Choosing Spouses and Houses: Impaired Congruence Between Preference and Choice Following Damage to the Ventromedial Prefrontal Cortex

Mark D Bowren; Katie E Croft; Justin Reber; Daniel Tranel

doi:10.1037/neu0000421

. Author manuscript; available in PMC: 2018 Apr 9.

Published in final edited form as: Neuropsychology. 2018 Mar;32(3):280–303. doi: 10.1037/neu0000421

Choosing Spouses and Houses: Impaired Congruence Between Preference and Choice Following Damage to the Ventromedial Prefrontal Cortex

Mark D Bowren ^1,^*, Katie E Croft ^2,^*, Justin Reber ³, Daniel Tranel ⁴

PMCID: PMC5890956 NIHMSID: NIHMS912710 PMID: 29620403

Abstract

Objective

A well-documented effect of focal ventromedial prefrontal cortex (vmPFC) damage is a deficit in real-world decision-making. An important aspect of this deficit may be a deficiency in “internal consistency” during social decision-making – that is, impaired congruence between expressed preferences versus actual behavioral choices. An example of low internal consistency would be if one expressed the desire to marry someone with impeccable moral character,” yet proceeded to marry someone convicted of multiple felonies. Here, we used a neuropsychological approach to investigate neural correlates of internal consistency in complex decision-making.

Method

Sixteen individuals with focal vmPFC lesions, 16 brain-damage comparison individuals, and 16 normal comparison individuals completed a three-option forced-choice preference task in which choices were made using attribute sets. Participants also completed visual-analogue preference ratings to indicate how much they liked each option, and rated the influence of each attribute on their decision-making. Options were either social (potential spouses) or non-social (potential houses). Internal consistency for a trial was defined as agreement between the choice and the most positively rated option.

Results

A mixed design ANOVA revealed that internal consistency between choices and preferences derived from summed attribute ratings was significantly lower for the vmPFC group relative to comparison participants, but only in the social condition (pη² = .09), 95% CI [.002, .163].

Conclusions

Internal consistency during social decisions may be deficient in patients with vmPFC damage, leading to a discrepancy between preferences and choices. The vmPFC may provide an important neural mechanism for aligning behavioral choices with expressed preferences.

Keywords: vmPFC, decision-making, lesion method

A well-documented effect of damage to the ventromedial prefrontal cortex (vmPFC) is a severe deficit in decision-making that can be notoriously difficult to detect using standard neuropsychological tests, and yet is plainly evident in the real-world behaviors of individuals suffering such lesions (e.g., Anderson, Damasio, Jones, & Tranel, 1991; Floden, Alexander, Kubu, Katz, & Stuss, 2008; Knutson et al., 2015). Several case studies have provided detailed descriptions of the elusive nature of this deficit (Angrilli, Palomba, Cantagallo, Maietti, & Stegagno, 1999; Cato, Delis, Abildskov, & Bigler, 2004; Dimitrov, Phipps, Zahn, & Grafman, 1999; Eslinger & Damasio, 1985; Satish, Streufert, & Eslinger, 1999). Despite sustaining damage to the prefrontal cortex, these patients demonstrated preserved general intellect upon neuropsychological testing. Perhaps even more striking is the fact that these patients often performed within the normative range on tests of executive functions – the very cognitive capacities traditionally associated with the prefrontal cortex (Tranel, Anderson, & Benton, 1994). It became apparent that the salient deficit associated with vmPFC lesions was most clearly manifested in real-world behaviors rather than in the laboratory. For example, patient C. D. (Cato et al., 2004), described as an exceptional student and prodigious military recruit prior to sustaining vmPFC damage, demonstrated a severe decline in social functioning which led to his forcible discharge from the military. Like other patients with focal vmPFC lesions, C. D.’s deficit was most conspicuous in his real-world decision-making.

Clinicians and researchers who studied carefully the behavior of patients with vmPFC damage also noted a stark dissociation between these patients’ ability to reason abstractly about social decisions and their ability to act upon this reasoning in a congruent manner. For example, patient E.V.R. (Eslinger & Damasio, 1985), in response to a series of social judgment problems, demonstrated intact knowledge of appropriate normative responses, but behaved in an incongruent manner that has even been described as “sociopathic” (Eslinger & Damasio, 1985, p. 1737). Building on this finding, subsequent research demonstrated that individuals like E.V.R. could provide well-reasoned, normative responses to questions about complex moral dilemmas in a laboratory setting, but still display social decision-making impairments in analogous everyday situations (Anderson, Bechara, Damasio, Tranel, & Damasio, 1999; Saver & Damasio, 1991).

This striking lack of congruence between these patients’ laboratory-measured social decision-making and real-world behaviors, referred to here as a deficit in “internal consistency” (cf. Henri-Bhargava, Simioni, & Fellows, 2012), has captivated neuropsychological research for decades. However, this deficit has been difficult to measure empirically. One of the few successful demonstrations of this impairment utilized the Iowa Gambling Task (IGT; Bechara, Damasio, Damasio, & Anderson, 1994). In this task, designed to capture the subtle decision-making impairments of patients with vmPFC damage, individuals are asked to make a series of choices from four decks of cards in order to maximize their profit on a loan of virtual currency. With each choice, a reward and a penalty are delivered. Two of the decks are disadvantageous – they consistently bring high rewards, but also bring severe penalties that lead to an overall net loss when choosing from them. The other two decks are advantageous – they lead to consistently smaller rewards, but also lead to much smaller penalties, which render them profitable over time. When individuals with vmPFC lesions completed the IGT, they could often explain the contingencies of all four decks by the end of the task (Bechara, Damasio, & Damasio, 2000). Despite this, they also consistently chose from the two disadvantageous decks, even at the end of the task. This pattern of performance on the IGT may reflect a deficiency in internal consistency during decision-making.

Importantly, however, the IGT was not formulated with the goal of empirically measuring internal consistency, thus limiting the utility of the task for this purpose. In contrast, Henri-Bhargava et al. (2012) designed a two-option forced-choice paradigm specifically intended to measure internal consistency. In this study, stimuli were rank-ordered based on the frequency with which they were chosen across the forced-choice trials. Preference for a stimulus was operationalized as the rank-order of that stimulus – the more often a specific stimulus was chosen across repeated trials, the higher ranked that stimulus was in terms of preference. Internal inconsistency was defined as the number of deviations from preference (i.e., the number of instances in which the higher ranked of the two options was not selected). The investigators found that, relative to comparison individuals, patients with vmPFC lesions demonstrated greater inconsistency. It is noteworthy, however, that preference was measured as an ordinal variable. This approach limited the investigators’ ability to consider the degree to which one stimulus is preferred over others. One can easily envision a scenario, for example, in which the most preferred option is only slightly preferred over the next best option. Poor internal consistency under such circumstances may reflect the difficulty of the decision. A rank-order approach to preference is unable to factor this possibility into an analysis. To our knowledge, no study has investigated internal consistency between a continuous measure of preference and choices among vmPFC patients within a multi-option, forced-choice decision-making context.

Previous research has not included empirical investigations of how various stimulus characteristics might contribute to deficient internal consistency among patients with vmPFC lesions. In particular, although previous studies have often emphasized the social nature of this decision-making deficit, none have empirically tested whether deficient internal consistency among such patients is more strongly associated with decisions involving the integration of social stimuli relative to non-social stimuli. A better understanding of the nature of the stimuli most strongly associated with deficient internal consistency could have wide-ranging implications for patients with vmPFC damage. For example, if such damage is associated with a greater impairment in internal consistency during decisions involving social information, then practitioners may be wise to closely monitor and advise caregivers and patients about how such circumstances could compromise the coherence between these patients’ stated preferences and final choices. A number of important life decisions (e.g., signing a will) often involve social information and would therefore fall under this jurisdiction. A more complete understanding of the stimulus correlates of deficient internal consistency may also carry implications for theories of the cognitive ontology (Poldrack, 2010) of decision-making and its neural correlates. For example, this would add to a growing body of research suggesting the presence of two distinct executive function systems: a dorsal “cognitive” system and a ventral “social-affective” system (Barbey, Koenigs, & Grafman, 2013; Floden et al., 2008; Gläscher et al., 2012; Grossman et al., 2010; Robinson, Calamia, Gläscher, Bruss, & Tranel, 2014; Stuss & Levine, 2002; Tsuchida & Fellows, 2009).

In addition to the social (versus non-social) nature of the stimuli, other aspects of the stimuli may have important influences on internal consistency. One intriguing possibility is that internal inconsistency is attributable to the complexity of the stimuli in a decision space. Although the empirical literature on this topic is sparse, some researchers have still speculated that patients with vmPFC damage excel at neuropsychological testing but flounder in everyday decision-making because real-world decision making is highly complex (Damasio, 1994). A forced-choice moral dilemma, for instance, the type commonly used in experimental approaches (e.g., Koenigs et al., 2007; Moll, de Oliveira-Souza, & Eslinger, 2003; Shenhav & Greene, 2014), dramatically constrains the decision space and possible outcomes; by contrast, even a seemingly simple choice between restaurants can take into account prices, distance, menu items, décor, likelihood of running into familiar people, and so on. In sum, there is some scant evidence that the complexity of the decision space is one reason behind these patients’ deficient congruence between preferences and choices.

To address some of the unanswered questions regarding how various stimulus characteristics might contribute to deficient internal consistency, we designed a multi-option, multi-attribute decision-making paradigm to study internal consistency among patients with focal vmPFC lesions. The “social-ness” of the stimuli involved in the decision was our primary manipulation of interest. The selection of a spouse from a set of potential spouses served as the social decision, whereas the selection of a house served as the non-social comparison decision. For both stimulus types, the process of selecting a single option based on a constellation of attributes requires one to: (1) weigh each individual attribute’s importance, (2) assign positive or negative affective valence to each attribute, (3) integrate the collection of weighted attributes, and (4) use all this information to guide behavior (i.e., in the selection of a single option). Given that the complexity of the decision space could influence internal consistency during decision-making, we added a manipulation of this factor as well. Complexity was operationalized as the number of attributes associated with each option during a decision. Specifically, two levels of complexity were included: four (simple condition) and eight attributes (complex condition).

In accordance with previous descriptions of the decision-making deficit associated with vmPFC lesions, we predicted that patients with such lesions would demonstrate a deficit in internal consistency during decisions involving social stimuli (spouses), but not during decisions involving non-social stimuli (houses). We did not have specific predictions about the effect of complexity on internal consistency. Rather, we included complexity so as to examine how it might interact with other factors, namely the social/non-social nature of the stimuli and participant group.

Methods

Participants

Participants were 16 individuals with lesions involving the ventromedial prefrontal cortex (vmPFC), 16 brain-damage comparison (BDC) individuals with lesions that did not involve the vmPFC, and 16 neurologically normal comparison (NC) individuals. All vmPFC and BDC patients were selected from the Neurological Patient Registry of the University of Iowa’s Division of Behavioral Neurology and Cognitive Neuroscience. Patients in the vmPFC group were recruited on the basis of having brain damage that included the vmPFC whereas BDC participants were recruited on the basis of having brain damage that did not include the vmPFC region. All patient data were collected during the chronic epoch of the lesion. The participants in the NC group were recruited from the Cognitive Neuroscience Registry for Normative Data of the Division of Behavioral Neurology and Cognitive Neuroscience and from local ads. All participants received detailed information describing the experiment and gave informed consent prior to participating in the study. This study was approved by the Institutional Review Board for Human Subjects Research at the University of Iowa.

Demographic data regarding the participants are provided in Table 1 of Appendix A. The three groups were well matched in terms of sex ratio, handedness, age, education, IQ, and for the two patient groups, time since lesion onset (chronicity). All three groups contained identical ratios of males to females (5:3). The frequencies of mixed-, right-, and left-handed individuals were similar across the groups. In terms of lesion laterality, the vmPFC group was comprised largely of bilateral cases, whereas the BDC group consisted of only unilateral cases. There were no statistically significant group differences in average age, F(2,45) = 1.15, p = 0.33, pη² = .02, education, F(2,45) = 0.44, p = 0.65, pη² = .01, or Full Scale Intelligence Quotient, F(2,45) = 0.52, p = 0.60, pη² = .01, across the groups. Furthermore, there were no statistically significant differences between the two patients groups in terms of average chronicity, t(27) = 1.33, p = 0.24, pη² = .06.

The lesion etiologies for the participants in the vmPFC and BDC groups are presented in Table 2. Lesion etiologies among the patients in the vmPFC group included cerebrovascular accidents (n = 8) and benign tumor resections (n = 8). Lesion etiologies among patients in the BDC group consisted of cerebrovascular accidents (n = 12), tumor resections (n = 3), and tissue resection for the treatment of epilepsy (n = 1).

Figure 1 depicts a lesion overlap map for the vmPFC group. This image was generated using Brainvox software (Frank, Damasio, & Grabowski, 1997), a program which utilizes the MAP-3 technique. This procedure reconstructed each patient’s lesion mask onto a common template brain in three-dimensional space. All the individual patient lesion maps were then overlaid onto a single template brain map. A color-coded scheme was applied to the template brain to represent the number of patients with a lesion mask present in each voxel. Neuroanatomical data were based on structural magnetic resonance imaging (MRI) for patients 770, 2391, 3001, 3032, and 3349 whereas computerized tomography (CT) images were utilized for patients 297, 318, 1652, 1815, 1983, 2025, 2352, 2577, and 3383. The neuroanatomical data for patient 3350 were deemed too difficult to trace due to tissue calcification. Additionally, patient 3456 did not have neuroanatomical research data available at the time of this study. However, this patient’s clinical radiological report indicated that there was bilateral involvement of the vmPFC. Therefore, the neuroanatomical data for patient 3350 and patient 3456 were excluded from the lesion overlap image. The behavioral data provided by these two participants were retained for analysis.

Apparatus

All instructions and tasks were constructed and presented using Presentation® software (Version 13.0, www.neurobs.com). All stimuli were presented on a Compaq Presario X1000 laptop with an ATI Mobility Radeon 9200, 15.4-inch color display. The connected keyboard was modified such that three keys were labeled with “A,” “B,” or “C.” The labels were written in red, blue, or green ink for A, B, and C respectively and were equally spaced across the length of keyboard.

Stimuli

Initially, 200 spouse attributes and 200 house attributes were created as the stimuli for this study. Some of the spouse attributes were selected from a pool of person attributes used in a previous study (Todorov & Uleman, 2004). The remainder of the spouse attributes were created by an author (KC) who explicitly excluded any references to sexual characteristics and/or physical attractiveness in an effort to promote the use of personal preference in making a decision rather than normative standards (e.g., most people would want an attractive spouse). House attributes were created by the same author (KC) who utilized several home buying and house building/contracting websites as guides. House attributes that were deemed overly complex or associated with people (e.g., “has noisy next door neighbors”) were excluded.

The superset of 400 initial attributes was tested in a pilot study of 27 neurologically normal adults recruited from the University of Iowa’s Neuroscience Registry for Normative Data of the Division of Behavioral Neurology and Cognitive Neuroscience. This sample of participants rated the attributes in terms of their affective valence and perceived influence in counterbalanced blocks. The affective rating scale was a measure of valence, or how positive or negative a person perceived each attribute/behavior (for example, “How would you feel about a person if they exhibited that behavior or attribute in the real world?”). Participants provided a rating on a 7 point Likert scale from −3 to 3, where −3 represented “Very Negative”, 0 represented “Neutral” and 3 represented “Very Positive.” The influence ratings measured how much influence or weight each attribute might have had on a person’s decision to choose that particular person as a spouse, or to choose that particular house in the real world (e.g., “When choosing a life partner, how much would it influence your decision if they…” or “When deciding which house to choose when buying a new house, how much would it influence your decision if a house…?”). Participants provided a rating on a 5 point Likert scale where 0 represented “Not at all” influential, 2 represented a “Medium” amount of influence, and 4 represented a “A Lot” of influence, thus providing influence ratings. Independent sample t-tests comparing males and females in terms of perceived influence were performed on all spouse and house attributes. Any attribute associated with a statistically significant difference was removed from the stimulus set. Furthermore, we removed attributes when the standard deviation of the affective ratings was greater than 1 (the standard deviation of the vast majority of attributes was less than this value). No such cutoff was applied to the influence ratings, as these values did not vary as much as the affective ratings. This led to the removal of 13 spouse stimuli and 13 house stimuli, leaving a total of 187 attributes in each condition.

The remaining attributes were assigned by an author (KC) to sets of either 4 (simple condition) or 8 (complex condition) based on their mean weighted additive (WADD) score (Dawes & Corrigan, 1974; Gigerenzer & Goldstein, 1996; Payne, Bettman, & Johnson, 1993) from the pilot data. The WADD score for an attribute was the product of the influence rating and the valence of the affective rating. Summing the WADD scores for all the attributes in a set resulted in a single metric representing the net desirability of the associated option. Specifically, we inspected the distribution of these WADD scores across sets to ensure that each trial contained one option with the greatest mean sum of WADD scores relative to the other options. The manual assignment of attributes to sets ensured that no option contained attributes that were directly contradictory (e.g., one house could not have both the attribute “has a significant crack in the foundation” and the attribute “has a solid foundation”). Furthermore, this allowed us to prevent situations in which all the positive attributes of a set were grouped together and all the negative attributes were grouped together. This helped to prevent primacy and recency effects from influencing how participants responded to the forced-choice task (Newell, Wong, Cheung, & Rakow, 2009). Correlations between the affective and influence ratings from the pilot data revealed that the two measures were highly correlated, which is not surprising due to the well-established and intimate relationship between valence and potency. For the social attributes, the Pearson correlation coefficient was r = 0.63 for the positive and r = −0.73 for the negative attributes. For the nonsocial attributes, r = 0.86 for the positive and r = −0.94 for the negative attributes. The valence (positive or negative) of each attribute was determined by the mean affective valence rating from the pilot study. The final attribute sets, developed by an author (KC), are listed in Appendices B1 and B2 along with the corresponding valence and average influence rating.

Procedure

Prior to the start of the experiment, an experimenter gave participants a detailed description of the study procedures and began the informed consent process in accordance with the Declaration of Helsinki. Once participants consented to take part in the study, an experimenter explained the nature and order of the tasks. Participants were first asked to complete a forced-choice preference task, followed intermittently by a visual analogue preference task, and finally an attribute rating task. Instructions were presented to the participants on the computer screen and simultaneously read aloud by an experimenter. All participants were tested by the same experimenter (KC) and underwent the same procedure, using a standard script of instructions. For two patients (1 vmPFC, 1 BDC), the stimuli were read aloud to them due to visual or reading comprehension impairments. In these cases, the experimenter read the stimuli aloud in a neutral tone but did not interact with the participants unless the participants had a question about the task protocol.

Forced-Choice Preference Task

The first portion of the study asked participants to complete a 3-option forced-choice preference task (see Figure 2 for examples). Participants were instructed that they would be presented with information about several groups of hypothetical people or houses and that each one would be described by various positive and negative attributes. They were instructed to read the attributes aloud as they appeared on the screen to ensure that they attended to each attribute. They were then informed that, after the presentation of the attribute sets for all three options, they would be asked to choose one person or house from each group of three. Participants were also explicitly informed prior to beginning the task that each group of options was unique and independent from all other groups (i.e., person/house A is not the same from trial to trial). For person trials, the participants were asked to select the person from each group of three that they would choose as a spouse or life partner. For house trials, they were asked to select the house from the group that they would choose to live in for the rest of their lives. In the interest of time, participants were encouraged to not spend too long making a choice and to go with their gut feeling if they were not sure which option to choose. Participants made their choices by pressing keys on the keyboard labeled with “A,” “B,” or “C.” The choice trials were presented in blocks (one spouse block and one house block) and the presentation order of the blocks was counterbalanced across the participants such that half of the participants saw the house block first and the other half saw the spouse block first.

To begin each trial of the forced-choice preference task, participants were first shown a list of attributes and were told that these attributes were to be associated with option A (i.e., Spouse A or House A). The attributes for option A were presented one at a time in a fixed order across participants. Each attribute was displayed for six seconds and then remained on the screen until all the attributes for option A were displayed on the screen. Using the same procedure, second and third lists of attributes were subsequently associated with options B and C, respectively. Following the presentation of all three attribute lists, a separate screen prompted the participants to select the person they would choose as a spouse or life partner or the house they would choose to live in for the rest of their lives. Lastly, the final decision screen was displayed to the participants. This screen showed all three options and their associated attribute sets at once, side by side, in columns separated by vertical lines. The left-most column contained the attribute set for option A in red text whereas the middle and right-most columns presented the attribute sets for options B and C in blue and green text, respectively. The final decision screen was presented in an effort to ease impression formation and working memory load for participants, particularly among those in the two patient groups. This decision screen remained onscreen until the participant made a choice by using the corresponding key on the keyboard. This format was repeated for a total of 10 house trials and 10 spouse trials. For each block (houses and spouses), a total of 6 trials contained options with 4 attributes per attribute set whereas 4 trials contained options with 8 attributes per attribute set. The former served as the simple condition trials whereas the latter served as the complex condition trials. In each block, trials 1, 3, 5, 7, 8, and 10 were simple trials whereas trials 2, 4, 6, and 9 were complex trials. In an effort to minimize the load on the participants, we included one fewer complex trial than simple trial. Each block was preceded by one practice trial.

Visual Analogue Preference Task

Immediately following each trial of the forced-choice preference task, participants were asked to rate their preference for each option in the preceding trial. They were instructed to indicate their attitude toward each option using the visual analogue scale shown in Figure 3. The distance from the left-most point on the scale to the right-most point was exactly 100 millimeters. The distance in millimeters from the left-most point on the scale to the mark made by the participants served as the visual analogue preference for each option. During this task, the final summary screen appeared five seconds after the end of the previous forced-choice preference trial (i.e., five seconds after the participants pressed a button to select an option) and remained on the screen while the participant completed the visual analogue scales for each person/house. After the completion of the visual analogue ratings, the participants were instructed to press the enter key on the keyboard to proceed to the next trial of the forced-choice preference task.

Attribute-Rating Task

After the completion of all forced-choice preference trials and visual analogue ratings, participants completed the last portion of the study: the attribute rating task. The attributes themselves were the same attributes used in the attribute sets of the forced-choice preference task. The participants were instructed that they would see all of the previously displayed person/house attributes presented individually and that for each one, they were to rate how much influence or importance each attribute had on their decision of which person (or house) to choose as a spouse (or house). To clarify, participants were instructed that this rating was not a measure of how good or bad an attribute was, but instead, how important it was to them, or how much it would factor into their decision of which person/house to choose. All attributes were rated sequentially in two counterbalanced blocks consisting of either all the spouse attributes or all the house attributes. Each block was preceded by three practice trials. Participants were asked to rate a total of 180 spouse attributes and 180 house attributes for a total of 360 attributes. They rated these attributes on a 5-point Likert scale in which 0, 2, and 4 represented “Not at all”, “Medium” amount of influence, and “A Lot” of influence, respectively. Examples of this task are depicted in Figure 4.

Measures

Internal Consistency

In order to calculate internal consistency, we first calculated WADD scores for each attribute to create a single numerical representation that takes both influence and valence into account. WADD models are considered standard in multi-option, multi-attribute decision research (Dawes & Corrigan, 1974; Gigerenzer & Goldstein, 1996; Payne et al., 1993). The WADD scores were calculated by multiplying each attribute’s influence rating by either 1 or −1 (where 1 represents positive valence and −1 represents negative valence). For example, if a participant rated a negative attribute’s influence as 4 (“A Lot” of influence), that attribute’s WADD score was −4.

Because the attributes used in the attribute rating task were the same attributes used in the forced-choice preference task, we were able to sum the WADD scores for any option in any trial in the forced-choice preference task. The resulting scores reflected the influence and valence of the attributes associated with an option. This summary score for an option was termed the sum of WADD scores. As an example, if the attributes for option A in a simple trial were rated as −1, 3, 2, and 1, the sum of WADD scores for that option was 5. The average sum of WADD scores for each option in each trial is presented in Appendices B3 and B4.

To calculate internal consistency, we examined participants’ sums of WADD scores in relation to their raw choices (i.e., their responses to the forced-choice preference task) across trials. To do this, we first compared the sum of WADD scores for choices A, B, and C for each trial. The option with the greatest sum of WADD scores was deemed to be the attribute-derived choice. Internal consistency within a trial was defined as agreement between the attribute-derived choice and the raw choice within a given forced-choice trial. If the sum of WADD scores for two (or all three) choices were equivalent, such a trial was considered internally consistent if the raw choice for that trial agreed with either (or any) of those options. For example, if the sums of WADD scores for options A and B were both 2, and the sum of WADD scores for option C was −1, then an internally consistent raw choice would be defined as a raw choice of either option A or option B. The final metric to represent internal consistency was defined as the number of internally consistent trials divided by the total number of trials. For example, if a participant was internally consistent on five of the ten spouse trials, then that participant’s internal consistency was 0.5 (or 50%) for the spouse trials.

Response Time

Patients with brain damage generally show greater variability in their response times relative to non-brain damaged comparisons. To the extent that this could affect decisional processes among the patient groups, it was important to measure the response time of the participants during the forced-choice preference task. The amount of time, in seconds, that elapsed from the presentation of the final summary screen of a forced-choice preference trial to the selection of an option by the participant was defined as the response time for each trial.

Attribute-Based Choice Difficulty

We might expect internal consistency to decrease for trials in which the option that was deemed the attribute-derived choice narrowly won out over the option with the next greatest sum of WADD scores. For example, if the sum of WADD scores for options A, B, and C were 3, 2, and −1, respectively, we might expect the associated average internal consistency for such a trial to be lower than the average internal consistency for a trial in which the sum of WADD scores for options A, B, and C were 4, −3, and −4. The average WADD score for each option was used in this calculation and for the analysis so that the simple and complex trials fell within the same scale, making them comparable in terms of this measure without changing the relative ranking of the associated options. Without this calculation, complex trials could range between −32 and 32 whereas simple trials could only fall within the range of −16 to 16. Comparing the averages of the WADD scores for the options in a trial restricted the range from to −4 to 4. To create a final continuous measure of this type of choice difficulty, we calculated the distance between the average sum of WADD scores for the attribute-derived choice and the average sum of WADD scores for the option with the next greatest sum of WADD scores. For example, if one trial’s average sums of WADD scores were 3, −3, 2, and −1, then the attribute-based choice difficulty for that trial would be 1 (calculated by subtracting the second highest score, 2, from the highest, 3).

Results

Because the NC and BDC groups did not differ statistically on any of the dependent variables, they were combined into a single comparison group for the purposes of these analyses. The mean response time, attribute-based choice difficulty, and internal consistency for the vmPFC and comparison groups are presented in Table 3. All analyses were performed using SPSS v 24. Follow-up t-tests were conducted using the Dunn-Sidak correction for multiple comparisons. Also, we analyzed participant sex by including it as a covariate in the main analyses. This variable was subsequently removed, as the results of the analyses were not impacted by participant sex. Also, the two groups were matched in terms of sex ratio, and our pilot study had removed any attributes that differed by participant sex.

Preliminary Analyses

It is important to test whether the groups differed in terms of the rate at which they selected each foil in the forced-choice preference task. To this end, we compared the response patterns of the two groups using three independent samples t-tests: (1) the average number of choice A responses, (2) the average number of choice B responses, and (3) the average number of choice C responses. The two groups did not differ in terms of the average number of option A responses, t(46) = .50, p = .62, pη² = .01, or the average number of option C responses, t(46) = .82, p = .15, pη² = .04. There was a statistically significant difference between the groups in terms of the number of option B responses, t(46) = −2.16, p < .05, pη² = .09, such that participants in the vmPFC group (M = 7.94, SD = 1.48) picked option B more often than the comparison group (M = 7.22, SD = .83). However, Levene’s test for the equality of variances for this test was near significance (p = .054), indicating that unequal variance may influence the probability of type II errors for this test. Correcting for this resulted in the t-test dropping to trend-level significance, t(19.87) = −1.80, p = .087, pη² = .14. Overall, there is no strong evidence of a group-related selection bias for a particular foil (A, B, or C).

Furthermore, it is important to inspect whether there were any group differences in terms of the proportion of participants that selected the “best” option across trials (i.e., the option that was set up to have the highest mean WADD score according to the pilot data). Treating this option in each trial as the “correct” option, we observed no group differences in selecting the correct option, F(2, 45) = .98, p = .39, pη² = .02.

Internal Consistency

Internal consistency data were analyzed using a 2 × 2 × 2 mixed-design analysis of variance (ANOVA) with group (vmPFC or comparison) as the between-subjects factor and stimulus type (house or spouse) and complexity (simple or complex) as the within-subjects factors. Because the underlying internal consistency data were in the form of a proportion, and because the underlying data were binary (consistent or inconsistent), the data were arcsine transformed prior to the analysis. Raw data are presented to avoid confusion about the meaning of the tail ends of the distribution. Response time and attribute-based difficulty for the forced-choice task were entered as covariates. There was a statistically significant main effect of stimulus type, F(1, 44) = 4.54, p < .05, pη² = .09. A follow-up test revealed that the average internal consistency was lower for house stimuli (M = .28, SD = .19) than for spouse stimuli (M = .76, SD = .15; p < .001). There were no significant main effects of group, F(1, 44) = 1.31, p = .26, pη² = .03, or complexity, F(1, 44) = 1.74, p = .19, pη² = .04. In support of our main hypothesis, there was a statistically significant interaction between group and stimulus type, F(1, 44) = 4.20, p < .05, pη² = .09. Follow-up tests on this interaction revealed a simple main effect of group when stimulus type was spouse, F(1, 44) = 4.11, p < .05, pη² = .09. Specifically, participants in the vmPFC group (M = .67, SD = .18) demonstrated less internal consistency relative to the comparison group (M = .75, SD = .18), 95% CI [.002, .163]. These data are displayed in Figure 5. No simple main effect was observed when the stimulus type was house, F(1, 44) = .71, p = .40, pη² = .02. There were no significant interactions between group and complexity, F(1, 44) = .36, p = .55, pη² = .01, complexity and stimulus type, F(1, 46) = .07, p = .79, pη² = .002, or group, stimulus type, and complexity, F(1, 44) = .03, p = .86, pη² = .001.

Exploratory Analysis

To complement the central finding from the main analysis, the internal consistency between visual analogue ratings and raw choice data was compared across the vmPFC and comparison groups. This was calculated for the spouse trials in a manner identical to the method described for the calculation of internal consistency between the attribute-derived choices and raw choices. Using this measure, most participants in both the vmPFC and comparison groups were either very high in internal consistency or internally consistent on all trials (i.e., 100%). This was expected given the fact that the visual analogue ratings immediately followed the raw choice data, likely eliciting strong demand characteristics. The distribution of these data resulted in a severe violation of the assumption of normality. As such, we analyzed the data using a non-parametric test.

The two groups of participants were classified as either internally consistent (100% internal consistency) or internally inconsistent (< 100% internal consistency) with respect to the visual analogue ratings and raw choice data. Table 4 presents the number of consistent and inconsistent participants in this analysis by group along with the corresponding percentage of the total sample for each count. The resulting distribution did not satisfy one of the assumptions for a Chi-Square analysis. Each cell in a Chi-Square must have a count of at least five cases, but only two participants from the comparison group were inconsistent. The Fisher’s Exact test correction was applied. This test revealed a significant difference in the distribution of participants, p < .05. The percent of internally consistent and inconsistent participants are compared across groups in Figure 6. The percent of participants was used in lieu of the count because the comparison group contained twice as many participants as the vmPFC group. A greater percent of the participants in the vmPFC group (31.25%) were internally inconsistent on at least one trial compared to the comparison group (6.25%).

Conclusions

Patients with damage to the vmPFC often display poor congruence between their cognitive abilities and real-world behaviors following the onset of their lesions. The current study examined stimulus correlates of this deficit in “internal consistency,” which was operationalized as agreement between stated or inferred preference and choice during a three-option forced-choice preference task. We tested the hypothesis that decisions involving social stimuli would be associated with attenuated internal consistency among patients with focal vmPFC damage. The results demonstrated that persons with focal vmPFC lesions exhibited reduced internal consistency between their preferences and choices relative to comparison individuals, but only for decisions involving the selection of a spouse as opposed to the selection of a house.

The attenuated congruence between preferences and choices among vmPFC patients was observed both when preference was inferred from individual attribute ratings as well as when preference was determined by asking the participants directly about how much they liked each option. There are likely strong demand characteristics at play in the latter scenario – the participants made their choices just prior to indicating how much they preferred each option. It might be expected, then, that most individuals would be internally consistent under these circumstances. Indeed, this is exactly what we observed for the comparison group. In contrast, patients with focal vmPFC lesions were much more likely to be internally inconsistent on at least one such trial. That these patients’ preferences and choices did not more frequently coincide despite being measured within seconds of one another adds important perspective to our primary finding.

In accordance with previous research, the present study suggests that the vmPFC may be involved in maintaining congruence between cognitive abilities and behavior. Furthermore, this study extends upon previous studies in two important ways: (1) it suggests that the nature of the information under deliberation plays an important role in the whether this deficit is manifested during a particular decision, and (2) it demonstrates that this deficit can be measured empirically in a multi-option, multi-attribute decision-making paradigm. Our finding that reduced internal consistency is observed among vmPFC patients during social decisions coincides with, and lends empirical support to, previous research (Eslinger & Damasio, 1985; Saver & Damasio, 1991) recognizing that social, real-world contexts most prominently evoke poor internal consistency among these patients.

However, our findings may appear to differ in some ways from those reported by Henri-Bhargava et al. (2012). Namely, those investigators found some evidence of internal consistency among vmPFC patients for a range of categories, including some potentially “non-social” categories. This discrepancy could be due to methodological differences between studies. One key difference between the present study and Henri-Bhargava et al. (2012) was the manner in which preference for a stimulus was measured. Henri-Bhargava et al. used a rank-order approach whereas the present study used a graded approach. The approach of the present study allowed us to account for how much more one option was preferred to the others (i.e., the attribute-based difficulty). In contrast, the rank order approach employed by Henri-Bhargava et al. did not allow for this. Beyond the measurement of preference, other key differences between these studies include the number of options in the forced-choice task (two options in Henri-Bhargava et al. compared to three options in the present study), and the nature of the non-social stimuli used in each study. With respect to the latter, it is important to note that the patients in the Henri-Bhargava et al. study only demonstrated significantly greater internal inconsistency compared to healthy comparison participants on two categories: “vegetables” and “puppies.” Although “vegetables” may certainly be non-social, “puppies” may certainly be interpreted as a social category. In sum, key methodological differences between the present study and Henri-Bhargava et al. may explain some apparent discrepancies between these two studies.

We did not find evidence that the complexity of the decision space interacted with any other factors. This finding stands in contrast to the notion that deficient congruence between preferences and choices among vmPFC patients during decisions is due to the complexity of those decisions. Our results suggest that the social nature of the decision itself may have been sufficient to impair the congruence between preferences and choices among these patients. Alternatively, it could be that the number of options, rather than the number of attributes per option, is associated with attenuated internal consistency. Indeed, complexity could conceivably be operationalized in a variety of ways that differ from how it was defined in the present study. As such, future studies should seek to clarify the role of the complexity of information in the maintenance of internal consistency.

In addition to our central finding, we found a large and unexpected main effect of the social/non-social manipulation on internal consistency. Specifically, the internal consistency for house trials was well below that of the spouse trials, across both groups and both levels of complexity (see Table 3). This suggests that the house and spouse stimuli were not well matched on overall internal consistency. However, this limitation does not detract from the central finding of this study, as it did not result in ceiling or floor effects for internal consistency in either the house or spouse condition.

It may be the case that this unexpected finding is attributable to differences between the spouse and house stimuli. For example, individuals may have more experience deciding between potential mates than deciding between potential houses. We might expect most people to have more previous dating experiences than house-purchasing experiences. Such disproportionate “practice” for one type of decision over the other might contribute to the discrepant internal consistency between houses and spouses. However, it is notable that practice with selecting a person to date is not equivalent to practice selecting a spouse (a decision which, for many, may only be made once). Another potential difference between the stimuli that could explain our unexpected finding may lie in the decision-making strategies that the participants used during the tasks. It may be the case that participants did not select houses using an additive model (i.e., the WADD model used to determine the most preferred option). Rather, participants may have employed a range of non-additive heuristic strategies. This latter possibility would be interesting given previous research demonstrating that patients with vmPFC lesions employ non-normative heuristics during decision-making (Fellows, 2006). The present study was not designed to tease apart these nuances of the decision-making approach. However, it may be worthwhile for future studies to explore whether the decision-making heuristics utilized by patients with vmPFC lesions are dependent upon the nature of the stimuli under deliberation.

It is important to note that the present study does not suggest that incongruence between preferences and choices is unitarily caused by focal vmPFC lesions. Imperfect internal consistency was also observed among the comparison group in this study – even between the visual analogue ratings and choices (a set of tasks which, as mentioned above, likely promotes internal consistency). That internal inconsistency is, to an extent, normative suggests that vmPFC damage is not the sole cause of incongruence between preferences and actions. Rather, our data only support the idea that vmPFC lesions are associated with more frequent instances of incongruence between preferences and choices than is to be expected among a normative population. Any conclusions drawn from this study about the role of the vmPFC in internal consistency should be interpreted in light of these considerations.

Although most studies of patients with focal and stable vmPFC damage include far fewer participants, the present study may still be limited by a somewhat small sample size. As is the case with other neuropsychological studies, small samples may give rise to concerns about the generalizability of results. Additionally, participants only completed a total of 20 trials. Given the various manipulations that we employed in this study (i.e., social/non-social and simple/complex), it may be the case that an insufficient number of trials were aggregated to properly measure internal consistency. Although this did not affect our ability to observe a significant effect of social stimuli on internal consistency, this limitation may have affected our ability to detect an effect of complexity.

The task utilized in the present study was novel, and thus warrants a discussion of its potential limitations and related alternative explanations for our central finding. Although the task was designed to mimic aspects of real-world decision-making, it was also necessarily simplified into a standardized format that was feasible for laboratory research. As such, the possibility remains that our results may not generalize to real-world situations. Another important limitation of this task involves some potentially important differences between houses and spouses. For example, it might be common to sit down and deliberate rationally over a house but not a spouse. Such unfamiliar circumstances for making decisions about potential spouses may have affected our results. Alternatively, the house and spouse stimuli may differ in the affective information associated with each stimulus type. It may be the case that the affective information contained in the spouse stimuli drives our central finding. Disentangling the contributions of affective and social information to the deficient internal consistency of patients with vmPFC lesions remains an important goal for future research.

A final consideration concerns the nature of the spouse attributes. Some attributes are single behaviors that may or may not be indicative of a person’s character across situations (e.g., “cursed at the flight attendant”). Others are more general statements that apply across situations (e.g., “becomes angry when people express ideas very different from their own”). It may be the case that participants focused on one or two attributes that apply across situations while discounting the others. We have the impression, however, that this possibility is accounted for by the nature of the influence ratings. If participants’ decisions were not influenced by an attribute that described a single behavior, they were able to indicate that in their influence rating (i.e., they could assign that attribute a low influence rating). Indeed, the instructions asked them to approach these ratings in this very way. Alternatively, if an attribute provided information that can apply across many situations, and the participants found that attribute to be influential in their decision, they could assign that attribute a high influence rating. This approach allowed participants to decide whether a single behavior was important to them while also leaving room for the possibility that some people may find some single behaviors to be highly adjudicative with regard to their decision.

Moving forward, it may be worthwhile to develop and validate a laboratory measure of internal consistency similar to the one used in this study. Item response theory (IRT) may hold promise as a method for the development and selection of items for such a measure. IRT could help elucidate the items (i.e., trials) that best distinguish between those with normative levels of internal consistency from those with abnormally low internal consistency. A more refined set of trials could improve the future measurement of internal consistency. A well-validated, standardized, and relatively short instrument to this end would almost certainly find both clinical and research applications.

In conclusion, the real-world decision-making deficit associated with vmPFC damage carries implications for decision-making research and clinical case management. Although widely recognized, this deficit has been difficult to measure empirically and has not yet been fully characterized. The present study sought to both provide a novel paradigm for measuring this deficit in the laboratory as well as elucidate stimulus correlates of this deficit. The results of the present study suggested that decisions involving social information are associated with attenuated congruence between preferences and choices among patients with focal vmPFC lesions. Our work elaborates upon previous research by empirically demonstrating that this deficit can be captured in a multi-option, multi-attribute paradigm, and that the nature of the stimuli under deliberation can influence internal consistency among patients with vmPFC damage. These findings help to refine our understanding of the role of the vmPFC in decision-making.

Public Significance Statement.

This study suggests that persons with vmPFC damage may display poor congruence between their thoughts and actions during social decisions. This raises concerns about their ability to make important life decisions that involve social information (e.g., signing a will).

Acknowledgments

This research was supported by grants from the National Institute of Mental Health [2P50 MH094258] National Institute of Neurological Disorders and Stroke [P50 NS19632], the National Institute on Drug Abuse [R01 DA022549], the McDonnell Foundation [Collaborative Award 220020387], and the Kiwanis Foundation.

Appendix A

Table 1.

Demographic Information and IQ for All Three Groups; Chronicity and Laterality of Lesion for Patient Groups

Group	Sex	Handedness	Age	Education	FSIQ	Chronicity	Laterality
NC^*	10M/6F	1Mi/1L/14R	58.8 (9.2)	15.0 (2.2)	111.8 (8.2)	n/a	n/a
BDC	10M/6F	0Mi/2L/14R	56.6 (9.1)	15.1 (2.8)	111.8 (11.2)	9.2 (6.5)	0B/7R/9L
vmPFC	10M/6F	2Mi/1L/13R	61.2 (7.0)	14.3 (13.1)	108.3 (13.1)	12.9 (9.1)	13B/1R/2L

Subject ID	Group	Laterality	Lesion Etiology
297	vmPFC	L	ACoA aneurysm clipping
318	vmPFC	B	Meningioma resection
770	vmPFC	B	Meningioma resection
1652	vmPFC	L	ACoA aneurysm clipping
1815	vmPFC	B	Meningioma resection
1983	vmPFC	B	ACoA aneurysm clipping
2025	vmPFC	B	ACoA aneurysm clipping
2352	vmPFC	B	SAH; ACoA aneurysm clipping
2391	vmPFC	B	Meningioma resection
2577	vmPFC	B	SAH; ACoA aneurysm clipping
3001	vmPFC	R	Meningioma resection
3032	vmPFC	B	Meningioma resection
3349	vmPFC	B	Meningioma resection
3350	vmPFC	B	Meningioma resection
3383	vmPFC	B	SAH; ACoA aneurysm clipping
3456	vmPFC	B	SAH; ACoA aneurysm clipping
1290	BDC	L	Temporal lobe hemorrhage
1656	BDC	R	Right hemisphere infarct
2107	BDC	R	Parietal/insular infarct
2374	BDC	L	Temporal lobe resection
2905	BDC	R	SAH; Temporal lobe resection
2934	BDC	L	Temporal/occipital infarct
2962	BDC	R	Temporal lobe resection
2995	BDC	L	Parietal lobe meningioma resection
3165	BDC	L	Frontal intraparenchymal hemorrhage and infarct
3195	BDC	L	Parietal lobe hemorrhage; AVM excision
3315	BDC	R	Right hemisphere infarct
3348	BDC	L	Temporal intraparenchymal hemorrhage
3396	BDC	L	Temporal lobe meningioma resection
3428	BDC	L	Left hemisphere infarct
3490	BDC	R	Occipital lobe infarct
3501	BDC	R	Right hemisphere infarct

	vmPFC Group				Comparison Group
	Spouses		Houses		Spouses		Houses
Measure	Simple	Complex	Simple	Complex	Simple	Complex	Simple	Complex
Response Time (sec)	16.28 (4.96)	21.15 (10.80)	14.30 (4.04)	13.85 (6.72)	19.09 (4.98)	22.57 (10.69)	12.26 (4.02)	13.11 (6.68)
Attribute-Based Choice Difficulty	1.03 (.20)	.80 (.24)	.73 (20)	1.11 (.27)	1.04 (.20)	.78 (.24)	.82 (.19)	.97 (.27)
Internal Consistency^*	.66 (.16)	.68 (.20)	.34 (.12)	.30 (.20)	.72 (16)	.78 (.21)	.29 (.12)	.29 (.19)

	Consistent	Inconsistent	Total
vmPFC	11 (22.9%)	5 (10.4%)	16 (33.3%)
Comparison	30 (62.5%)	2 (4.2%)	32 (66.7%)
Total	41 (85.4%)	7 (14.6%)	48 (100%)

	Valence	Mean vmPFC	SD vmPFC	Mean Comparison	Comparison SD
Person A
ate the omelet that had fallen on the floor	Negative	1.56	1.26	1.22	1.10
likes to pick fights with their friends	Negative	2.63	1.02	2.94	0.98
offered to help the woman lift her heavy bags onto the bus	Positive	2.13	1.02	1.84	0.92
took cookies to the orphans during the holidays	Positive	2.19	0.91	1.72	1.05
Person B
wore the same dirty shirt to work for 3 days	Negative	1.88	1.02	1.47	1.14
brought juice and fruit for a sick friend	Positive	2.25	1.29	1.75	1.02
doesn’t give up, even when things look hopeless	Positive	3.13	0.50	2.97	0.93
took their nephew to the playground	Positive	1.63	1.09	1.22	1.04
Person C
gets irritated when people ask favors of them	Negative	2.25	0.93	2.06	0.76
started an argument with the policeman	Negative	2.50	0.89	2.38	0.98
often pretends to be sick in order to get out of obligations	Negative	2.69	1.08	2.66	0.94
is pretty conscious of other’s feelings	Positive	2.63	0.81	2.91	0.73

Person A
continues to drive drunk, even after several accidents	Negative	3.81	0.40	3.88	0.42
are never willing to admit when they’re wrong	Negative	3.06	0.68	3.09	0.89
participated in the walk for breast cancer	Positive	2.44	1.31	1.38	0.91
reads books to an elderly woman every Sunday	Positive	2.13	1.09	1.56	1.05
adopted a stray dog from the pound	Positive	1.63	0.96	1.41	1.10
picked up a friend at the airport	Positive	1.81	1.11	1.28	1.05
helped a friend prepare for a job interview	Positive	2.13	0.89	1.59	1.04
switched shifts with a co-worker when they needed the day off	Positive	2.13	1.02	1.56	1.31
Person B
ripped pages out of the library book	Negative	2.63	1.09	2.16	0.85
talked on their cell phone loudly on a crowded bus	Negative	1.81	1.17	1.44	0.98
always expects the worst to happen	Negative	2.69	0.85	2.25	1.18
refused to congratulate the winner of the game	Negative	2.06	0.50	1.81	0.67
keeps their promises faithfully	Positive	3.63	0.96	3.41	0.93
is open-minded to different views and opinions	Positive	2.56	1.03	3.09	0.78
smiles often	Positive	2.63	0.81	2.41	1.05
is warm and caring with others	Positive	2.88	1.13	2.84	0.98
Person C
is very insecure and always second-guessing themselves	Negative	2.31	0.79	2.28	1.02
made fun of the cripple’s walk	Negative	3.06	1.06	2.94	1.01
sent their father to a nursing home unnecessarily	Negative	3.38	0.81	3.47	0.80
tells racist jokes in public	Negative	3.31	0.79	2.75	1.30
never says please or thank you	Negative	2.50	0.97	2.34	1.10
took credit for their coworker’s work in order to get promoted	Negative	3.25	0.86	3.13	0.94
nursed the homeless and sick kitten back to health	Positive	1.69	1.20	1.69	1.03
returned the money when given the wrong amount of change	Positive	2.69	1.20	2.38	0.98

Person A
insists on having things their way	Negative	2.81	0.66	2.94	0.88
makes loud noises when eating	Negative	1.88	1.31	1.38	0.91
stood up a friend at dinner and never called to explain why	Negative	2.44	0.89	2.44	0.95
held the door open for the boy on crutches	Positive	2.38	1.02	1.81	0.97
Person B
likes to tell lies	Negative	3.44	0.96	3.53	0.67
always turns the lights off when they walk out of the room	Positive	1.75	1.13	0.78	1.16
helped the newcomer set up their office	Positive	2.88	1.09	2.19	1.12
coaches a co-ed soccer team on the weekends	Positive	1.69	0.95	0.88	0.87
Person C
cursed loudly in front of the children	Negative	3.25	0.86	2.28	1.35
slammed the door in the salesman’s face	Negative	1.31	0.95	1.59	1.19
puts forth their best effort, no matter what the outcome may be	Positive	2.94	1.00	3.13	0.75
is kind-hearted	Positive	2.94	0.93	3.16	1.02

Person A
often stares rudely at strangers	Negative	2.13	0.81	1.78	1.21
refused to yield to the emergency vehicle	Negative	2.94	1.06	2.22	1.16
didn’t smoke at home while their roommate was trying to quit	Positive	2.56	1.09	2.41	0.91
gave their seat on the bus to a pregnant woman	Positive	2.25	1.24	2.19	1.06
helped the blind man pick out items in the grocery store	Positive	2.44	1.26	2.13	1.16
returned the lost wallet to the owner	Positive	3.06	0.93	2.63	1.04
tries to be courteous all the time, even to people who are disagreeable	Positive	2.69	1.01	2.66	1.04
pulled over to the side of the road to check on the stranded travelers	Positive	2.50	1.10	2.28	1.02
Person B
jumped to the front of the line without asking	Negative	2.81	0.91	2.50	1.02
loudly interrupted the private conversation	Negative	2.63	0.81	1.94	0.91
offends people by being so blunt	Negative	2.31	1.01	2.41	0.91
pushed the old lady out of their way	Negative	3.38	0.72	2.97	0.90
lost half of their savings to a scam artist	Negative	1.88	1.09	2.16	1.27
always underpays their portion of the bill	Negative	2.44	0.81	2.47	0.92
organized a toy drive for a local charity	Positive	2.13	1.02	1.81	0.97
feels really guilty when they tell lies	Positive	2.44	1.15	2.56	1.19
Person C
does many things that is simply for their own gain and at the expense of others	Negative	2.94	0.93	3.38	0.79
made fun of the boy’s lisp	Negative	3.00	0.97	2.63	1.29
picked dry gum from under the desk, put it in their mouth and chewed it	Negative	2.94	1.44	2.59	1.36
tripped their friend and then laughed at her	Negative	2.88	1.09	2.88	1.18
gives money to charity on a regular basis	Positive	2.31	0.95	1.72	1.08
helped the old lady across the road	Positive	2.13	0.96	1.59	1.01
stayed calm enough to help their friends out of the car after the accident	Positive	2.81	0.98	2.47	1.08
brought food over to the grieving family	Positive	2.50	0.89	2.19	0.90

Person A
ate the soup, even though there was a fly floating in the bowl	Negative	1.38	1.15	1.53	1.32
talked continuously while sitting behind a couple at the movie theater	Negative	2.56	1.15	2.13	1.01
encourages their less talented co-workers	Positive	2.25	0.86	2.13	1.01
bought food and gave it to a homeless person	Positive	1.94	1.06	2.00	1.16
Person B
has to be reminded to pay back money they borrow	Negative	2.31	0.95	2.22	1.01
arranged a birthday party for their little brother	Positive	1.94	1.06	1.19	1.00
likes to help others by listening to their problems	Positive	2.69	1.01	2.38	0.94
comforted a sad friend	Positive	2.44	0.81	1.94	1.08
Person C
ate moldy food out of the refrigerator	Negative	2.13	1.26	1.97	1.28
broke an expensive item in a store and put it back without telling anyone	Negative	2.69	1.14	2.41	1.04
skipped their sister’s wedding to go to Las Vegas with friends	Negative	3.13	0.62	2.44	1.32
tries to practice what they preach	Positive	3.00	0.63	2.84	0.85

Person A
missed their niece’s birthday party because they were too drunk to attend	Negative	3.25	0.86	3.00	1.30
rebels against people in authority even when they know the authorities are right	Negative	2.94	1.06	2.91	0.93
stole merchandise from the store where they worked	Negative	3.63	0.72	3.59	0.71
yelled at the waitress for the 15 minute delay	Negative	2.50	1.21	2.47	0.92
routinely forgets to show up at appointments	Negative	2.50	0.82	2.38	1.10
is really bossy	Negative	2.81	0.91	2.81	1.09
is a patient teacher	Positive	2.88	0.81	2.47	1.08
gave a stranger change when they were short on money at the store	Positive	1.94	1.00	1.97	0.90
Person B
becomes angry when people express ideas very different from their own	Negative	2.63	0.89	3.13	0.87
has been fired from 7 jobs	Negative	3.31	0.87	3.19	0.86
doesn’t like to try new experiences	Negative	1.65	0.62	2.53	1.02
is very critical of their friends	Negative	3.13	0.96	2.88	0.87
tries not to gossip about other people’s business	Positive	2.19	1.17	2.38	0.98
usually shows up on time, if not early	Positive	2.31	1.08	2.47	1.11
volunteered at the local soup kitchen	Positive	2.13	1.09	1.81	0.97
sent their mother flowers on Mother’s Day	Positive	1.88	1.09	1.63	1.10
Person C
threw their trash on the sidewalk and kept walking	Negative	2.19	1.22	1.66	0.90
has chronic bad breath	Negative	1.88	1.15	1.59	1.13
doesn’t hesitate to go out of their way to help someone in trouble	Positive	2.94	0.93	2.84	0.88
helped their little brother with a school project	Positive	2.06	1.18	1.78	1.18
is an organ donor	Positive	2.75	0.93	1.34	1.18
loves to cook for others	Positive	2.19	0.91	1.72	1.11
tries to be a good listener, no matter who they’re talking to	Positive	2.63	1.09	2.72	1.11
lent a friend money when they forgot their wallet	Positive	2.19	0.98	1.94	0.91

Person A
coughs without covering their mouth	Negative	1.81	1.05	1.44	1.01
made fun of the girl’s outfit	Negative	2.38	0.89	1.78	1.10
is very agreeable and easy-going	Positive	3.44	0.63	3.13	0.83
strives to be fair when dealing with other people	Positive	2.75	1.00	2.88	0.91
Person B
burst the pimples on their face in public	Negative	1.81	1.42	2.19	1.09
farted loudly on the airplane	Negative	2.19	1.11	1.63	1.34
is always running late	Negative	2.13	1.09	1.91	1.06
helped a lost person find their destination	Positive	1.88	1.02	1.72	0.99
Person C
often drinks before work	Negative	3.38	0.81	3.47	0.84
appropriately declares everything at customs	Positive	1.63	1.26	1.53	1.08
held open the door for everyone exiting the bus	Positive	2.00	1.10	1.50	1.05
brought a cake to work for a co-worker’s birthday	Positive	1.44	0.96	1.28	0.96

	Mean vmPFC	SD vmPFC	Mean Comparison	SD Comparison
B3.1 Trial 1
Person A	0.13	1.96	−0.59	2.24
Person B	−5.63	1.93	−4.97	1.71
Person C	−4.81	2.01	−4.19	1.96
B3.2 Trial 2
Person A	6.39	4.01	3.23	3.92
Person B	1.38	4.01	2.80	3.92
Person C	−13.46	4.01	−12.86	3.92
B3.3 Trial 3
Person A	−4.75	2.01	−4.94	1.96
Person B	−1.38	2.01	−2.13	1.96
Person C	1.31	2.01	2.41	1.96
B3.4 Trial 4
Person A	0.82	4.01	1.10	3.92
Person B	−10.89	4.01	−10.09	3.92
Person C	−6.64	4.01	−6.95	3.92
B3.5 Trial 5
Person A	−4.25	2.01	−3.78	1.96
Person B	0.88	2.01	0.91	1.96
Person C	−4.94	2.01	−3.97	1.96
B3.6 Trial 6
Person A	−12.83	4.01	−12.74	3.92
Person B	−2.45	4.01	−3.61	3.92
Person C	4.84	4.01	3.42	3.92
B3.7 Trial 7
Person A	2.00	2.01	2.78	1.96
Person B	−4.25	2.01	−4.00	1.96
Person C	−1.56	2.01	−2.22	1.96
B3.8 Trial 8
Person A	2.13	2.01	1.13	1.96
Person B	−5.31	2.01	−4.63	1.96
Person C	1.88	2.01	2.50	1.96
B3.9 Trial 9
Person A	0.01	4.01	3.14	3.92
Person B	5.27	4.01	2.86	3.92
Person C	−20.64	4.01	−20.97	3.92
B3.10 Trial 10
Person A	−10.31	2.01	−10.13	1.96
Person B	4.75	2.01	3.78	1.96
Person C	−1.31	2.01	−2.19	1.96

	Mean vmPFC	SD vmPFC	Mean Comparison	SD Comparison
B4.1 Trial 1

House A	0.56	2.01	0.34	1.96
House B	−1.38	2.01	−1.50	1.96
House C	2.88	2.01	3.41	1.96
B4.2 Trial 2
House A	−16.52	4.01	−15.62	3.92
House B	6.08	4.01	4.11	3.92
House C	−6.96	4.01	−5.74	3.92
B4.3 Trial 3
House A	5.81	2.01	5.47	1.96
House B	5.38	2.01	4.72	1.96
House C	5.81	2.01	5.91	1.96
B4.4 Trial 4
House A	1.19	4.01	1.98	3.92
House B	−11.21	4.01	−12.05	3.92
House C	0.25	4.01	1.10	3.92
B4.5 Trial 5
House A	1.63	2.01	1.88	1.96
House B	−4.56	2.01	−3.97	1.96
House C	−1.13	2.01	−1.44	1.96
B4.6 Trial 6
House A	−2.52	4.01	−4.14	3.92
House B	−1.39	4.01	−0.72	3.92
House C	1.82	4.01	0.38	3.92
B4.7 Trial 7
House A	−1.69	2.01	−2.56	1.96
House B	−2.00	2.01	−3.88	1.96
House C	3.56	2.01	3.78	1.96
B4.8 Trial 8
House A	3.56	2.01	2.94	1.96
House B	−4.00	2.01	−4.13	1.96
House C	5.19	2.01	5.31	1.96
B4.9 Trial 9
House A	−4.89	4.01	−4.39	3.92
House B	−6.02	4.01	−7.56	3.92
House C	11.58	4.01	10.83	3.92
B4.10 Trial 10
House A	−0.88	2.01	0.34	1.96
House B	3.00	2.01	1.50	1.96

Person A
is so emotional that everyone walks on eggshells around them so as not to upset them	Negative	3.38	0.62	3.06	1.27
loaned their notes to someone who had missed the meeting at work	Positive	1.94	1.18	1.59	0.80
threw a surprise birthday party for their friend	Positive	1.94	0.93	1.22	0.91
helped a friend paint a room in their house	Positive	1.63	0.89	1.38	0.91
Person B
cursed at the flight attendant	Negative	2.75	1.13	2.47	1.08
is very unapproachable	Negative	2.88	1.09	3.00	0.76
often cuts off drivers on the road	Negative	2.63	0.81	2.28	0.99
always obey laws, even if they’re unlikely to get caught	Positive	2.94	0.93	3.13	0.84
Person C
routinely take more than their share of food	Negative	1.56	0.96	1.75	1.02
stands too close when they talk to you	Negative	1.63	1.02	1.13	0.94
is very courteous to everyone they meet	Positive	2.88	0.89	2.63	1.01
tries to see the humorous side of things when faced with problems	Positive	2.19	0.98	2.75	1.08

Person A
picked their nose before serving the guests’ food	Negative	3.13	1.09	2.47	1.29
threw a fit at the store when the credit card was declined	Negative	2.69	1.14	2.25	0.92
yawned loudly at the funeral	Negative	1.81	0.91	1.09	1.12
chews with their mouth open	Negative	1.69	1.14	1.50	1.19
is genuinely good-natured	Positive	2.94	0.93	3.13	0.75
is very flexible and can adapt quickly to changing circumstances	Positive	2.69	0.87	2.94	0.76
is willing to admit when they make a mistake	Positive	2.06	1.24	3.16	0.85
is extremely loyal to friends and family	Positive	3.38	0.62	3.56	0.56
Person B
humiliated you in front of their friends	Negative	3.25	0.86	3.31	1.06
kicked the neighbor’s dog for fun	Negative	3.13	1.09	3.03	1.18
let the excited passenger have the window seat on the plane	Positive	1.81	1.17	1.31	0.97
offered to help their neighbor mow the lawn	Positive	2.25	0.77	1.69	0.90
smiles often	Positive	2.63	1.09	2.41	1.36
teaches dance classes for the elderly	Positive	1.44	1.09	1.25	0.95
carpools to work every day, even though they don’t have to	Positive	1.69	1.14	1.16	0.92
picked up a neighbor’s mail while they were away	Positive	1.81	1.11	1.38	0.98
Person C
is undisciplined and never get things done on time	Negative	2.81	1.22	3.00	0.88
minimized the seriousness of the assault victim’s injuries	Negative	2.38	0.89	2.50	1.08
often tries to get even, rather than forgive and forget	Negative	3.31	0.70	3.03	0.86
pouts when they don’t get their way	Negative	2.38	0.96	2.72	1.02
takes criticism very poorly	Negative	2.69	0.87	2.75	0.92
snuck out of the restaurant without paying	Negative	3.19	0.91	3.22	0.91
baked cookies for sick children in the hospital	Positive	2.13	0.62	1.72	0.99
gave the homeless person some old clothes	Positive	1.75	1.00	2.00	1.16

Person A
spent their monthly paycheck buying lottery tickets	Negative	3.06	1.06	3.44	0.76
watches 6–8 hours of TV every day	Negative	2.56	1.03	2.38	1.29
won’t ever lend their possessions to friends	Negative	2.00	1.15	1.88	0.91
devotes themselves completely to others when they need help	Positive	2.69	0.79	2.44	0.91
Person B
yells at you if you put your drink on the table without a coaster	Negative	1.44	0.73	1.22	1.04
helped the newcomer set up their office	Positive	2.13	1.02	1.31	0.97
reported that they scratched the rental car while driving it	Positive	2.13	1.02	1.75	0.98
ate their friend’s bad cooking without complaining	Positive	1.94	0.85	1.94	0.98
Person C
made fun of the poor family living out of their car	Negative	2.75	1.24	2.59	0.98
threw a tantrum to get attention	Negative	3.38	0.72	3.06	0.84
remodeled their parents’ house for them	Positive	2.63	1.09	1.91	0.93
tutored the neighbor’s child for free	Positive	2.19	0.91	1.56	1.01

	Valence	Mean vmPFC	SD vmPFC	Mean Comparison	Comparison SD
House A
does not have much natural light	Negative	1.88	0.89	2.06	1.24
has a large finished-out attic	Positive	2.38	1.09	2.06	0.98
has pet stains on the carpet	Negative	2.06	0.57	1.91	1.12
is near local police and fire stations	Positive	1.94	0.77	1.72	1.25
House B
has a stove with only two working burners	Negative	1.38	0.81	1.53	0.95
has bare concrete floors	Negative	1.81	1.28	2.28	1.02
is at the end of a quiet cul-de-sac	Positive	2.56	1.03	2.63	1.21
is missing shingles on the roof	Negative	2.44	1.21	2.47	1.05
House C
has a built-in water filter	Positive	1.25	0.93	1.25	1.05
has a wrap-around porch	Positive	2.44	0.73	2.41	1.07
has an oven that cooks unevenly	Negative	1.56	0.96	1.25	0.92
has lush green grass	Positive	1.88	1.09	1.75	1.44

House A
has a green swampy pool	Negative	2.69	1.08	2.66	1.15
has driveways and sidewalks that drain water away from house	Positive	2.81	0.98	2.59	1.01
has drywall with no cracks and straight corners	Positive	2.25	1.29	2.75	1.05
has no garbage pick-up service	Negative	2.31	1.40	2.50	1.14
has windows that leak air	Negative	2.25	1.00	2.53	1.02
has wood burning stove heating that only heats one room of the house	Negative	3.19	1.11	2.75	1.22
smells like cigarette smoke	Negative	2.81	0.98	2.75	1.08
smells like wet dog	Negative	2.56	1.15	2.88	1.10
House B
comes with a flat-panel TV mounted on the wall	Positive	1.06	1.00	1.16	1.22
has a jacuzzi bathtub	Positive	0.75	1.24	0.97	1.03
has a new top-of-the-line refrigerator installed	Positive	2.06	1.06	1.91	1.15
has an exercise room furnished with all new equipment	Positive	1.19	1.17	1.34	1.00
has bedrooms painted with lead paint	Negative	2.94	1.44	3.06	1.05
has custom-made sinks in the bathroom	Positive	1.63	1.02	1.25	1.05
has high speed wireless internet	Positive	1.00	1.03	1.13	1.07
has termites in the wood frame	Negative	3.69	0.60	3.91	0.39
House C
does not have a dishwasher	Negative	0.75	0.68	1.13	1.01
has a cracked driveway and sidewalk	Negative	1.38	1.02	1.66	1.04
has a full bathroom for every bedroom	Positive	2.44	0.73	2.25	1.05
has fully-grown trees in the yard	Positive	2.13	1.02	2.34	1.31
has hidden solar panels on the roof to generate power for the house	Positive	2.94	1.06	2.78	1.01
has torn screens on the windows	Negative	1.06	1.06	1.47	0.92
is available with low mortgage interest rates	Positive	3.25	1.00	3.53	0.62
is not wired for cable TV	Negative	0.81	0.91	1.09	0.93

House A
floods every time it rains heavily	Negative	3.88	0.50	3.63	0.71
has a built-in microwave oven	Positive	0.63	0.89	0.91	0.82
has a heated towel rack in the bathroom	Positive	0.25	0.45	0.53	0.67
has a light dimmer installed in the dining room	Positive	0.44	0.73	0.72	0.89
House B
doorbell doesn’t work	Negative	0.38	0.62	0.36	0.75
has a broken toilet seat	Negative	0.50	0.73	0.75	0.88
has a reasonable monthly mortgage payment	Positive	3.25	0.68	3.34	0.75
has an old mailbox that is not weather-proof	Negative	0.25	0.58	0.47	0.62
House C
has a sagging roof	Negative	3.81	0.54	3.72	0.52
has a tall cedar privacy fence	Positive	1.25	1.13	1.66	1.10
has a toilet that overflows	Negative	2.38	1.09	2.03	1.09
is close to a good grocery store	Positive	1.25	1.24	1.69	1.12

House A
does not have a bathroom door	Negative	1.38	1.36	1.50	0.84
has a leaky hot water heater	Negative	2.06	1.06	1.97	1.15
has a well-maintained garden	Positive	1.50	1.03	1.50	1.11
has brand-new drawers that open and close smoothly	Positive	1.81	1.11	1.75	1.19
has carpet that is tight and the seams don’t show	Positive	1.81	0.83	1.97	1.15
has dark black-out window shades to reduce heat in the summer	Positive	1.75	1.06	1.50	1.02
has newly finished hardwood floors	Positive	2.25	1.34	2.59	1.04
is made of strong, beautiful brick	Positive	2.94	1.06	2.84	0.95
House B
has a large open kitchen	Positive	2.50	1.10	2.31	0.90
has a major draft through the front door	Negative	1.98	0.89	2.25	1.14
has lots of natural light	Positive	2.38	1.09	2.28	1.22
has no air conditioning and poor ventilation	Negative	3.38	0.72	2.94	0.98
has plenty of storage space	Positive	2.50	0.89	2.59	0.80
has secured and clear gutters	Positive	2.06	1.18	2.03	1.28
has walls made entirely of cinder blocks	Negative	2.69	1.01	2.50	1.11
has water-marked ceilings	Negative	3.06	0.85	3.09	0.96
House C
has a stained and mildew-covered bathtub	Negative	2.34	1.21	2.13	1.04
has an energy-saving programmable thermostat	Positive	1.63	1.31	1.56	1.19
has an unreliable furnace	Negative	2.88	1.15	3.00	0.92
has cold, broken tile floors in the bathroom	Negative	2.38	0.89	2.16	0.95
has gopher tunnels in the front yard	Negative	1.75	1.24	1.94	1.24
has peeling paint in the bathroom	Negative	1.50	0.89	1.59	1.07
is very close to a great walking trail	Positive	1.88	1.36	1.47	1.08
previously had a fire in the attic	Negative	3.50	1.03	2.97	0.97

House A
has a front door that is warped and doesn’t close properly	Negative	2.06	1.12	2.13	1.16
has bad tasting water	Negative	3.38	0.72	2.63	1.13
has excellent topsoil for gardening	Positive	1.63	1.09	1.56	1.19
includes snow removal from the walkway during winter	Positive	1.00	1.15	1.91	1.20
House B
does not have a washer and dryer	Negative	0.75	0.93	1.34	1.04
has outdated wallpapering throughout the entire house	Negative	1.44	1.03	1.63	1.07
has windows that open and close easily	Positive	2.50	1.03	2.19	0.93
has wooden doors with scratches and splinters	Negative	1.19	0.75	1.59	1.13
House C
has a large, oversized pantry	Positive	1.31	1.01	1.88	0.91
has brand-new carpet	Positive	2.06	1.12	2.09	1.06
has extremely small closets	Negative	2.19	0.75	2.00	0.88
is located in a peaceful neighborhood	Positive	2.88	0.96	3.13	0.94

House A
has a clean-burning natural gas stove	Positive	2.00	1.10	2.28	1.14
has a deck with rotting wood	Negative	2.44	0.89	2.66	1.00
has a lawn overgrown with weeds	Negative	1.63	1.20	1.78	1.31
has a spacious guest room above the garage	Positive	1.94	0.77	1.84	0.95
has no insulation in the attic	Negative	3.13	0.89	2.60	1.23
has weathered, cracked siding on the outside	Negative	3.25	0.68	2.97	0.74
is near a noisy highway	Negative	2.81	1.28	2.94	0.95
is near a sewage treatment plant	Negative	3.38	0.96	3.09	0.96
House B
has a broken step leading upstairs	Negative	1.06	0.77	1.44	0.67
has a built-in air purifier	Positive	1.13	1.31	1.31	1.06
has a library with plenty of built-in shelves	Positive	1.50	1.21	2.06	1.05
has a very large supply of hot water	Positive	1.75	1.00	2.19	0.93
has granite countertops in the kitchen	Positive	1.81	1.17	1.78	1.18
has gutters plugged with leaves and debris	Negative	1.56	0.96	1.31	0.93
has smoke and carbon monoxide detectors pre-installed in prime locations	Positive	2.56	1.09	2.03	0.97
has strong water pressure in the showers	Positive	2.06	0.93	2.00	1.08
House C
comes with free cable TV for 3 years	Positive	0.44	0.73	1.31	1.12
has a large laundry room	Positive	1.63	1.20	1.88	1.18
has a top-of-the-line oven and range	Positive	1.69	1.08	1.94	1.01
has a wide, easily-accessible driveway	Positive	1.88	0.96	2.06	1.13
has asbestos in the attic	Negative	3.56	0.81	3.34	0.90
has no garage or carport	Negative	2.44	1.26	2.78	1.07
has water pipes not constructed according to building codes	Negative	3.56	0.73	3.41	0.91
shakes when trains pass by on the railroad	Negative	3.38	0.62	2.97	1.00

House A
has a lot of junk left in the attic	Negative	1.19	0.91	1.41	0.95
has an energy efficient furnace	Positive	2.81	1.17	3.00	0.92
has reliable plumbing	Positive	3.06	0.85	3.22	0.91
has toilets with water stains	Negative	1.06	1.18	1.16	1.11
House B
comes with a garage-door opener	Positive	0.69	0.60	1.13	1.04
has a mosquito prevention system in the back yard	Positive	1.06	1.29	1.34	1.26
has water accumulate in the basement whenever it rains	Negative	3.88	0.34	3.56	0.76
is furnished with beautiful furniture	Positive	1.50	1.32	1.97	1.20
House C
has a large hole in the drywall	Negative	1.56	1.03	2.06	0.98
has a modern, but classic design	Positive	1.56	1.03	1.97	1.06
has a toilet which often plugs up	Negative	2.44	0.96	2.09	1.12
has nails poking out from the walls	Negative	2.13	0.96	2.28	1.08

PERMALINK

Choosing Spouses and Houses: Impaired Congruence Between Preference and Choice Following Damage to the Ventromedial Prefrontal Cortex

Mark D Bowren

Katie E Croft

Justin Reber

Daniel Tranel

Abstract

Objective

Method

Results

Conclusions

Methods

Participants

Apparatus

Stimuli

Procedure

Forced-Choice Preference Task

Visual Analogue Preference Task

Attribute-Rating Task

Measures

Internal Consistency

Response Time

Attribute-Based Choice Difficulty

Results

Preliminary Analyses

Internal Consistency

Exploratory Analysis

Conclusions

Public Significance Statement.

Acknowledgments

Appendix A

Table 1.

Table 2.

Table 3.

Table 4.

Figure 1.

Figure 2.

Figure 3.

Figure 4.

Figure 5.

Figure 6.

Appendix B Social (Spouse) and Non-Social (House) Stimuli

B1 Spouse Stimuli and Influence Ratings

B1.1 Practice Trial

B1.2 Trial 1

B1.3 Trial 2

B1.4 Trial 3

B1.5 Trial 4

B1.6 Trial 5

B1.7 Trial 6

B1.8 Trial 7

B1.9 Trial 8

B1.10 Trial 9

B1.11 Trial 10

B2 House Stimuli and Influence Ratings

B2.1 Practice Trial

B2.2 Trial 1

B2.3 Trial 2

B2.4 Trial 3

B2.5 Trial 4

B2.6 Trial 5

B2.7 Trial 6

B2.8 Trial 7

B2.9 Trial 8

B2.10 Trial 9

B2.11 Trial 10

B3 Spouse WADD Scores

B4 House WADD Scores

Contributor Information

References

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases

House A
has a filthy fireplace and chimney	Negative	1.94	1.12	1.94	0.98
has a newly installed washer and dryer	Positive	1.63	1.09	1.84	0.81
has cigarette burns in the carpet	Negative	2.38	0.89	2.06	1.11
has rotten wooden door frames	Negative	2.69	0.95	2.75	1.02
House B
does not have locks on the front door	Negative	1.50	1.21	1.47	1.05
has a leaky faucet in the kitchen	Negative	1.13	0.96	1.31	0.86
is a great buy for the price	Positive	3.13	0.81	3.59	0.61
recently had a new roof installed	Positive	3.56	0.63	3.28	0.81
House C
has a fragrant rose garden in the backyard	Positive	0.81	0.98	1.88	1.02
has a heated garage	Positive	1.56	1.31	1.50	1.08
has black mold behind the drywall	Negative	3.94	0.25	2.66	0.79
has luxurious showerheads in all the showers	Positive	1.25	1.24	1.06	1.11

House A
has a bathtub with a slow-running drain	Negative	1.31	0.87	1.78	0.91
has a solid foundation	Positive	3.44	1.03	3.66	0.55
has creaky doors	Negative	0.81	0.75	1.34	0.87
has a newly installed washer and dryer	Positive	2.75	1.13	2.78	1.01
has energy efficient and cooling fans throughout the house	Positive	2.38	1.31	2.69	1.00
has huge oil stains on the driveway in front of the house	Negative	0.81	0.83	1.25	0.84
has low property taxes	Positive	3.13	1.15	3.22	0.91
has no smoke detectors installed	Negative	1.06	1.06	1.38	1.31
House B
has a hole in the kitchen floor	Negative	2.56	1.15	2.47	1.08
has an uneven floor	Negative	2.81	0.66	2.56	0.91
has bright wide windows	Positive	2.19	1.22	2.38	1.04
has high electricity bills	Negative	3.25	0.93	3.03	0.82
has noisy squirrels living in the attic	Negative	2.81	1.17	2.94	1.11
has very large walk-in closets	Positive	2.00	1.10	2.56	0.91
is in a neighborhood with declining home values	Negative	2.81	1.33	3.16	0.88
is located at the corner of a loud intersection	Negative	3.38	0.72	2.63	1.16
House C
has a built-in surround sound stereo	Positive	0.69	0.95	1.16	1.11
has a faulty septic tank	Negative	3.25	1.06	3.13	1.10
has a luxurious fireplace next to the bathtub	Positive	0.75	0.93	1.19	1.18
has a media room with movie seating	Positive	1.25	1.13	1.25	1.30
has a sprinkler system installed	Positive	1.69	1.30	1.63	1.04
has an extra freezer in the garage	Positive	0.81	1.05	1.06	1.05
has an infestation of cockroaches	Negative	3.63	0.62	3.41	0.84
has newly installed blinds on the windows	Positive	1.25	1.00	1.63	1.01

House A
has a broken window	Negative	1.00	0.89	0.97	0.90
has birds living in the chimney	Negative	2.44	1.09	2.13	1.18
has old dusty shag carpet in the den	Negative	1.56	0.96	1.94	0.95
has plenty of outdoor lighting at night	Positive	1.31	1.08	1.53	1.11
House B
has a rusting chain link fence with holes in it	Negative	1.44	1.21	1.78	0.97
has a screened-in porch to keep the bugs out	Positive	1.75	1.39	2.34	0.94
has an alarm system monitored by the police	Positive	1.56	0.73	1.47	1.11
has very little storage space	Negative	2.56	0.63	2.44	0.84
House C
has a storm shelter	Positive	1.69	1.25	1.84	1.19
has air conditioning ducts that have never been cleaned	Negative	1.13	1.02	1.44	0.91
has an excellent neighborhood park nearby	Positive	2.13	1.02	2.41	1.19
is built from the finest wood available	Positive	2.69	1.30	2.84	1.17