Animal Cognition

, Volume 18, Issue 1, pp 393–397

‘The thieving magpie’? No evidence for attraction to shiny objects


    • Centre for Research in Animal BehaviourUniversity of Exeter
  • S. E. G. Lea
    • Centre for Research in Animal BehaviourUniversity of Exeter
  • N. Hempel de Ibarra
    • Centre for Research in Animal BehaviourUniversity of Exeter
Short Communication

DOI: 10.1007/s10071-014-0794-4

Cite this article as:
Shephard, T.V., Lea, S.E.G. & Hempel de Ibarra, N. Anim Cogn (2015) 18: 393. doi:10.1007/s10071-014-0794-4


It is widely accepted in European culture that magpies (Pica pica) are unconditionally attracted to shiny objects and routinely steal small trinkets such as jewellery, almost as a compulsion. Despite the long history of this folklore, published accounts of magpies collecting shiny objects are rare and empirical evidence for the behaviour is lacking. The latter is surprising considering that an attraction to bright objects is well documented in some bird species. The present study aims to clarify whether magpies show greater attraction to shiny objects than non-shiny objects when presented at the same time. We did not find evidence of an unconditional attraction to shiny objects in either captive or free-living birds. Instead, all objects elicited responses indicating neophobia in free-living birds. We suggest that humans notice when magpies occasionally pick up shiny objects because they believe the birds find them attractive, while it goes unnoticed when magpies interact with less eye-catching items. The folklore may therefore result from observation bias and cultural inflation of orally transmitted episodic events.


CorvidObject attractionMagpieNeophobiaNest ornamentation


It is widely believed across Europe that magpies (Pica pica) are unconditionally attracted to, and collect, shiny objects. A Google search on the term ‘magpie shiny object’ produced 60,100 results in 0.28 s (February 2014). Most of these are descriptions of magpies that make reference to their ‘weakness for shiny things’ (Winterman 2008) or reputation for ‘stealing jewellery’ (BBC 2003). Although such assertions are usually stated as fact, no references or examples are given. The belief is so ingrained in our culture that one definition given for ‘magpie’ in the Collins English Dictionary is ‘a person who hoards small objects’ (Collins 2014). Perhaps, the most famous example of this folklore penetrating popular culture is Rossini’s opera ‘The thieving magpie’, first performed in 1817, where a servant girl is executed for a series of silver thefts that were actually committed by a magpie. Despite the folklore’s long history and widespread acceptance, published accounts of magpies stealing or collecting shiny objects are rare and empirical research on the subject is lacking.

An extensive Internet search uncovered just two published accounts of magpies stealing shiny things: a missing engagement ring that was discovered in a magpie nest (Telegraph 2008) and a magpie in Rochdale stealing keys, coins, and a spanner (a shiny tool) from an automotive garage (Manchester Evening News 2007). Without empirical research, it is not known whether these birds had an unconditional attraction to shiny objects, or whether the folklore has caused observation bias in the human population. In other words, we notice when magpies collect shiny objects because we ‘know’ they are attracted to shiny objects but we do not notice when they interact with less eye-catching items. A non-specific attraction to objects was demonstrated by Heinrich (1995) with four hand-raised juvenile ravens, birds who are also reputedly attracted to shiny objects. The young birds contacted novel items placed in their aviary significantly more than familiar items; however, their attraction depended solely on the novelty of the item ‘without relevance to its palatability or shininess’.

Research investigating cognitive abilities and object interaction in magpies is scarce. However, attraction to shiny or bright objects has been documented in several other bird species, with many using them as non-body ornaments, to attract mates or influence parental investment (outlined in Schaedelin and Taborsky 2009). Black kites (Milvus migrans) place them in the nest to act as threats to conspecifics, revealing the viability and conflict dominance of the signaller (Sergio et al. 2011). Some corvid species are known to collect and cache non-food objects, and this may be an expression of exploration, although this behaviour does not appear to be colour-specific (Jacobs et al. 2014). We performed experiments with captive wild-born magpies and free-living magpies (during breeding and non-breeding periods) to test for an unconditional attraction to shiny objects or any sign of collection of non-food objects, and to investigate possible explanations for such behaviour. In the absence of such an attraction, a neophobic response to objects might be expected; accordingly, we recorded behaviour that could indicate either neophobia or ambivalence towards the objects. Neophobia was measured as a reduction in the rate of feeding in the presence of the test objects. Five behaviours, such as approach/withdrawals (see Greenberg and Mettke-Hofmann 2001), walk-bys, flyovers, jumping jacks (see Heinrich 1988), and neck extensions, were classified as ambivalent behaviour, and ambivalence scores were calculated as the sum of the numbers of all these behaviours exhibited in a session. Full operational definitions of these behaviours are given in Table S1.


Captive study

Eight magpies were obtained from three wildlife rescue centres in the United Kingdom. All birds were deemed to be suitable for release back into the wild at the time of acquisition. They were housed in an aviary on the roof of the laboratory. The accommodation provided a shed and a flight cage (Fig. S1). There was a large, hinged hatch in the wall of the shed adjacent to the flight cage. A large tray filled with soil was situated in the flight cage (outside of the test arena) so the birds could cache food or other objects. A second large tray filled with water was provided for bathing. The birds were colour banded for individual identification. All birds that arrived in a group were housed as a group. They were locked in the shed overnight, and while test materials were being set out, during experimental sessions, birds not being tested were also locked in. At all other times, all birds had free access to both the shed and the flight cage.

On arrival at the laboratory, the birds were habituated to feeding from two plastic trays situated at the opposite ends of the main aviary (test arena, Figure S1). The shiny object tests formed part of a random sequence of novel object tests that followed habituation; the other tests involved presenting birds with familiar shaped trays in novel colours (black, silver, and red or blue) and novel shaped trays (8 cm deeper) in the familiar colour. In all tests, the trays contained mealworms. In the shiny objects test, two piles of six zinc-plated steel woodscrews were placed in the aviary, each 20 cm from a feeding tray. One pile contained silver screws and the other blue screws. This test was only performed once with each bird. The screws were 7 cm long. Before the study began, half of the screws were painted blue with matt aerosol paint. The remaining screws were left in their original shiny silver colour.

All sessions were videotaped with a tripod-mounted video camera situated inside an observation hide located approximately 20 m from the flight cage. The tests were performed between May and August 2010.

Field study

The field experiment was conducted at eight sites on the university campus, where magpies are accustomed to regular human activity, allowing observations to be conducted from a close proximity (<20 m). To prevent pseudoreplication, the sites were separated by a minimum of 500 m, about twice the maximum extent of territories defended by breeding magpies across northern Europe (Birkhead 1991, p. 43). Each study site had a resident pair of magpies, identified as the same pair in each session by their habituation to the observer and to the feeding protocol, as well as their defence of the feeding area when other magpies approached. This provided a total of 16 birds in the field study, but, as the two birds at each site could not be reliably individually identified, they are treated as one unit for the analyses of mean feeding latency and frequency and the mean ambivalence score.

The experiment took place in two phases. The first series of tests were performed at six sites during the non-breeding season, from August to September 2011. The same tests were then repeated at four of the same sites and two new sites throughout the breeding season, February to April 2012. The new sites were required as the birds from two summer sites did not re-engage in the breeding season.

The test objects were screws as used in the captive study, small foil rings (3 cm diameter), and a small rectangular piece of aluminium foil (7 cm × 5 cm). Half of the screws and rings were painted blue with matt aerosol paint, and the rest left in their original shiny silver colour, as was the piece of aluminium foil.

Before beginning each phase, food (monkey nuts) was provided in two loose piles on the ground daily at each study site for 2 weeks and the observer recorded data as in the intended tests. The birds were considered to be habituated when all of the food provided was taken in the presence of the observer and within 20 min of provisioning. Food continued to be provided throughout the study, with test objects placed next to the food during test sessions. Once the birds were habituated, three control sessions were conducted to determine the birds’ feeding motivation and whether they had a side preference (right or left side of the feeding area). Two equal portions of nuts were placed in loose piles on the ground approximately three metres apart, and at roughly equal distance from the observer and from tree cover. Figure S2 illustrates these procedures.

Shiny object tests were conducted once three successful control sessions had been completed (success was defined as the birds appearing within 15 min and feeding from the piles). Two loose piles of nuts were placed on the ground in the same locations as in the control sessions but with the addition of two piles of objects (screws or rings), each placed 30 cm from a nut pile. One object pile contained four silver objects and the other four blue objects. During the breeding season, a further test was conducted with a rectangular piece of aluminium foil placed on the ground next to one nut pile and secured with a small hair pin (Fig. S2). In the non-breeding phase, two object tests were performed at each site, one with screws and one with rings. In the breeding phase, a minimum of eight tests were performed at each site, with multiple exposures to the rings and screws and one test with the foil rectangle. The tests were spaced out to incorporate early nest-building, final nest-building, and incubation and were performed in a random sequence. In both phases, the sequence of tests varied across sites. Control sessions lasted for 30 min or until the food was depleted, whichever came first. All test sessions lasted for 30 min, even if the food was depleted sooner, to give the birds ample time to investigate the objects. If no magpies arrived in the first 15 min of an experimental session, it was abandoned and its data were discarded.

Binoculars (10× magnification) were used to confirm the number of food items taken in each visit. All sessions were videotaped with a tripod-mounted video camera situated 15 to 20 m from the food and positioned to give a panoramic view of the feeding area. The video footage was viewed to measure feeding latency and feeding frequency, which were considered to be the measures of neophobia, and to record ambivalent behaviours. One person coded the sessions from the non-breeding phase and another coded those from the breeding phase, and also recoded three sessions from the non-breeding phase. Pearson’s correlations and comparisons of means confirmed consistent coding: the correlations between coders were 1.00, 0.99, and 0.99 for feeding latency, feeding frequency, and ambivalence scores, respectively, P < 0.05 in all cases, and the mean ± SE for the two observers was 246 s ± 202 and 245 s ± 202 for feeding latency, 7.33 ± 1.09 and 8.67 ± 1.61 for feeding frequency, and 11 ± 5.86 and 12 ± 6.49 for ambivalence score.

Statistical procedures

Data from the captive test and the foil test were analysed using Wilcoxon rank sum tests. General estimating equations (GEE) were used to analyse the remaining field data to accommodate the non-normal distribution and the use of repeated measures (Garson 2012). The most appropriate model type, working correlation matrix structure, and subset of predictors were chosen based on the goodness-of-fit statistics QIC (quasi-likelihood under independence criterion) and QICC (a corrected version that rewards parsimony). Feeding latency data were analysed using identity-inverse Gaussian regression, whereas feeding frequency and ambivalence data were analysed using negative binomial regression. Least significant difference tests were used to examine pairwise contrasts. Spearman’s correlation coefficients were used to explore whether ambivalent behaviour changed with repeated exposure to the objects in the field study. GEE analyses on feeding latency, feeding frequency, and ambivalence scores were first run with the control sessions split into multiple phases, each comprised of the three sessions preceding a test. No qualitative differences between the control sessions across the different phases were found so their data were combined and treated as one control condition. This resulted in a better model fit for each of the analyses. All analyses were carried out using SPSS version 20.


Captive study

None of the eight birds made contact with any object, shiny or blue. The presence of the objects did not deter them from feeding from the trays next to the objects. Feeding latencies in this test were not significantly different to the preceding control session (mean ± se: control = 42 s ± 13, screw test = 23 s ± 19; W = 16, N = 8, P = 0.313). Additionally, there was no significant difference in feeding latency between the trays next to silver and blue screws (mean ± SE: silver = 325 s ± 161, blue = 370 s ± 223; W = 4, N = 8, P = 0.844). The birds exhibited little ambivalence in the presence of the screws, with scores in this test lower than the other novel object tests conducted during the study (mean score ± SE: shiny object test = 3.13 ± 1.34, other tests in captive study ranged from 8.88 ± 2.70 to 25.88 ± 7.97).

Field study

Magpies only made contact with a shiny object twice in 64 tests. Both incidents occurred at the same site, once in the non-breeding phase and once during early nest-building. One silver ring was picked up and immediately discarded in both instances. No other object interactions occurred at any site.

Figure 1 shows that birds in the field appeared to find the test objects aversive, with significant differences in feeding latencies and frequencies for the food piles across three conditions (control, ring test, and screw test; feeding latency: χ2 = 23.52, df = 2, P < 0.001; feeding frequency: χ2 = 12.68, df = 2, P = 0.002). Pairwise contrasts between conditions confirm that magpies fed from the piles significantly faster and took significantly more nuts in the control condition than when objects were present. In the foil test, magpies showed a preference for the control pile over the pile next to the foil rectangle, feeding from the control pile sooner (W = 1, N = 6, P = 0.080; Fig. 1a) and taking significantly more nuts from it (W = 0, N = 6, P = 0.041; Fig. 1b). The aversion to feeding from food piles beside objects remains when the non-breeding and breeding phases are examined separately (test*season interaction: feeding latency: χ2 = 5.87, df = 2, P = 0.053; feeding frequency: χ2 = 1.28, df = 2, P = 0.527).
Fig. 1

Feeding behaviour of free-living magpies (6 sites, 202 total trials). Birds a fed more quickly and b took more food items when no objects were present. The two columns in control condition represent the left and right food piles. Error bars show the standard error. *P < 0.05; **P < 0.005

Figure 1 also shows that screw colour (shiny or blue) did not affect magpies’ reactions to the food piles during either phase (breeding and non-breeding). The birds took significantly longer to feed from piles beside blue rings in the breeding season (test*colour*season interaction: feeding latency: χ2 = 14.16, df = 4, P = 0.007; Fig. 1a) but there was no difference in the non-breeding phase. Ring colour did not affect feeding frequency in either phase (test*colour*season interaction: feeding frequency: χ2 = 5.09, df = 4, P = 0.278, Fig. 1b). Colour comparisons were not confounded by a positional bias (mean feeding latency ± SE: right pile = 350 s ± 49, left pile = 298 s ± 40, χ2 = 1.62, df = 1, P = 0.203; mean feeding frequency ± SE: right pile = 4.65 ± 0.31, left pile = 4.90 ± 0.41, χ2 = 2.62, df = 1, P = 0.106).

Figure 2 shows that the pattern of ambivalent behaviour also implies that the birds were wary of the objects. There was a significant difference in ambivalence scores across the four conditions (control, ring test, screw test, and foil test; χ2 = 27.5, df = 3, P < 0.001). Pairwise contrasts between conditions reveal the magpies displayed significantly more ambivalence in the ring and screw tests than in the control sessions. Ambivalence scores decreased with repeated exposure to the objects (N = 6, n = 163, r = −0.187, P = 0.017).
Fig. 2

Ambivalence scores (±se) for field-tested magpies (6 sites, 202 total trials). Birds showed more ambivalence when test objects were present. *P = 0.050; **P ≤ 0.026


This study exonerates Rossini’s magpie, finding no evidence to suggest the species is unconditionally attracted to shiny objects. The birds either ignored or avoided both shiny and blue objects, with little difference observed between breeding and non-breeding seasons. There were two interactions with shiny rings at one site, but these occurred after the food was depleted rather than when the objects were first presented, suggesting investigations into a potential food item rather than unconditional attraction.

Rather than being attractive to magpies, the objects appeared to induce neophobia in most of the birds in the field, demonstrated by increased feeding latency, decreased feeding frequency, and more ambivalent behaviour when objects were present than during control sessions. The free-living birds reacted this way to both shiny and blue objects, as well as to different shaped objects. The decline in ambivalence scores with repeated exposure to the objects suggests the birds did habituate to their presence over time. This agrees with the descriptions of object neophobia in other corvid species (Greenberg 2003; Greenberg and Mettke-Hofmann 2001), and we have investigated it more fully in other studies (Vernelli 2013). Birds in captivity did not show any object-related avoidance. Similar differences between wild and captive populations have been observed in kea (Huber and Gajdon 2006) and spotted hyenas (Bensom-Amram et al. 2013). They may reflect the artificial nature of captive settings where animals are confined in close proximity to novel stimuli, whereas wild animals are free to avoid fear-provoking objects.

We suggest the widely held belief that magpies are unconditionally attracted to shiny objects results from cultural generalisation of anecdotal evidence, demonstrating how biases in observations of animals may determine the broad public perception of species and their impact on the environment.


We thank RSPCA, Secret World Wildlife Rescue, and Crows are Us for providing rescued magpies for the captive study. TVS was funded by the University of Exeter Research Studentship.

Conflict of interest

The authors declare that they have no conflict of interest.

Supplementary material

10071_2014_794_MOESM1_ESM.docx (1.6 mb)
Supplementary material 1 (DOCX 1594 kb)

Copyright information

© Springer-Verlag Berlin Heidelberg 2014