家猫的偏好评估与强化物有效性。 机翻

Abstract

Understanding domestic cat (Felis catus) preferences is important for identifying effective reinforcers that can support training, enrichment, and welfare. We assessed preference across three stimulus categories (food, toys, and social interaction) using free‐operant assessments with four adult cats in familiar home environments. The primary dependent measure was duration of quadrant engagement with concurrently available stimuli. Within‐category assessments evaluated preference among multiple stimuli within each category to identify the most preferred item, and between‐category assessments presented the highest ranked stimuli from each category concurrently to evaluate cross‐category preference. A single‐operant reinforcer assessment evaluated whether the most and least preferred stimuli identified in the between‐category assessment differentially maintained operant responding, measured as a nose‐poke response to access the target stimulus. Results showed that three of four cats responded at higher levels when the most preferred stimulus was available relative to the least preferred stimulus. This study is the first to combine within‐ and between‐category free‐operant preference assessments with a direct evaluation of reinforcer effectiveness in domestic cats using procedures that are feasible for home and applied settings, including shelters. These findings support the utility of systematic preference assessments for identifying candidate reinforcers in domestic cats.

Domestic cats (Felis catus) are among the most common companion animals and are found in a variety of settings including private homes, shelters, and managed care environments (American Pet Products Association, 2025; American Society for the Prevention and Cruelty to Animals, 2016; Miller et al., 2014). In the United States alone, approximately 49 million households own at least one cat (American Pet Products Association, 2025). Despite their widespread presence, cats are often perceived as less trainable and less responsive to human‐directed interaction than other companion animals, particularly dogs (Potter & Mills, 2015; Seksel, 2008; Spotte, 2014). Experimental research, however, has long demonstrated that cats acquire behavior through both respondent and operant conditioning. Thorndike's pioneering puzzle‐box experiments established many of the fundamental principles of instrumental learning using cats (Thorndike, 1898), and subsequent research has demonstrated their ability to acquire a wide range of behaviors under operant and respondent contingencies (e.g., Mayes et al., 2015; Pisa & Argillo, 2009; Witte & Kipke, 2005). Collectively, this body of research suggests that perceived limitations in feline trainability might reflect challenges in identifying effective reinforcers rather than species‐specific constraints on learning.

Preference assessments are widely used to identify stimuli that are likely to function as reinforcers across both human and nonhuman populations. Common assessment formats include paired‐stimulus, multiple‐stimulus (with and without replacement), and free‐operant procedures, each of which differs in how stimuli are presented and how preference is quantified (DeLeon & Iwata, 1996; Fisher et al., 1992; Roane et al., 1998). Preference assessments may also be conducted within categories of stimuli (e.g., identifying the most preferred food item from several food options) or between categories (e.g., comparing the highest ranked food, toys, and social stimuli) to evaluate relative preference across broader classes of reinforcers. In free‐operant preference assessments, multiple stimuli are available concurrently and preference is typically measured by the allocation of time or engagement with each stimulus. Relative to trial‐based methods, free‐operant arrangements can reduce assessment time, minimize repeated stimulus presentation, and provide subjects with continuous opportunities to choose among available stimuli (Hagopian et al., 2004; Roane et al., 1998). However, because all stimuli remain concurrently available, free‐operant assessments might be less sensitive for establishing a complete rank ordering of preference, particularly when prolonged engagement with one stimulus limits opportunities to interact with others (Clay et al., 2021). Despite this limitation, free‐operant procedures have been widely used to evaluate preference in situations in which minimizing response requirements and maintaining naturalistic patterns of interaction are important.

Experimental investigations have demonstrated that cats exhibit systematic patterns of stimulus selection under controlled conditions. For example, Church et al. (1996) examined frequency‐dependent food selection in domestic cats and found that food choices varied as a function of nutritional history and available food distributions. More recently, Vitale Shreve et al. (2017) conducted the first systematic evaluation of domestic cat preferences across multiple enrichment categories using a free‐operant preference assessment arrangement. Adult cats from both pet and shelter populations were exposed to stimuli within four categories: human social interaction, food, toys, and scent. Within‐category assessments identified the most preferred stimulus in each category, which was subsequently compared during a between‐category assessment. Preference was measured as the proportion of time spent in locations associated with each stimulus. Results demonstrated substantial individual variability, although social interaction was most frequently preferred, followed by food. These findings challenged the assumption that cats are generally less socially motivated and highlighted the importance of individualized preference assessments. However, preference assessments alone do not establish whether a stimulus will function as an effective reinforcer. Research with other species has demonstrated that preferred stimuli do not always correspond with stimuli that maintain operant responding (e.g., Vicars et al., 2014), emphasizing the importance of directly evaluating the relation between preference and reinforcer effectiveness. Although formal preference assessments have been conducted with domestic cats (i.e., Vitale Shreve et al., 2017), no published study has directly evaluated whether preferred stimuli identified through these procedures function as effective reinforcers.

The present study evaluated the relation between preference and reinforcer effectiveness in domestic cats by comparing outcomes of a free‐operant preference assessment with performance during reinforcer assessments. Preference was first evaluated using procedures modeled after those described by Vitale Shreve et al. (2017) to identify each cat's most and least preferred stimuli. These stimuli were then evaluated as consequences for operant responding to determine whether highly preferred stimuli functioned as more effective reinforcers than less preferred stimuli. We predicted that stimuli identified as highly preferred would maintain higher levels of responding than less preferred stimuli, demonstrating alignment between preference assessment outcomes and reinforcer effectiveness. Support for this prediction would provide empirical evidence for using systematic preference assessments to identify individualized reinforcers for domestic cats, with potential applications for training, enrichment, husbandry, and behavioral interventions.

METHOD

Subjects and setting

Four healthy domestic cats participated: three females (Lana, Miracle, Zarf) and one male (Sterling), ranging in age from 10 months to 6 years. Cats were recruited through a convenience sample using word‐of‐mouth recruitment among researchers' personal and professional networks. Inclusion criteria required cats to be at least 10 months of age and in good health, with no known medical or behavioral concerns. All cats lived in private homes, and caregivers reviewed the study protocol before providing consent for participation. Cats were assessed in familiar indoor environments (e.g., living rooms or other open household spaces). Sessions were conducted in areas providing a minimum unobstructed space of approximately 2.13 × 2.13 m. Furniture was not systematically rearranged for experimental purposes, although caregivers could adjust movable items as needed. Temporary boundaries (e.g., closed doors or lightweight barriers) were used to maintain the designated session area. To minimize the effects of satiation on food preference assessments and prior exposure on toy and social interaction preference assessments, access to target stimuli was restricted for approximately 2.5 hr prior to the first session of each day. Specifically, food and toys were withheld, whereas opportunities for social interaction were minimized as much as possible, consistent with procedures described by Vitale Shreve et al. (2017). Sessions were conducted over 4 to 6 days depending on cat, caregiver, and researcher availability. When multiple sessions were conducted on the same day, sessions were separated by a minimum of 5 min.

Materials and stimuli

Three stimulus categories were evaluated: food, toys, and social interaction. These categories were adapted from the preference assessment procedures described by Vitale Shreve et al. (2017), with the exception that scent stimuli were omitted. Table 1 lists the individualized stimuli evaluated for each cat.

Food and toy stimuli were individualized for each cat in consultation with the caregiver to ensure that all stimuli represented familiar items available in the cats' home environments. Three food stimuli and three toy stimuli were identified for each cat based on caregiver report. Food stimuli were presented using three identical custom‐constructed puzzle boxes, each consisting of a paperboard enclosure (7.6 × 50.8 × 21.6 cm [H × L × W]) with a clear glass (Pyrex) bowl securely attached to the top and bottom surfaces of the enclosure using a rope and bowl collar. Each puzzle box contained one food stimulus and was placed in the quadrant assigned to that stimulus. The apparatus was modeled after the puzzle box used by Vitale Shreve et al. (2017), which was adapted from the apparatus described by Miklósi et al. (2005). The puzzle boxes prevented consumption while allowing cats visual, olfactory, and limited tactile access to the food during preference assessments. The same individualized stimuli were used throughout the preference and reinforcer assessments for each cat.

The social interaction category consisted of three caregiver‐delivered interaction conditions: vocal interaction, petting, and joint play. Joint play involved the caregiver manipulating a toy (e.g., laser pointer) to engage the cat and was categorized as social interaction because access to the toy was mediated entirely through the caregiver, consistent with the procedures described by Vitale Shreve et al. (2017). During all social interaction conditions, caregivers were instructed to interact naturally while maintaining consistency within each interaction type across sessions (e.g., during vocalization conditions, caregivers could say the cat's name in any manner they typically used when calling the cat).

Testing environment and recording

Sessions were conducted in designated areas of the cats' homes that provided sufficient space for stimulus presentation and movement. The testing area was marked with painter's tape to form a square grid (approximately 1.68 × 1.68 m) divided into four quadrants with a central origin point. Three stimuli were assigned to separate quadrants during each session; the fourth quadrant (“No‐Stim”) contained no programmed stimuli and was included as a nonstimulus comparison area during data collection and analysis. This arrangement was adapted from Vitale Shreve et al. (2017). Sessions were video recorded using a mobile device and later scored using VLC Media Player (Version 3.0.20) and paper data sheets.

Dependent measure and operational definition

The primary dependent measure during the preference assessments was quadrant engagement duration, defined as the cumulative amount of time a cat remained within a quadrant. Entry into a quadrant was scored when both front paws crossed the quadrant boundary and remained within the area for at least 2 s. Exit was scored when both front paws exited the quadrant or the cat left the grid area for at least 2 s.

Experimental design

A within‐subject repeated‐measures design was used. Each cat completed within‐category preference assessments, followed by a between‐category preference assessment and a single‐operant reinforcer assessment.

Procedure

Within‐category preference assessments

Within‐category preference assessments were conducted separately for the food, toy, and social interaction categories to identify the most preferred stimulus within each category. Each assessment evaluated three individualized stimuli from a single category using procedures adapted from Vitale Shreve et al. (2017). During each session, stimuli were assigned to separate quadrants of the assessment area. The location assigned to each stimulus was counterbalanced across sessions and cats to minimize potential position bias. Sessions lasted 3 min, and quadrant engagement duration was recorded for each stimulus. Following completion of the assessments, stimuli were ranked within each category according to total engagement duration across sessions. The stimulus associated with the greatest total duration was identified as the most preferred stimulus for that category.

For food and toy assessments, all three stimuli were available concurrently throughout the 3‐min session. During food assessments, three identical puzzle boxes were placed in the assigned quadrants, with each puzzle box containing one of the three food stimuli. During toy assessments, one toy stimulus was placed in each assigned quadrant. Cats were free to move throughout the assessment area and interact with any of the available stimuli.

Social interaction assessments were conducted for 3 min using three sequential 1‐min intervals, with each interval consisting of a different social interaction, consistent with the procedures described by Vitale Shreve et al. (2017). The cat's primary caregiver served as the social partner during these sessions. At the beginning of each 1‐min interval, the caregiver moved to the quadrant assigned to the corresponding social interaction stimulus and delivered the programmed social interaction continuously throughout the interval. At the conclusion of each interval, the caregiver moved to the quadrant assigned to the next stimulus and presented the subsequent interaction. Social interaction stimuli consisted of caregiver‐delivered vocal interaction, petting, and joint play, with joint play involving caregiver manipulation of an interactive toy to engage the cat.

Within‐category preference assessments continued until a stable preference was identified or a maximum of six sessions had been completed. A stable preference was defined as the same stimulus being identified as the most preferred across two consecutive sessions. The highest ranked stimulus from each category was subsequently included in the between‐category preference assessment.

Between‐category preference assessments

Following completion of the within‐category preference assessments, the highest ranked stimulus identified within each category (food, toy, and social interaction) was included in a between‐category preference assessment. The purpose of this assessment was to identify each cat's most and least preferred stimuli across categories. Procedures were identical to those described for the within‐category preference assessments, except that only the highest ranked stimulus identified within each category was evaluated and the selected social interaction stimulus was available throughout the full 3‐min session. Preference across categories was determined according to total quadrant engagement duration across sessions. The stimulus associated with the greatest total duration was identified as the most preferred, whereas the stimulus associated with the lowest total duration was identified as the least preferred. The most and least preferred stimuli were subsequently evaluated in the single‐operant reinforcer assessment.

Single‐operant reinforcer assessment

A single‐operant reinforcer assessment was conducted to evaluate whether the most preferred (MP) and least preferred (LP) stimuli identified in the between‐category preference assessments differentially maintained responding. Each cat completed sessions in which a nose‐poke response produced access to either the MP or LP stimulus. Sessions were conducted in the same familiar home environments used for the preference assessments but without the quadrant arrangement or stimulus locations used during those assessments. Each cat experienced an equal number of MP and LP sessions, and the order of MP and LP sessions was randomized while ensuring no more than two consecutive sessions occurred under the same reinforcement condition. The initial reinforcement condition was counterbalanced across cats.

During each session, the caregiver sat on the floor in a familiar location within the testing area. At the beginning of each trial, the caregiver presented either an open hand or closed fist, which served as the response stimulus. Hand signals were assigned to reinforcement conditions and counterbalanced across cats. The same hand signal was used throughout an individual session, and hand signals changed across sessions. A nose‐poke response was defined as contact between any part of the cat's nose and the caregiver's presented hand; contact with whiskers or other body parts did not constitute a response. A nose‐poke response produced 20 s of access to the designated stimulus (i.e., MP or LP), consisting of social interaction, a toy, or a small food item depending on the condition and cat. If a nose‐poke response did not occur within 40 s of hand presentation, the trial terminated without stimulus presentation. Following either a response or trial termination, the caregiver withdrew the hand and a 20‐s intertrial interval followed before the beginning of the next trial. For food conditions, the intertrial interval began following consumption of the food item.

No formal response training, prompting, shaping, or modeling was conducted prior to the reinforcer assessment. Instead, the assessment relied on the cats' emission of the nose‐poke response within a familiar caregiver interaction context. Each session consisted of 10 discrete trials and lasted approximately 11–19 min (M = 14 min). The dependent measure was the occurrence of a nose‐poke response on each trial, scored dichotomously as a response or no response. Trials in which a nose‐poke response did not occur within 40 s were scored as no‐response trials and were included in the denominator when calculating the percentage of trials with a response. Reinforcer effectiveness was evaluated as the percentage of trials in which a nose‐poke response occurred across MP and LP sessions.

Interobserver agreement and procedural fidelity

Independent observers blind to the study's purpose scored approximately 40% of sessions (37 of 91) across all assessment types. For preference assessments, interobserver agreement was calculated on quadrant engagement duration using a second‐by‐second comparison method, dividing the smaller duration by the larger duration and converting to a percentage (M = 92%, range = 80%–100%). For the reinforcer assessment, interobserver agreement was calculated on a trial‐by‐trial basis as a percentage of agreement on response occurrence (M = 100%). Procedural fidelity was assessed using a checklist of essential procedural components, including grid setup, stimulus placement, timing accuracy, and reinforcement delivery. Fidelity averaged 91% (range = 83%–100%).

RESULTS

Figure 1 displays outcomes from the within‐category preference assessments, between‐category preference assessments, and the single‐operant reinforcer assessment for each cat. Table 2 summarizes total engagement durations obtained during the within‐ and between‐category preference assessments. Within‐category preference assessments continued until stable preference was identified or the maximum number of sessions was completed. Sterling completed six social interaction and toy sessions and four food sessions, Lana completed four social interaction and toy sessions and two food sessions, Miracle completed four sessions within each category, and Zarf completed four social interaction and food sessions and two toy sessions.

[图：Individual performance across within‐category, between‐category, and reinforcer assessments. Rows represent assessment c]

The highest ranked stimulus from each within‐category preference assessment was subsequently evaluated during the between‐category preference assessment. Sterling's selected stimuli were chicken (food), feather (toy), and vocalization (social interaction); Lana's were meat (food), tube (toy), and joint play (social interaction); Miracle's were chicken (food), ball (toy), and joint play (social interaction); and Zarf's were chicken (food), string (toy), and joint play (social interaction). Between‐category preference assessments used the same completion criteria as the within‐category assessments. Sterling and Miracle completed three between‐category sessions, whereas Lana and Zarf completed four sessions. Based on total engagement duration during the between‐category preference assessment, food was identified as Sterling's MP stimulus, social interaction was identified as the MP stimulus for Lana and Miracle, and toy was identified as the MP stimulus for Zarf. The LP stimuli identified during the between‐category preference assessment were social interaction for Sterling, food for Lana and Miracle, and social interaction for Zarf.

During the single‐operant reinforcer assessment, responding occurred at higher levels in the MP condition relative to the LP condition for three of four cats. Trials without a nose‐poke response were scored as no‐response trials and were included when calculating the percentage of trials with a response. Sterling completed eight reinforcer assessment sessions, whereas Lana, Miracle, and Zarf each completed six sessions, with equal numbers of MP and LP sessions conducted for each cat. Across cats, responding occurred during an average of 59.2% of MP trials (SE = 7.2%, median = 70.0%, range = 10.0%–90.0%) and an average of 13.1% for LP trials (SE = 3.1%, median = 10.0%, range = 0.0%–30.0%). Sterling responded during 67.5% of MP trials and 2.5% of LP trials. Lana responded during 20.0% of MP trials and 13.3% of LP trials. Miracle responded during 70.0% of MP trials and 20.0% of LP trials. Zarf responded during 79.3% of MP trials and 20.7% of LP trials.

DISCUSSION

The present study extended the work of Vitale Shreve et al. (2017) by evaluating whether stimuli identified through free‐operant preference assessments functioned as effective reinforcers during an operant task. Although previous research has demonstrated that domestic cats exhibit systematic stimulus preferences, no published study has directly evaluated whether preferences identified through structured preference assessments predict subsequent reinforcer effectiveness. The present findings indicate that free‐operant preference assessments successfully identified individualized preferred stimuli and that those stimuli generally maintained higher levels of operant responding than LP stimuli. Specifically, the MP stimulus identified during the between‐category preference assessment maintained higher levels of responding than the LP stimulus for three of four cats. Collectively, these findings extend previous research by demonstrating that preference assessments can identify candidate reinforcers whose effectiveness can subsequently be evaluated using operant procedures.

The relation between preference and reinforcer effectiveness observed in the present study is consistent with previous research demonstrating that preferred stimuli frequently but not invariably function as effective reinforcers (e.g., DeLeon et al., 1997; Vicars et al., 2014). Sterling, Miracle, and Zarf responded more frequently when responding produced access to their MP stimulus than when it produced access to their LP stimulus, whereas Lana responded at similar levels across reinforcement conditions. One possible explanation for this variability is the relative value of the stimuli identified during the preference assessment. For example, Lana's MP and LP stimuli during the between‐category preference assessment might not have differed substantially in reinforcing value, resulting in similar levels of responding across conditions. These findings highlight that preference and reinforcer effectiveness are related but distinct constructs. Preference assessments identify stimuli that are likely to function as reinforcers, whereas reinforcer assessments directly evaluate the extent to which those stimuli maintain operant responding. Consequently, reinforcer assessments provide an important complement to preference assessments by confirming whether identified preferences translate into functional reinforcement effects.

The single‐operant reinforcer assessment provided an important extension of previous preference assessment research by directly evaluating whether stimuli identified as preferred maintained responding under an operant contingency. An important methodological consideration is that the nose‐poke response was not formally trained prior to the reinforcer assessment. Therefore, acquisition of the response and sensitivity to the reinforcement contingency might have occurred concurrently. Previous research evaluating reinforcer effectiveness has often used established responses or procedures in which the target behavior was acquired before reinforcement effects were assessed (e.g., Grant & Warrior, 2019; Saito et al., 2019; Vicars et al., 2014). In the present study, however, the purpose of the reinforcer assessment was not to evaluate response acquisition but rather to compare responding under equivalent response requirements when access to MP and LP stimuli was contingent on responding. All cats emitted the target response without prompting, shaping, or modeling, and responding occurred at higher levels in the MP condition than in the LP condition for three of four cats. Nonetheless, future research should evaluate whether similar patterns are observed when responses are established prior to reinforcer assessment, thereby allowing acquisition and reinforcement effects to be evaluated separately.

Free‐operant assessments offer several advantages for evaluating stimulus preferences in domestic cats because they allow animals to choose among simultaneously available stimuli with minimal experimenter‐imposed response requirements. This arrangement might be particularly useful when evaluating companion animals in naturalistic settings because it permits subjects to allocate their behavior across available options without requiring repeated approaches, selections, or handling. The present study used a free‐operant arrangement based on procedures described by Vitale Shreve et al. (2017), allowing direct comparison with previous research while maintaining feasibility within cats' home environments. However, free‐operant procedures also have limitations. Because all stimuli are available concurrently, engagement with one stimulus might reduce opportunities to contact other stimuli, potentially limiting the precision with which lower ranked stimuli can be differentiated. In contrast, trial‐based procedures such as paired‐stimulus and multiple‐stimulus assessments might provide more controlled comparisons among stimuli and may produce more reliable preference assessment hierarchies in some populations (e.g., Cameron & Mitchell, 2024; Martin et al., 2018; Vicars et al., 2014). Thus, free‐operant assessments should not necessarily be viewed as a replacement for other preference assessment formats but rather as one approach for identifying candidate reinforcers, particularly when naturalistic choice and feasibility are important considerations.

Several limitations should be considered when interpreting the present findings. First, the small sample size limits the generality of the results, and replication with larger and more diverse samples will be necessary to determine the extent to which these findings generalize across domestic cat populations. However, the repeated‐measures design and evaluation of individualized stimuli across multiple sessions allowed for within‐subject comparisons of preference and reinforcer effectiveness. Second, the use of puzzle boxes during food preference assessments introduced a difference in interaction requirements across categories. The puzzle boxes were used to prevent food consumption from terminating access during preference assessment while still allowing cats to contact food‐related stimuli through visual, olfactory, and limited tactile cues. Nevertheless, the additional interaction required to access the food stimuli might have influenced food‐related engagement relative to toy and social interaction stimuli. Future research should evaluate procedures that maintain the advantages of controlled stimulus presentation while minimizing differences in response effort across stimulus categories. Third, although stimulus access was restricted prior to sessions to reduce potential effects of satiation and prior exposure, motivational variables were not directly measured or experimentally manipulated. Future research should evaluate how factors such as deprivation, recent feeding history, and prior stimulus exposure influence preference assessment outcomes and reinforcer effectiveness. Finally, the present study did not evaluate scent stimuli, which were included in the preference assessment procedures of Vitale Shreve et al. (2017), because the current investigation focused on stimulus categories that could be directly evaluated as consequences during reinforcer assessments. Additional research should examine whether other stimulus classes, including scent‐based stimuli, demonstrate similar relations between preference and reinforcer effectiveness.

The present findings also have practical implications for the assessment and treatment of behavior in domestic cats. Identifying effective reinforcers is a fundamental component of behavior‐change programs, and the present results suggest that systematic preference assessments might provide a practical method for identifying individualized stimuli that are likely to maintain responding. Such procedures might assist caregivers, trainers, veterinarians, and animal‐care professionals in selecting reinforcers for teaching husbandry behaviors, facilitating voluntary participation in examinations and grooming, increasing comfort with transportation or confinement (e.g., entering a carrier), and developing individualized enrichment programs. These procedures might also have broader applications in shelters and other managed‐care environments, where identifying effective reinforcers might improve the efficiency of training and behavioral interventions while promoting animal welfare (Ellis et al., 2017; Houser & Vitale, 2022). Importantly, the present findings also suggest that preference assessments should be viewed as a means of identifying candidate reinforcers rather than definitive measures of reinforcer effectiveness. When feasible, directly evaluating whether preferred stimuli maintain responding might provide additional confidence when selecting reinforcers for behavioral interventions.

AUTHOR CONTRIBUTIONS

All authors contributed to the development, completion, and revision of this manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare that there are no conflicts of interest.

ETHICS APPROVAL

All procedures were approved by the Institutional Animal Care and Use Committee.