material.
Laland & Williams 1997 / guppies,Poecilia reticulata / Founder shoals of 4 fish were trained to prefer one of two doorways through which to swim from one tank compartment to a second containing food; then on 7 consecutive sessions, a naïve fish was introduced, replacing an experienced one. / Differential use of the two doorways was preserved along chains. The difference in preference approximately halved over the 7 transitions, but remained clear at this point. / Demonstration of socially mediated transmission along ‘cultural generations’ was clear, but extrapolation of graphs suggests behaviour of the different groups would converge by about 14 days.
Laland & Williams 1998 / guppies,Poecilia reticulata / Similar to the authors’ prior paper but fish had a choice of the modelled route or a shorter, energetically less costly alternative route. / Over 7 transmission sessions, differential usage of the long versus short routes was maintained. However, after 4-5 days differences between populations that used one long route versus another long route had eroded. / This study demonstrated both that a relatively maladaptive (energetically costly) behaviour pattern can be maintained by social learning, and that individual learning in this context can work in opposition to gradually erode the tradition in favour of an adaptive one.
Freeberg 1998; Freeberg et al. 2001 / cowbirds,
Molothrus ater / Young cowbirds were housed with either of two groups of adults from different regions, that sang distinctly different songs. This process was repeated later once the first generation of young had matured and could be used as models. / When mature, birds from the different rearing groups sang songs that on some measures showed virtually no overlap, but matched the songs of the original adults, and they preferentially paired with unfamiliar birds that had been reared with the same social background, in both the first (70 birds) and the second (56 birds) ‘cultural generations’. / It appears true that as the author claims, this is ‘the first experimental evidence for the cultural transmission of courtship patterns in animals’.
Reader & Laland 2000 / guppies,Poecilia reticulata / Spread of accessing a food source via a novel route was studied in populations containing both naïve and experienced fish that varied systematically in sex, age and size. No no-model controls. / Significant effects were found among the individual differences variables’ effects on the spread of use of the route. / Because no no-model controls were used, whether the spread of behaviour documented was due to social learning was apparently not determined. It is thus not clear if ‘diffusion’ is the correct term here, if diffusion is defined as occurring through social rather than individual learning.
Swaney et al. 2001 / guppies,Poecilia reticulata / Spread of accessing a food source via a novel route was studied in populations containing both naïve and experienced fish, the latter varying in familiarity and experience. No no-model controls. / Significant effects of model variables were found. / As in the previous study, no no-model controls were used, so that whether the spread of behaviour documented was due to social learning was not determined.
Brown & Laland 2002 / guppies,Poecilia reticulata / Spread of learning to use one of either of two routes to escape a threat was studied in groups containing naïve fish and models experienced in using one route, or sham (inexperienced) models. / Naïve fish strongly tended to follow the route of the model, but did not persevere in this choice of route when models were removed. / We interpret the finding that fishes’ tendency to use the same route as the model disappeared when no model was present as a lack of evidence of social learning, despite the title of the paper employing this title.
Cloutier et al. 2002 / chickens,Gallus gallus domesticus / Nine flocks of four chickens were each exposed to two models feeding ‘cannibalistically’ on blood that could be pecked from a mock hen, then later tested in pairs with no model present. Testing for control birds, that saw no model initially, was also in pairs. / Feeding on blood spread in both control and experimental conditions but was significantly enhanced in groups that saw initial models, especially in a condition where observers were in contact with model and target. Spread of the behaviour conformed to a sigmoidal diffusion curve. / The social learning effect demonstrated means that the flocks exposed to a model could be described as developing a tradition of cannibalism. However, by the third trial of the experiment, control birds already approximated the level of attack shown in the groups originally seeded with models.
Gajdon et al. 2004 / keas,
Nestor notabilis / One model was discretely trained to climb a vertical stick and use its beak to push a small food-holding tube off the stick. The foraging device was then offered to the model and a population of 12 banded birds, of which 5 typically were present. / Only one bird succeeded in baseline control tests. However only two observers later succeeded and did not do so immediately, so the authors concluded evidence of social learning was weak. This was contrasted with informal evidence of fast social learning in two captive birds tested. / The evidence for social learning in captive birds is quite anecdotal, but the apparent difference between learning in captivity and the wild appears striking. Such field experiments remain so rare it may be premature to draw firm conclusions on the extent to which these results reflect the true status of social learning in these birds’ lives.