Shifting the balance between goals and habits: Five failures in experimental habit induction

Abstract

Habits are repetitive behaviors that become ingrained with practice, routine, and repetition. The more we repeat an action, the stronger our habits become. Behavioral and clinical neuroscientists have become increasingly interested in this topic because habits may contribute to aspects of maladaptive human behavior, such as compulsive behavior in psychiatry. Numerous studies have demonstrated that habits can be induced in otherwise healthy rats by simply overtraining stimulus-response behaviors. However, despite growing interest in this topic and its application to psychiatry, a similar body of work in humans is absent. Only a single study has been published in humans that shows the effect of extensive training on habit expression. Here, we report five failed attempts to demonstrate that overtraining instrumental behavior leads to the development of inflexible habits in humans, using variants of four previously published outcome devaluation paradigms. Extensive training did not lead to greater habits in two versions of an avoidance learning task, in an appetitive slips-of-action task, or in two independent attempts to replicate the original demonstration. The finding that these outcome devaluation procedures may be insensitive to duration of stimulus-response training in humans has implications for prior work in psychiatric populations. Specifically, it converges with the suggestion that the failures in outcome devaluation in compulsive individuals reflect dysfunction in goal-directed control, rather than overactive habit learning. We discuss why habits are difficult to experimentally induce in healthy humans, and the implications of this for future research in healthy and disordered populations. (PsycINFO Database Record

Figure 1

Task design for Experiment 1A. Panel A lists the training contingencies, with the location of a white square against a black background signaling which foot pedal needed to be pressed to avoid an electric shock to the wrist on the corresponding side. A square in the middle of the screen signified that no shock would be delivered. This stimulus was included in the study to measure false alarms. Panel B shows an example of a trial on which the discriminative stimulus signals that the left pedal needs to be pressed to avoid a shock to the left wrist. Panel C shows the outcome-devaluation manipulation consisted of plugging an electrode from one of the wrists out of the stimulator, in clear view of the participant: for half of the participants this was the left and for the others the right wrist. Panel D: During the extinction test, the discriminative stimuli were again shown on the screen in random order while the participants had the opportunity to press the foot pedals. During this brief test, the shocks were no longer delivered regardless of behavior (i.e., the test was conducted in extinction). This prevents new learning during the test phase.

Figure 2

Main results of Experiment 1A. Panel A: The brief (1-day) group received two trials of training per stimulus in total, whereas the extended (1-day) group received 33 trials per stimulus. Shown here are the average accurate response percentages for the final two stimulus presentations in both the 1-day brief (black bars) and 1-day extended training groups (gray bars). Panel B: The critical outcome-devaluation test results are shown in the right panel. Error bars represent standard error of the means.

Figure 3

Task design for Experiment 1B. Panel A lists the training contingencies, with two fractal stimuli signaling which foot pedal needed to be pressed in order to avoid an unpleasant noise to the ear on the corresponding side. The S-R assignment was counterbalanced across participants. Panel B shows an example of a trial on which the discriminative stimulus signals that the left pedal needs to be pressed in order to avoid an unpleasant sound to the left ear. Panel C shows the outcome-devaluation manipulation consisted of removing an earpiece from one of the ears, and laying this on the table in front of the participant: for half of the participants this was the left and for the others the right ear. Panel D: During the extinction test, the discriminative stimuli were again shown on the screen in random order while the participants had the opportunity to press the foot pedals. During this brief test, the sounds were no longer delivered.

Figure 4

Main results of Experiment 1B. Panel A shows response accuracy of the three groups (1-day brief: white bars; 1-day extended: gray bars; 2-day: black bars) separately for to-be-devalued and to-remain-valuable outcomes. Panel B: The percentage of responses for valued and devalued outcomes during the outcome-devaluation test is shown separately for the three groups. Error bars depict standard error of the means.

Figure 5

Task Design for Experiment 2. Panel A lists the training contingencies, with four symbols signaling which key needs to be pressed to earn a fictitious coin that was worth points. The stimulus–outcome assignment was counterbalanced across participants; and the pictures used as discriminative stimuli versus outcomes (on the surface of the coins) were reversed for half of the participants. Panel B shows an example of a trial on which the discriminative stimulus signals (the moon symbol) that the left key needs to be pressed to earn a coin with a square symbol. Panel C: The outcome-devaluation manipulation consisted of instructing participants (at the start of each test block) that two coins are still worth points (the still-valuable outcomes), but that the two outcomes with a cross superimposed would now lead to deduction of points from the total score (the devalued outcomes). Across the four test blocks, each outcome was devalued two times. Panel D: During the nominal extinction test, the discriminative stimuli were again shown on the screen in random order, and in rapid succession, while the participants had the opportunity to press the two keys. During this brief test, no outcomes were presented, but participants were instructed that they were still earning (and losing) points, and that their total score would determine their chance to earn cinema tickets at the end of the experiment.

Figure 6

Main results of Experiment 2. The effect of training duration was investigated with a within-participant design. Panel A shows accuracy during the three days of training separately for the extensively trained stimuli (empty dots; 3-day) and for the briefly trained stimuli (filled dots; 1-day). The brief stimuli were each presented 32 times in total during eight block sets, whereas the extensively trained stimuli were each presented 96 times. Panel B: The critical outcome-devaluation test results are shown in the right panel. The black bars represent the 1-day training stimuli, and the gray bars the 3-day training stimuli. Error bars represent standard error of the means.

Figure 7

Task design for Experiment 3A and 3B. Panel A lists the training contingencies, with two fractal stimuli signaling which key needed to be pressed (on a variable interval [VI]-10s schedule) to earn M&Ms and Fritos. The S-R assignment was counterbalanced across participants. Panel B shows an example of a trial on which the discriminative stimulus signals that the left key needs to be pressed to earn M&Ms. Panel C: The outcome-devaluation manipulation consisted of satiating participants on one of the two food rewards: For half of the participants these were the M&Ms and for the other half the Fritos. Panel D: During the extinction test, the three discriminative stimuli (corresponding to rest, devalued and valued blocks) were again shown on the screen in random order while the participants had the opportunity to press the keys. During this 3-min test, responding no longer resulted in the food outcomes.

Figure 8

Main results of Experiment 3A. Panel A: End of training. Responses per second for the to-be-devalued outcome and the to-be-valuable outcome at the end of training (averaged across last three blocks of valued and devalued block-types). Response rates are shown separately for the 1-day training group (black bars) and the 3-day training group (gray bars). Panel B: Manipulation check. Food “wanting” ratings on 0 to 20 Likert scales (“How much do you want to eat this type of food right now?”) at pre- and postdevaluation manipulation. Scores are shown separately for the 1-day training group (black lines) and the 3-day training group (gray lines). Dashed lines represent ratings of the devalued food, and solid lines ratings of the valuable food. Panel C: Response rates during the outcome-devaluation test, shown separately for the brief and extended training groups. Error bars represent standard error of the means.

Figure 9

Main results of Experiment 3B. Panel A: End of training. Responses per second for the to-be-devalued outcome and the to-remain-valuable outcome at the end of training (averaged across last three blocks of valued and devalued block-types). Response rates are shown separately for the 1-day training group (black bars) and the 3-day training group (gray bars). Panel B: Manipulation check. Food ‘wanting’ ratings on 0 to 10 Likert scales (“How much do you want to eat this type of food right now?”) pre- and postdevaluation manipulation. Scores are shown separately for the 1-day training group (black lines) and the 3-day training groups (gray lines). Dashed lines represent ratings of the devalued food, and solid lines ratings of the valuable food. Panel C: Response rates during the outcome-devaluation test, shown separately for the 1-day and 3-day training groups. Error bars represent standard error of the means.