Real and Phantom Decoys in Capuchin Monkey (Sapajus spp.) Decision-Making

doi:10.21203/rs.3.rs-3287219/v1

Decision-making has been observed to be systematically affected by decoys, i.e., options that should be irrelevant, either because unavailable or because manifestly inferior to other alternatives, and yet shift preferences towards their target. Decoy effects have been extensively studied both in humans and in several other species; however, evidence in non-human primates remains scant and inconclusive. To address this gap, this study investigates how choices in capuchin monkeys (Sapajus spp.) are affected by two types of decoys: asymmetrically dominated decoys, i.e., options that are inferior to one of the other alternatives, and phantom decoys, i.e., unavailable options that are superior to another available alternative. After controlling for the subjective strength of initial preferences and the distance of each decoy from its target in attribute space, results demonstrate a systematic shift in capuchins’ preference towards the target of both asymmetrically dominated decoys (whether they are available or not) and phantom decoys, regardless of what options is being targeted by such decoys. This provides the most comprehensive evidence to date of decoy effects in non-human primates, with important theoretical and methodological implications for future comparative studies on context effects in decision-making.

context effects

attraction effect

decoys

phantom decoy effect

capuchin monkeys

comparative decision-making

Mainstream economic theories of rational decision-making assume that preferences are consistent and context-independent. Consistency implies that the relative preferences among two options should not be reversed if a third alternative is added to the choice set. Similarly, context independence imposes that the evaluation of an option should not be affected by the presence of an inferior, irrelevant, or unobtainable alternative (for a review, see Rieskamp et al., 2006). Whatever their merits as a normative model of rational choice, these economic axioms failed the empirical test. Decision makers' preferences are sensitive to the context and change according to the quantity and quality of the available alternatives (Lichtenstein & Slovic, 2006). Context effects prove that manipulating the choice set by adding irrelevant alternatives shifts the preference order and changes the attribution of subjective values (Pettibone & Wedell, 2000; Prelec et al., 1997).

One of the most studied context effects is the attraction effect (AE; Huber et al., 1982; also known as the decoy effect or asymmetric dominance effect). It shows that adding to a binary choice set a new alternative (decoy), that is built to be clearly inferior to one of the pre-existing options, increases the relative choice share of the dominating alternative (target), at the expense of the other option (competitor; Huber et al., 1982). For example, the probability of choosing a print-and-web subscription to a journal that costs 125$ compared to the online subscription at 59$ is higher when the choice set also includes the only print subscription at 125$, a clearly irrelevant and inferior option (Ariely, 2008). To elicit AE, alternatives are usually described on at least two attribute dimensions (i.e., quality and price), thus eliciting so-called “multiattribute decision making”; moreover, the target alternative must clearly dominate the decoy option, which in turn must not be clearly dominated by the competitor option. Most of the previous literature employed decoys built to be equally rewarding on one dimension (i.e., quality) but evidently disadvantageous on the other attribute (i.e., price) compared to the dominant target option (“asymmetrically dominated decoys” - AD; Lichters et al. 2015).

To date, context effects have been repeatedly observed in humans, both in laboratory studies and real-life scenarios, across various domains. To mention a few: political elections (Herne, 1997), risky and intertemporal decisions (Marini & Paglieri, 2019), consumer choices (Huber et al., 1982), medical and legal judgments (Schwartz & Chapman, 1999; Kelman et al., 1996), dating preferences (Ariely, 2008), and many others. Importantly, in recent years, context effects have also been demonstrated in merely perceptual tasks (Trueblood et al., 2013; Liao et al., 2021), where decisions require a simple perceptual discrimination (i.e., choosing between stimuli based on their physical characteristics, namely width or height), in contrast with preferential-based decisions, that are instead managed by high-level cognitive processes (Busemeyer et al., 2019). Since perceptual choice tasks entail quick and intuitive decisions that do not need deliberation or justifications, it is reasonable to assume that context effects would have emerged early in ontogenetic and phylogenetic development (Zhen & Yu, 2016). Integrated models of the role of context effects on, respectively, perceptual and value-based choices have also been proposed: for instance, a recent process tracing study suggests that context effects may, at first, be elicited by perceptual cues of the choice set (i.e., attribute salience), before being strengthened by higher cognitive operations (such as value integration and attribute-wise comparative processes; Marini et al., 2020; 2022).

Interestingly, another, markedly different, type of decoys can also affect the decision-making process: unavailable (phantom) decoys (PDs; Pettibone & Wedell, 2007). A phantom decoy is an option inserted into the choice set that can be compared with the others and evaluated against them, yet it remains unavailable for actual selection. Contrary to AD decoys, PD alternatives dominate the target option. Despite its unavailability, the unobtainable option increases preferences for the dominated target. This means that the asymmetrically dominated option A gains a significant share when a superior unavailable option P is added to the choice set (Pettibone & Wedell, 2007). Phantom superior unobtainable decoys, firstly introduced by Pratkanis & Farquhar (1992) in opposition to real decoys (available inferior options), have been subsequently divided into known and unknown phantoms (Scarpi & Pizzi, 2013). Known PDs are clearly labeled as unavailable (i.e., sold-out products) from the stimulus onset, while unknown PDs seem to be authentic available options until the decision maker tries to select them. Whereas known PDs usually strengthen target preferences, unknown PDs trigger a reactance process that leads to higher competitor selections (Scarpi & Pizzi, 2013).

Several explanations of these context effects have been proposed. With respect to ADs, available theories include weight-based explanations (Huber et al., 1982), prospect theory (Kahneman, 1979), high-level cognitive explanations (Simonson, 1989; Hedgcock & Rao, 2009), perceptual explanations (Dimara et al., 2017), and salience-driven processes (Bordalo et al., 2013). As regards PDs, the two main theories that have been used to explain its elicitation are the relative advantage model (Tversky & Simonson, 1993) and the similarity-substitution hypothesis (Pettibone & Wedell, 2000). Tversky and Simonson (1993) suggested that the phantom option serves as the reference point during decision-making, with only the relative advantages and disadvantages of the alternatives being considered. In contrast, Pettibone and Wedell (2000) suggest that decision-makers, using a heuristic strategy, tend to select the option most similar to the unavailable one without reordering their preferences. However, given the pervasiveness of context effects in our everyday decisions, scholars recently shifted their attention towards models aimed at accounting for the underlying cognitive mechanisms responsible for multiple decoy effects – that is, capable of explaining both ADs and PDs within the same theoretical framework. For example, sequential sampling models (SSMs), such as the drift-diffusion model (Krajbich & Rangel, 2011), the linear ballistic accumulator model (Trueblood et al., 2014), and the multialternative decision field theory (Roe et al., 2001), assume that the decision maker accumulates evidence for the alternatives using options' attributes as input of a comparative process. These comparative processes affect the allocation of subjective values and are translated into choice once a certain threshold is reached (Noguchi & Stewart, 2014). Indeed, the most plausible reason a non-chosen irrelevant alternative can affect a decision maker's preference order is that the preferred alternative is previously compared with other options (Choplin & Hummel, 2005). From a theoretical point of view, SSMs explain the elicitation of context effects both in value-based and perceptual choices (Busemeyer et al., 2019) and recently showed a good fit to empirical data (see Turner et al., 2018 for a comparison of the most influential models).

As mentioned above, recent findings on the merely perceptual nature of context effects suggest that they could be rooted in our evolutionary history and may have played an adaptive role in shaping our decisional mechanisms. Indeed, context effects appear to be widespread in non-human animals (Parrish et al., 2015), possibly pointing at an interspecific use of some comparative evaluation mechanisms, which results in systematic violations of basic rationality axioms. ADs have been reported to affect choice in many non-human species, such as slime molds (Latty & Beekman, 2011), starlings and hummingbirds (Bateson et al., 2002; Morgan et al., 2012), honeybees (Shafir et al., 2002) and dogs (Jackson & Roberts, 2021). As regards PDs, there is only preliminary evidence supporting their effectiveness in shifting preferences in non-human animals. Tan and colleagues (2015) proved that empty food feeders affected honeybees' choices. Scarpi (2011) reported a preference change in cats choosing between different food options in which one alternative was unreachable. Lea & Ryan (2015) documented an influence of an unavailable decoy in mate choices of túngara frogs. In all three studies, adding a phantom alternative increased the selection of the most similar dominated alternative.

Regarding primates, existing evidence on ADs is more controversial and inconclusive. A first perceptual task by Parrish and colleagues (2015) found an attraction effect in macaques choosing among geometric figures. However, the following value-based study with the same species failed to replicate the effect (Parrish et al., 2018). Similar negative results have been reported in comparable studies on other primate species, such as capuchins monkeys (Cohen & Santos, 2017) or great apes (Sanchez-Amaro et al., 2019), choosing between different food options. Despite this corpus of negative evidence, two recent studies on capuchin monkeys made a methodological improvement directly measuring monkeys' preferences at baseline levels, which lead to observe significant preference shifts towards the target option using ADs (Watzek & Brosnan, 2020; Marini et al., 2023). Human literature recently clarified that the context effect is muted when decision makers already have a strong preference among alternatives, prior to the introduction of the AD decoy (Huber et al., 2014; Farmer et al., 2016; Gaudeul & Crosetto, 2019). In Watzek and Brosnan’s (2020) and Marini and colleagues (2023) studies, ternary choice sets that included a decoy were built after assessing the order of preference for the two foods involved and estimating the indifference point between the less favorite and the most favorite food for each animal. These studies represented the first observations of the decoy effect in primates in a value-based task.

Moreover, in Marini and colleagues (2023) study, capuchin monkeys seemed to be sensitive to the distance between the decoy and the target in the attribute space. More specifically, building the dominated decoys by halving the target amount and employing a fixed amount of the favorite food (2 units) and different amounts of the non-favorite food (up to 8 units) led Marini and colleagues to employ decoys with a different degree of similarity with the target options. In doing so, they reported a violation of context independence in two opposite directions. Capuchin monkeys exhibited a strong attraction effect when the decoy was very similar to the target (in the favorite food condition) and a consistent repulsion effect (i.e., an increase in competitor preferences) when the irrelevant option was farther from the dominant alternative (in the less favorite food condition). This result is not new in the human literature (Spektor et al., 2018; Evans et al., 2021), and it is coherent with a recent study that examined the impact of the distance of the decoy in the attribute space (Liao et al., 2021). Unfortunately, Marini and colleagues’ study (2023) does not allow to discriminate whether the elicitation of the repulsion effect was due either to the different distance of the decoy from the target alternative, or to the type of food (more or less preferred) being targeted.

Whereas evidence of Ads in non-human primates is mixed, it appears to be entirely non-existent with respect to PDs: to the best of our knowledge, no study has yet investigated the phantom decoy effect in non-human primates. Thus, the present study intends to fill this gap by exploring the elicitation of context effects in capuchin monkeys, using an experimental protocol in which capuchins were presented with choices including, across different trials, both Ads and PDs.

As in previous studies, we first estimated capuchins' preferences in binary choices, to later administer ternary trials in which target and competitor options were near the subjective indifference points. We had three different decoy conditions, depending on the type of decoy being included: (a) a real asymmetric decoy dominated by the favorite or the non-favorite food option to verify if the elicitation of the attraction effect is sensitive to the food type (Marini et al., 2023); (b) an unavailable (phantom) version of the same asymmetric decoy dominated by the favorite or the non-favorite food, to assess the influence of irrelevant unavailable decoys on capuchins' preferences; (c) a standard phantom decoy that dominated the favorite or the non-favorite food option, to investigate the influence of potentially relevant yet unavailable options on subjects’ preferences.

This paradigm was developed to test the following hypotheses:

H ₁) We hypothesized that adding an asymmetrically dominated decoy would have affected capuchins’ preferences, increasing target selections regardless of the decoy direction (favorite or non-favorite food). In short, we expected the elicitation of the attraction effect in both real AD conditions. We supposed that previous findings on the repulsion effect were, therefore, due to an asymmetric location of the decoy in the attribute space (Liao et al., 2021; Marini et al., 2023).

H ₂) Since previous literature confirmed that non-human animals are able to recognize the dominance relationship within the choice set (target-decoy), choosing decoy alternatives in a negligible number of trials, we assumed that making the dominated alternative unavailable would not interfere with AE elicitation. Coherently with SSMs, we hypothesized that also the PD would be able to play its comparative role and thus increase target selections, despite its unavailability.

H ₃) Lastly, regarding the PD condition, we assumed an impact of the unavailable superior option in both the decoy directions. Since the only way for the unavailable option (PD) to affect capuchins' preferences is to be compared with other options in the choice set, this result would represent further evidence supporting SSMs and comparative models.

2.1 Ethical note

The study protocol was approved by the Italian Ministry of Health (337/2017-PR to EA) and fully complied with the Directive 2010/63/EU on the protection of animals used for scientific purposes.

2.2 Subjects

. A group of twelve capuchin monkeys (six females and six males) with an average age of 24.6 years (ranging from 12 to 36 years) was selected for the purposes of the present study. These monkeys were housed in four different groups at the Primate Center of the Institute of Cognitive Sciences and Technologies of the National Research Council of Italy (ISTC-CNR) located in Rome. The capuchin groups were accommodated in compartments that provided both indoor and outdoor spaces. The size of the outdoor compartments varied depending on the group size, ranging from 65.4 m³ to 139.5 m³. The two indoor compartments measured a total of 25.4 m³. Three experimental compartments (1 m³ in total) were attached to one of the two indoor compartments. Indoor-outdoor compartments were equipped with various enriching materials. Animals were never food deprived for testing, and they received their main meal (a variety of seasonal fruit and vegetables, along with primate monkey chow -- Altromin-A, A. Rieper S.p.A., Vandoies, Italy) in the afternoon after testing was completed. Water was continuously available.

2.3 Experimental apparatus

The apparatus was located on a 70.5 x 50.5 cm table (height: 87.5 cm) placed in front of the experimental compartment. In the Preliminary Food Preference phase two food options were presented by means of a Plexiglas tray (27 × 40 cm); each food option was positioned at 5.2 cm from the frontal edge of the tray and at 8 cm from a central panel that divided the tray into two equally sized portions. In all other phases, the options were presented by means of a wooden tray (43 × 50 cm) on which a vertical frame surrounding a plastic panel (50 x 40 cm) was inserted (see Fig. 1). In the Decoy phase, in “PD Unreachable” and “AD Unreachable” trials (see paragraph – Decoy phase), decoys were clearly visible to the experimental subject but could not be selected (a 7x7 cm transparent panel prevented the subject from reaching for the decoy).

2.4 General procedure

Capuchins were tested in dedicated experimental compartments. The testing sessions took place between 09:30 a.m. and 13:00 p.m. At the beginning of the testing procedure, a subject was isolated from its social group using sliding doors, which allowed the subject to move into the experimental compartment. After the testing session, the subject immediately rejoined the social group. The study involved five experimental phases, namely Preliminary Food Preference, Pretest I, Bound Pretest, Pretest II, and Decoy Phase, which were presented in a sequential order. During the experiment, capuchins were presented with either binary or ternary choices, and they were required to select only one food option in each trial. In all phases, a first experimenter positioned the food options on the tray and pushed it towards the subject, and a second experimenter scored the data. The first experimenter prevented the subject from seeing the food options by placing a black vertical panel between the tray and the experimental subject. A 10-second intertrial interval was implemented between each trial. During the entire experimental process, video recording was conducted using a Canon Legria HF R806 camera.2.5 Experimental phases

Preliminary Food Preference phase

This phase aimed to identify for each subject a high-preferred familiar food (A-food) and a less-preferred familiar food (B-food), such that the A-food was chosen over the B-food in 70%-80% of the trials in two consecutive sessions. Exclusively for the first session and just once, if the subject preferred one of the two foods in 65%-85% of the trials, the session was repeated rather than immediately discarded. In each session, capuchins performed 20 trials of binary choices between 1 unit of food A and 1 unit of food B. The order of presentation was randomized ensuring a counterbalanced right-left placement of the options. Food was cut into pieces of approximately the same size and weight (Raisins: 0.20 g; Pumpkin seed: 0.08 g; Rice Krispies: 0.04 g; Cheerios: 0.35 g) and each food item was weighed on a digital scale (Gibertini Europe 1700; 0.1 g accuracy). Each option consisted of a single food item. The options were presented by means of a Plexiglas tray (2.3 Experimental apparatus). After the subject entered the experimental compartment and located in its centre, the first experimenter pushed the Plexiglas tray towards the experimental compartment, so that the subject could select one of the options.

Pretest I phase

This phase aimed to make sure that capuchins were able to discriminate between the food quantities involved in the subsequent phases and that quadrupling the amount of the less-preferred food (B) offered to them would be sufficient to let them forsake a smaller amount of the high-preferred food (A). Each session consisted of 18 binary choices between the food pairs 2A vs. 1B or 2A vs. 8B (9 trials for each combination) presented in a pseudo-random order counterbalancing right-left position. The options were presented by means of a wooden tray (2.3 Experimental apparatus). After the subject entered the experimental compartment and located in its centre, the first experimenter pushed the wooden tray towards the experimental compartment. Monkeys moved to the next experimental phase once they demonstrated a consistent pattern of choosing 2A over 1B and 8B over 2A in at least seven out of nine trials. This criterion had to be met in two consecutive sessions to successfully complete this phase.

Bound pretest phase

To assess the upper bound (UB) for each monkey, which represented the minimum quantity of B-food chosen over A-food more than 50% of the time, we conducted five consecutive sessions. Each session involved 20 binary choices between 2 units of A-food and varying amounts of B-food. Within each session, subjects performed five types of trials, repeated four times, that were presented in a pseudo-random order. The right-left presentation of the options was also counterbalanced. The five types of trials included: 2A vs. 1B, 2A vs. 2B, 2A vs. 4B, 2A vs. 6B, and 2A vs. 8B. The food choices were presented using the same wooden of the Pretest I phase. Each subject completed a total of five sessions.

Pretest II phase

This phase aimed to make sure that capuchins correctly perceived the dominance relationship between various types of decoys and their targets (T): for phantom decoys (PD), decoys provided the same food of the target option plus one food unit, and we expected that PD was chosen over T; for asymmetrically dominated decoys (AD), decoys always consisted of the same type of food as their target minus one food unit, and the we expected that T was chosen over AD. Each session consisted of 36 binary choices between these options:

2A vs. 1A (AD targeting high-preferred food)
2A vs. 3A (PD targeting high-preferred food)
UB vs. (U-1) B (AD targeting less-preferred food)
UB vs. (U + 1) B (PD targeting less-preferred food)

In each session, there were 9 trials of each type, presented in a pseudo-random order, counterbalancing right-left presentation. The food options were presented on the same wooden tray used in the Pretest I and in the Bound pretest phases. Pretest II was considered completed once the subject chose the target option in at least seven out of nine trials across two consecutive sessions.

Decoy phase

The food amounts used during the Decoy phase were chosen based on capuchins’ preferences during the Bound pretest phase. In this phase, a 3 x 2 within-subjects design was employed (3 decoy types: phantom decoys, PD; unreachable asymmetrically dominated decoys, ADU; standard asymmetrically dominated decoys, AD; 2 target food types: High Preferred, A-Food; Less Preferred, B-food). Thus, there were six experimental conditions: PD on A, PD on B, ADU on A, ADU on B, AD on A, AD on B. Within each session, subjects were presented with 4 trials of each type, plus 4 baseline binary choices 2A vs. UB (where U is the upper bound for each subject), resulting in a total of 28 trials per session. The list of conditions is provided in Table 1.

Table 1

List of experimental conditions. Baseline (2 units of the preferred food) and UB ( the threshold at which the monkeys consistently displayed a preference for the B-food option over the A-food option in the majority of their choices) were always available. In ternary conditions we added a decoy as either a superior (unavailable) or inferior (available or unavailable) option to the choice set.
Condition			Options
Choice	Decoy type	Target option	Preferred food amount	Less Preferred food amount	Decoy
Binary			2A	UB
Ternary	Phantom	High-preferred food	2A	UB	3A
Ternary	Phantom	Less-preferred food	2A	UB	UB + 1
Ternary	AD Unreacheable	High-preferred food	2A	UB	1A
Ternary	AD Unreacheable	Less-preferred food	2A	UB	UB-1
Ternary	AD Standard	High-preferred food	2A	UB	1A
Ternary	AD Standard	Less-preferred food	2A	UB	UB-1

Trial presentation was pseudo-randomized within each session, ensuring that two consecutive trials were never of the same type. Decoy placement (left, center, right) was also pseudo-randomized within the four trials of the same type in any given session, making sure decoy and target were always next to each other in ternary trials. Decoys were designed as in Pretest II: PD consisted of the same type of food as their target, and in the same amount plus one food item; AD consisted of the same type of food as their target, and in the same amount minus one food item. Both decoys were designed in such a way as to be asymmetrically dominated by (AD) or dominant over (PD) their target, but never their competitor.

In PD and ADU trials, decoys were clearly visible to the subject but could not be selected (a 7x7 cm transparent panel prevented the animal from reaching for the decoy; see 2.3 Experimental apparatus); in AD trials, in contrast, decoys were accessible and could be chosen. Target and competitor were always accessible to the subject, across all trials. The food options were presented using the wooden tray used in previous phases.

2.5 Data Analysis

Only the data for the Decoy phase were statistically analysed by means of the software Stata IC (Version 14). The dependent variable was the choice for the 2A option; To control for individuals’ variability, subjects ID was modeled as a random effect, sex and condition as categorical predictors, and age, session number and trial number as continuous predictors. Conditional fixed-effects logistic regressions with robust standard errors were employed and the significance of interaction effects was tested using the Wald test.

Preliminary Food Preference phase

In this phase, a food pair was chosen for each subject. This food pair was selected ensuring that A-food was preferred over B-food in 65%-85% of the trials in two consecutive sessions. All monkeys showed consistent preferences, with an average of 77% of choices favoring the A-food. The food pairs employed for each animal can be found in Table S1 in the Appendix.

Pretest I phase

All subjects preferred two pieces of the high preferred A-food over one piece of the less preferred B-food and eight pieces of the B-food over two pieces of the A-food. Ten subjects reached the criterion of choosing 2A over 1B and 8B over 2A in at least 7 trials out of 9 in two sessions and two subjects (Quincy and Patè) in three sessions (Appendix, Table S2).

Bound pretest phase

Table 2 reports the results of the Bound pretest phase, which aimed to find for each subject an amount of B-food defined UB (upper bound), such that it was the smallest quantity of B-food that the subject chose in more than 50% of the trials over the A-food.

Subject	Preferred food	Upper Bound	2A vs. UB
Patè	2A	4B	0.20
Penelope	2A	6B	0.25
Peonia	2A	4B	0.25
Pepe	2A	6B	0.30
Quincy	2A	4B	0.40
Roberta	2A	4B	0.30
Robinia	2A	4B	0.45
Robot	2A	4B	0.45
Sandokan	2A	6B	0.20
Saroma	2A	4B	0.00
Totò	2A	4B	0.35
Vispo	2A	4B	0.35

Table 2. Bound pretest phase. For each subject, upper bound (number of B-food units) and proportion of preferences for 2A in the comparisons 2A vs. UB

Pretest II phase

In the Pretest II phase, all subjects successfully chose the target option over the decoy option in at least 7 out of 9 trials. Nine subjects reached the criterion in two sessions, two subjects (Penelope and Pepe) in three sessions and one subject (Quincy) needed five sessions. Preferences are reported in Table S3 in the Appendix.

Decoy phase

Descriptive statistics are reported in Table 3.

There was a significant interaction between condition and session (χ²₆ = 318.81; p < 0.001), as shown in Figure 2. In pro-A conditions (N = 480 trials), the choices for the 2A option significantly increased across sessions in the AD (z = 4.00, coeff. = 0.184, p < 0.001) and ADU conditions (z = 2.15, coeff. = 0.150, p = 0.032), and showed a positive but non-significant trend in the PD condition (z = 1.79, coeff. = 0.108, p = 0.073). In pro-B conditions (N = 480 trials), the choices for the 2A option significantly decreased across sessions in the ADU (z = -2.24, coeff. = -0.113, p = 0.025) and PD conditions (z =-2.37, coeff. =-0.101, p = 0.018), whereas for the AD condition there was a non-significant decrease (z =-1.26, coeff. =-0.057, p = 0.207; see figure 1). There was also a significant interaction between condition and trial (χ²₆= 16.93; p < 0.01), but post-hoc analysis did not show any significant variation in 2A choices over trials within the same session (for complete details of the statistical outputs, please see the Appendix).

As shown in Figure 3, capuchins chose the 2A option more frequently in all pro-A conditions than in the Baseline. After Bonferroni correction, which lowered the alpha level to 0.0042 (alpha = 0.05/12 comparisons), this difference resulted significant for the comparisons ADU vs. Baseline (z = 6.83, coeff. = .961, p < 0.001) and PD vs. Baseline (z = 8.31, coeff. = 1.473, p < 0.001), but not for the comparison AD vs. Baseline (z = 2.75, coeff. = .529, p = 0.006). Conversely, capuchins chose the 2A option significantly less frequently in all pro-B conditions than in the Baseline (AD vs. Baseline: z = -6.08, coeff. = -1.494, p < 0.001; ADU vs. Baseline: z = -6.41, coeff. = -1.631, p < 0.001; PD vs. Baseline: z = 5.95, coeff. = 1.507, p < 0.001; Figure 3). Moreover, in pro-A conditions, capuchins chose the 2A option significantly more frequently in PD than in AD trials (z = 6.00, coeff. = 0.944, p < 0.001) and in PD than in ADU trials (z = 3.07, coeff. = 0.512, p = 0.002), but not in ADU than in AD trials (z = 2.34, coeff. = .431, p = 0.020; Bonferroni corrected alpha = 0.0042). From pairwise comparisons of the pro-B conditions, no significant difference emerged, as can be seen in Figure 3. For complete details of the statistical outputs, please see the Appendix.

Subject	Baseline	PD on A	PD on B	ADU on A	ADU on B	AD on A	AD on B
Subject	2A-UB	2A-1A-UB	2A-UB-1-UB	2A-3A(u)-UB	2A-UB+1(u)-UB	2A-1A(u)-UB	2A-UB-1(u)-UB
Patè	0.12	0.55	0.05	0.45	0.02	0.20	0.15
Penelope	0.82	0.90	0.67	0.87	0.60	0.80	0.47
Peonia	0.25	0.80	0.25	0.52	0.17	0.65	0.22
Pepe	0.82	0.85	0.57	0.90	0.52	0.65	0.65
Quincy	0.50	0.85	0.27	0.70	0.42	0.67	0.27
Roberta	0.62	0.82	0.17	0.75	0.17	0.77	0.35
Robinia	0.72	0.90	0.07	0.97	0.30	0.90	0.12
Robot	0.42	0.80	0.05	0.75	0.02	0.57	0.05
Sandokan	0.87	0.92	0.50	0.90	0.27	0.87	0.45
Saroma	0.67	0.97	0.27	0.80	0.22	0.70	0.17
Totò	0.65	0.85	0.27	0.82	0.15	0.82	0.20
Vispo	0.44	0.80	0.02	0.75	0.10	0.62	0.07

Table 3. Decoy phase. For each experimental subject, proportion of 2A preferences in each experimental condition.

Our results confirm all three hypotheses of this study: asymmetrically dominated decoys were effective in shifting capuchins’ preferences regardless of their target (whether high-preferred or less-preferred food), thus suggesting that previous anomalies in the effectiveness of ADs were due to an asymmetric location of the decoy in the attribute space (Liao et al., 2021; Marini et al., 2023); moreover, ADs remained effective even when they could not be selected by capuchins (ADU conditions), thus proving that their role in the decision-making process responsible for the attraction effect is merely comparative and therefore unaffected by actual availability; finally, also traditional phantom decoys influenced the preferences of capuchin monkeys, once again regardless of their target. The latter result is of particular significance since it provides the first experimental evidence of phantom decoy effects in non-human primates available in the literature.

More generally, the specific design of this study allowed us to demonstrate that making the decoys unavailable does not diminish their impact on capuchins’ preferences: this remains true both for unavailable options that are superior to their target (traditional PDs) and to those that are asymmetrically dominated by it (phantom versions of traditional ADs). Indeed, the influence of unavailable decoys on choice was indistinguishable from that of available decoys in pro-B conditions, and even slightly stronger in pro-A conditions. Such a significant role of unavailable options in decision making is consistent with sequential sampling models (SSMs), according to which decision makers gain information on the alternatives by using the attributes of each option as input in a comparative process. Making an option unavailable does not make its attributes any less accessible as information input for the sake of comparison: in fact, removing decoys from the choice set might even prompt the subject to regard them more as a yarding stick than as a palatable option, therefore enhancing the salience of the relevant dominance relationship with their target (which would explain why phantom effects appear to be no weaker, and are occasionally stronger, than those elicited by available decoys). Further studies will be needed to investigate the specific factors responsible for the effectiveness of unavailable options as decoys: nonetheless, the data collected in this study, and in particular the comparison between AD and ADU conditions, provides a first valuable set of evidence on the relative effectiveness of available and unavailable decoys.

From a methodological standpoint, the symmetry observed for context effects across all conditions, regardless of the type of food being targeted by decoys, vindicate the soundness of the paradigm adopted in this study. In particular, it suggests that three aspects of this methodology should become a gold standard for future research on decoy effects in non-human animals: (i) the indifference point of each animal in binary trials should be estimated (see also Watzek & Brosnan, 2020; Marini et al., 2023), so that their preferences between baseline options are malleable enough to allow observing decoy effects in either direction (Huber et al., 2014; Farmer et al., 2016; Gaudeul & Crosetto, 2019); (ii) decoys must be designed in ways that keep constant their distance from their target in attribute space (Liao et al., 2021), at the same time ensuring that their dominance relationship is asymmetrical, i.e., it applies only to their target and not to the competitor; (iii) baseline trials should be included also within decoy sessions, to guarantee that context effects are not biased by independent shifts in baseline preferences across sessions. Taking care of these methodological constraints lead to rather elaborate experimental protocols, like the one employed in this study. Yet we believe it is the only way of guaranteeing sound results, in order to continue exploring context effects and what they reveal on the comparative nature of decision-making processes across various species.

5.1 Funding

No funding was received for conducting this study.

5.2 Conflict of Interest

The authors have no relevant financial or non-financial interests to disclose.

5.2 Author Contributions

MM, EA, and FP designed the study and wrote the paper; EC and SG conducted the experiment; MM and EA analyzed the data.

5.4. Data availability

The data that support the findings of this study are openly available in OSF at: https://osf.io/pg3zx/?view_only=ac0928fbdb974c56be965405cbb9f964

5.5. Ethical approval

The study protocol was approved by the Italian Ministry of Health (337/2017-PR to EA) and fully complied with the Directive 2010/63/EU on the protection of animals used for scientific purposes.

Ariely, D., & Jones, S. (2008). Predictably irrational (pp. 278-9). New York: Harper Collins.
Bateson, M., Healy, S. D., & Hurly, T. A. (2002). Irrational choices in hummingbird foraging behaviour. Animal Behaviour, 63(3), 587-596.
Bordalo, P., Gennaioli, N., & Shleifer, A. (2013). Salience and consumer choice. Journal of Political Economy, 121(5), 803-843.
Busemeyer, J. R., Gluth, S., Rieskamp, J., & Turner, B. M. (2019). Cognitive and neural bases of multi-attribute, multi-alternative, value-based decisions. Trends in Cognitive Sciences, 23(3), 251-263.
Choplin, J. M., & Hummel, J. E. (2005). Comparison-induced decoy effects. Memory & Cognition, 33, 332-343.
Cohen, P. M., & Santos, L. R. (2017). Capuchins (Cebus apella) fail to show an asymmetric dominance effect. Animal Cognition, 20, 331-345.
Dimara, E., Bailly, G., Bezerianos, A., & Franconeri, S. (2018). Mitigating the attraction effect with visualizations. IEEE Transactions on Visualization and Computer Graphics, 25(1), 850-860.
Evans, N. J., Holmes, W. R., Dasari, A., & Trueblood, J. S. (2021). The impact of presentation order on attraction and repulsion effects in decision-making. Decision, 8(1), 36-54.
Farmer, G. D., Warren, P. A., El‐Deredy, W., & Howes, A. (2017). The effect of expected value on attraction effect preference reversals. Journal of Behavioral Decision Making, 30(4), 785-793.
Gaudeul, A., & Crosetto, P. (2019). Fast then slow: A choice process explanation for the attraction effect. hal-02408719f
Hedgcock, W., & Rao, A. R. (2009). Trade-off aversion as an explanation for the attraction effect: A functional magnetic resonance imaging study. Journal of Marketing Research, 46(1), 1-13.
Herne, K. (1997). Decoy alternatives in policy choices: Asymmetric domination and compromise effects. European Journal of Political Economy, 13(3), 575-589.
Huber, J., Payne, J. W., & Puto, C. (1982). Adding asymmetrically dominated alternatives: Violations of regularity and the similarity hypothesis. Journal of Consumer Research, 9(1), 90-98.
Huber, J., Payne, J. W., & Puto, C. P. (2014). Let's be honest about the attraction effect. Journal of Marketing Research, 51(4), 520-525.
Jackson, S. M., & Roberts, W. A. (2021). Irrational behavior in dogs (Canis lupus familiaris): a violation of independence from irrelevant alternatives. Behavioural Processes, 193, 104512.
Kahneman, D. (1979). Prospect theory: An analysis of decisions under risk. Econometrica, 47, 263-292.
Kelman, M., Rottenstreich, Y., & Tversky, A. (1996). Context-dependence in legal decision making. The Journal of Legal Studies, 25(2), 287-318.
Krajbich, I., & Rangel, A. (2011). Multialternative drift-diffusion model predicts the relationship between visual fixations and choice in value-based decisions. Proceedings of the National Academy of Sciences, 108(33), 13852-13857.
Latty, T., & Beekman, M. (2011). Irrational decision-making in an amoeboid organism: transitivity and context-dependent preferences. Proceedings of the Royal Society B: Biological Sciences, 278(1703), 307-312.
Lea, A. M., & Ryan, M. J. (2015). Irrationality in mate choice revealed by túngara frogs. Science, 349(6251), 964-966.
Liao, J., Chen, Y., Lin, W., & Mo, L. (2021). The influence of distance between decoy and target on context effect: Attraction or repulsion?. Journal of Behavioral Decision Making, 34(3), 432-447.
Lichtenstein, S., & Slovic, P. (Eds.). (2006). The construction of preference. Cambridge University Press.
Lichters, M., Sarstedt, M., & Vogt, B. (2015). On the practical relevance of the attraction effect: A cautionary note and guidelines for context effect experiments. AMS Review, 5(1), 1-19.
Marini, M., & Paglieri, F. (2019). Decoy effects in intertemporal and probabilistic choices the role of time pressure, immediacy, and certainty. Behavioural Processes, 162, 130-141.
Marini, M., Ansani, A., & Paglieri, F. (2020). Attraction comes from many sources: Attentional and comparative processes in decoy effects. Judgment and decision making, 15(5), 704-726.
Marini, M., Boschetti, C., Gastaldi, S., Addessi, E., & Paglieri, F. (2023). Context-effect bias in capuchin monkeys (Sapajus spp.): exploring decoy influences in a value-based food choice task. Animal Cognition, 503-514.
Marini, M., Sapienza, A., & Paglieri, F. (2022). There is more to attraction than meets the eye: Studying decoy‐induced attention allocation without eye tracking. Journal of Behavioral Decision Making. 2023;36:e2299.
Morgan, K. V., Hurly, T. A., Bateson, M., Asher, L., & Healy, S. D. (2012). Context-dependent decisions among options varying in a single dimension. Behavioural Processes, 89(2), 115-120.
Noguchi, T., & Stewart, N. (2014). In the attraction, compromise, and similarity effects, alternatives are repeatedly compared in pairs on single dimensions. Cognition, 132(1):44–56.
Parrish, A. E., Afrifa, E., & Beran, M. J. (2018). Exploring decoy effects on computerized task preferences in rhesus monkeys (Macaca mulatta). Animal Behavior and Cognition, 5(2), 235-253.
Parrish, A. E., Evans, T. A., & Beran, M. J. (2015). Rhesus macaques (Macaca mulatta) exhibit the decoy effect in a perceptual discrimination task. Attention, Perception, & Psychophysics, 77, 1715-1725.
Pettibone, J. C., & Wedell, D. H. (2000). Examining models of nondominated decoy effects across judgment and choice. Organizational Behavior and Human Decision Processes, 81(2), 300-328.
Pettibone, J. C., & Wedell, D. H. (2007). Testing alternative explanations of phantom decoy effects. Journal of Behavioral Decision Making, 20(3), 323-341.
Pratkanis, A. R., & Farquhar, P. H. (1992). A brief history of research on phantom alternatives: Evidence for seven empirical generalizations about phantoms. Basic and Applied Social Psychology, 13(1), 103-122.
Prelec, D., Wernerfelt, B., & Zettelmeyer, F. (1997). The role of inference in context effects: Inferring what you want from what is available. Journal of Consumer Research, 24(1), 118-125.
Rieskamp, J., Busemeyer, J. R., & Mellers, B. A. (2006). Extending the bounds of rationality: Evidence and theories of preferential choice. Journal of Economic Literature, 44(3), 631-661.
Roe, R. M., Busemeyer, J. R., & Townsend, J. T. (2001). Multialternative decision field theory: A dynamic connectionst model of decision making. Psychological Review, 108(2), 370-392.
Sánchez-Amaro, A., Altinok, N., Heintz, C., & Call, J. (2019). Disentangling great apes’ decoy-effect bias in a food choice task. Animal Behavior and Cognition, 6(3), 213–222.
Scarpi, D. (2011). The impact of phantom decoys on choices in cats. Animal Cognition, 14, 127-136.
Scarpi, D., & Pizzi, G. (2013). The impact of phantom decoys on choices and perceptions. Journal of Behavioral Decision Making, 26(5), 451-461.
Schwartz, J. A., & Chapman, G. B. (1999). Are more options always better? The attraction effect in physicians' decisions about medications. Medical Decision Making, 19(3), 315-323.
Shafir, S., Waite, T. A., & Smith, B. H. (2002). Context-dependent violations of rational choice in honeybees (Apis mellifera) and gray jays (Perisoreus canadensis). Behavioral Ecology and Sociobiology, 51, 180-187.
Simonson, I. (1989). Choice based on reasons: The case of attraction and compromise effects. Journal of Consumer Research, 16(2), 158-174.
Spektor, M. S., Kellen, D., & Hotaling, J. M. (2018). When the good looks bad: An experimental exploration of the repulsion effect. Psychological Science, 29(8), 1309-1320.
Tan, K., Dong, S., Liu, X., Chen, W., Wang, Y., Oldroyd, B. P., & Latty, T. (2015). Phantom alternatives influence food preferences in the east-ern honeybee Apis cerana. Journal of Animal Ecology, 84, 509–517.
Trueblood, J. S., Brown, S. D., & Heathcote, A. (2014). The multiattribute linear ballistic accumulator model of context effects in multialternative choice. Psychological Review, 121(2), 179-205.
Trueblood, J. S., Brown, S. D., Heathcote, A., & Busemeyer, J. R. (2013). Not just for consumers: Context effects are fundamental to decision making. Psychological Science, 24(6), 901-908.
Turner, B. M., Schley, D. R., Muller, C., & Tsetsos, K. (2018). Competing theories of multialternative, multiattribute preferential choice. Psychological Review, 125(3), 329.
Tversky, A., & Simonson, I. (1993). Context-dependent preferences. Management Science, 39(10), 1179-1189.
Watzek, J., & Brosnan, S. (2020). Capuchin monkeys (Sapajus [Cebus] apella) are more susceptible to contrast than to decoy and social context effects.
Zhen, S., & Yu, R. (2016). The development of the asymmetrically dominated decoy effect in young children. Scientific Reports, 6(1), 22678.

No competing interests reported.

Appendix.docx

Real and Phantom Decoys in Capuchin Monkey (Sapajus spp.) Decision-Making

Status:

Journal Publication

Version 1

Abstract

Figures

1. Introduction

2. Methods

2.1 Ethical note

2.2 Subjects

2.3 Experimental apparatus

2.4 General procedure

2.5 Data Analysis

3. Results

4. Discussion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1