Impaired movie taste in patient RD:
A quantitative assessment
Reginald Dwight
University of California, Davis
Abstract
Despite widespread belief that bad movie-taste results from focal brain lesions, few studies have systematically tested this hypothesis. Here we present data from RD, a patient suspected to have impaired movie taste, amovia, following an anoxic episode. RD ranked 1140 movies, and multiple regression analyses were performed to determine whether he had deficits in his movie selecting or judging behavior. Although his selecting behavior deviated from Internet Movie Database norms, his overall rankings highly correlated with these same norms, indicating a selective impairment in movie choice, but not judgment. Additionally, the genre predictors used in this analysis appear to be insufficient to fully characterize his movie enjoyment, but he does show a general dislike of romance, family, crime, and more recent films. Combined with a lack of a large general bias toward any movie type, there appears only to be sparse evidence of amovia in RD.
Although movie preference is generally thought to be a variable personality attribute, there is mounting evidence that bad movie taste is specifically mediated by damage to frontal region μV. For instance, Ivory et al. (2001) have reported a group of patients in Ottawa whose damage to area μV causes them to wander aimlessly through Blockbuster Video only to rent The Mexican. This seemingly benign behavior is primarily believed to affect movie specific socialcognitive functioning, but Ivory also reports an overall decrease in neural mass in such patients. His evidence is confounded, however, because it is unclear whether the additional atrophy stems from plasticity induced changes in areas adjacent to μV, or general brain deterioration following repeated viewings of the The Mexican.
To better characterize the role of area μV in maintaining quality movie preferences, we present here a case-study of patient RD, a man with severe damage to μV. RD’s movie-selectve impairment has sparked much interest within the neuropsychology community (Smacks, 1999), and this is largely due to his ownership of movies with titles such as Redneck Zombies and Rabid Grannies. Yet despite numerous anecdotal and qualitative reports, no studies to the authors’ knowledge have quantitatively assessed RD, or any other patient’s movie-related deficits (amovia). We used multiple regression to determine the factors mainly influencing RD’s current movie taste, and will discuss them in terms of behavioral norms acquired from the Internet Movie Database (IMDB).
Several criticisms of the amovia literature have focused on the inconsistencies between what researchers call bad taste. Throughout this paper, bad taste will refer to (1) genre bias and (2) a tendency to orient to, laugh at, quote, or sing songs from movies scoring lower than 6.2/10 on the Internet Movie Database. Genre bias, also known as the Emperor’s New Clothes Syndrome (Smacks, 1998), refers to a situation where a person favors a movie regardless of its actual content. Rather, they show favoritism toward the movie because it is supposed to be “good,” it is part of a genre they reliably favor, or they are too embarrassed to admit that they wasted $9.50 on a movie about talking babies. The second aspect of bad taste is found in patients with VSND, Video Store Navigational Deficits. Symptoms of this include walking directly toward the horror section, completely failing to attend to the Independent, Classics, or Foreign film sections of the video store. The owners of Blockbuster Video are believed to suffer from this disorder. Similarly, patient RD has been observed to possess a mild form of this deficit (Smacks, 1998), but recent evidence challenges this diagnosis, suggesting instead that he was actually searching for a respectable horror movie and not Dr. Giggles (Slackter, 1999).
Method
Patient
In 1998, patient RD suffered an anoxic episode while cleaning his refrigerator. Mistaking a six-month old spaghetti frittata for corn chowder, he accidentally opened a Tupperware™ without proper breathing apparatus and subsequently lost consciousness. He was revived 6 minutes later, and despite a strong desire to see the new Godzilla movie, he appeared normal. His roommates referred him to a neurologist after he was witnessed renting Leprechaun in the Hood.
As reported in Slackter (1999), RD has various cognitive and emotional deficits , but they only manifest themselves in film-related contexts. For instance, his desire to see Surf Nazis Must Die!!! was intentional, and not due to semantic dementia as had once been suspected. Similarly, his high vocabulary, high sentence comprehension skills, and ceiling performance on semantic decision tasks (Covey, 1997) make it unlikely that he purchased Sgt. Kabukiman NYPD without knowing what to expect.
A structural MRI (Figure 1) confirmed the behavioral reports, revealing extensive damage to frontal regions, extending dorsally from the precentral sulcus toward midline structures. This appears to lie within the boundaries of area μV which Slacks (1999) has defined as crucial for orienting and selecting decent movies in a video store.
Figure 1. Patient RD’s reconstructed and cortically inflated brain.
Sample
RD provided experimenters with a list of 1250 feature-length movies (mean year = 1985, standard deviation = 12.7) that he had seen in full. Frontal damage often leads to impairments in source memory (see Covey, 1997), so we were not unable to verify that he had seen each and every film and we were likewise unable to determine when he last saw each film on the list. Therefore, items for which RD could not recall any specific viewing conditions, details, scenes, dialogue, etc. were immediately removed from the stimulus set. Using a recursive sorting technique, RD was instructed to rank the remaining 1140 films according to how much he would enjoy the movie were he to watch it again today (see Appendices A and B for his highest and lowest rated films). During the first test session, ranking was done via a binary response (bottom vs. top 50%). Each of these bins was then broken into 5 further groups, giving a rough ranking of 1-10. Preliminary analyses revealed this to be too gross a ranking, so each bin was further broken into 4 subsets. Thus, ratings ranged from a minimum of -19.5 to a maximum of 19.5, with 28 or 29 samples per bin. Please note that the retrospective nature of the study addresses his current taste, not his movie taste prior to the lesion. We are not concerned about how much he enjoyed each movie upon initial viewing. Rather, this study assumes that his movie taste has changed as a function of the lesion, and his projective, not retrospective, opinions are what concerns our assessments of his amovia.
Procedure:
Each film’s entry on the IMDB was queried via a C-Shell Unix script. Data were collected for two continuous predictor variables, year and IMDB weighted ranking (0 min, 10 max); and categorical variables, foreignness and genre. Movies that were produced solely by non-American companies were considered foreign, and genre reflected the 14 major genres that the IMDB lists for each movie: Action, Adventure, Animation, Comedy, Crime, Documentary, Drama, Family. Fantasy, Film-Noir, Horror, Independent, Musical, Mystery, Romance, Sci-Fi, Thriller, War, and Western. These were not mutually exclusive categories, and over 50% of the movies were in two or more categories.
A 0-1 dummy variable was created for each categorical variable with 75 or more instances (foreign (19%), action (22%), adventure (10%), comedy (48%), crime (8%), family (7%), fantasy (10%), horror (13%), romance (10%), sci-fi (15%), and thriller (15%)). There were only a handful of films that were not members of those genres, and they did not constitute a large enough comparison group. Because drama is highly general category, with 81% of dramas having an additional representation in another genre, exclusive dramas were added to the comparison group. This total comparison group represented approximately 10% of all films in the sample.
A preliminary analysis was also performed to determine any patterns in RD’s movie viewing habits independent of his attitude towards them. It is possible that RD’s rental habits reflect his amovia, but he still has preserved ability to discriminate an generally good from bad film. To assess this, we performed a regression analysis on IMDB rating, with the same predictors as outlined above for predicting RD’s own ratings. This analysis assumes that the ratings for the category have comparable means and distributions within the IMDB, and that there are no systematic biases to favor certain film genres within the IMDB.
Results and Discussion
Sample characteristics
While there was no relationship between IMDB rating and romance, family, adventure, or crime genres (all p’s > 2), the sample appears to be biased towards other dimensions, hinting at severe amovia with respect to RD’s film choosing behavior (R2 = .19, F(15,1124) = 17.539, p < .001) (Table 1). There is a negative relationship between year and IMDB rating, with older movies in the sample having higher ratings than more recent films (B = 0, β = -.105, r = -3.807, p < .001). Additionally, foreign films in the sample tended to be highly rated (B = -, β = .025, r = - 4.349, p < .001). These two effects are likely non-causal since the availability of older and foreign films is often modulated by their quality. That is, low quality old or foreign films are probably not available in video stores.
Fantasy and thriller were the only other film categories for which the sample seemed biased towards highly ranked movies (all p’s < .001). On the other hand, RD has strong inclinations to watch bad action, comedy, horror, and sci-fi films (all p's < .001) reflecting the large representation of B-movies in the sample. Together, these results suggest that RD has a severe Video Store Navigation Deficit.
Table 1.
Multiple Regression Equation Predicting Internet Movie Database Ranking of a Film Based on its Selection by Patient RD
| IMDB Ratings | |||
|---|---|---|---|
| Variable | b | SE | b |
| Year | -1.042E-02 | .003 | -.105*** |
| Action | -.384 | .092 | -.126*** |
| Adventure | -2.565E-02 | .128 | -.006 |
| Comedy | -.565 | .082 | -.224*** |
| Crime | 4.939E-02 | .140 | .010 |
| Fantasy | .187 | .124 | .044 |
| Family | -1.751E-02 | .142 | -.003 |
| Foreign | .555 | .090 | .175*** |
| Horror | -1.209 | .125 | -.326*** |
| Romance | 8.636E-02 | .190 | .020 |
| ci-Fi | -.401 | .100 | -.114*** |
| Thriller | .608 | .104 | .170*** |
| Comedy x Horror | .361 | .231 | .051 |
| Crime x Romance | .358 | .239 | .069 |
| Foreign x Crime | .236 | .379 | .018 |
| Intercept | .319 | 0.73 | 0*** |
Note: Sample consists of 1140 feature-length films seen by patient RD, and for which RD can recollect something about the viewing conditions, film content, etc. R2 = .190. * p < .05. ** p < .01. *** p < .001
Zero-Order Correlations
The dependent variable, RD’s movie ratings, significantly correlated with many of the predictors. RD’s rating had a .627 (p < .001) correlation with the IMDB ratings, and this was by far the strongest zero-order correlation in the analysis. Thrillers (r = .091, p = .001) and foreign films (r = .108, p < .001) also had small but significant positive correlations with RD's ratings. Alternatively, comedies (r = -.07, p = .009), sci-fi (r = -.049, p < .05), and horror movies (r = - .133, p < .001 were negatively related to rating. These results are puzzling when one considers the frequency with which RD actively pursues comedies, sci-fi, and horror films (48%, 15%, and 13%, respectively). The discrepancy between his choosing and judging behavior hints at VSND, and could be due to a lack of meta-awareness of his own movie tastes, undiagnosed perceptual deficits, or masochism.
Additionally, RD’s well known dislike of family films (Dewey et al., 1999) was reflected in the association between family films and his rankings (r = -.067, p = .01).
Table 2.
Multiple Regression Equations Predicting Whether RD Will Like A Movie
| RD’s Ratings | |||
|---|---|---|---|
| Variable | b | SE | b |
| IMDB Ranking (IMDB) | 5.401 | 0.364 | 0.589*** |
| Year | -0.109 | 0.025 | -0.12*** |
| Year x IMDB | 0.107 | 0.018 | 0.161*** |
| e^IMDB | 0.415 | 0.243 | 0.068 |
| Action (A) | -0.261 | 0.755 | -0.009 |
| Adventure (Adv) | -0.261 | 0.99 | -0.007 |
| Comedy (Com) | 0.465 | 0.673 | 0.02 |
| Crime (Cri) | -2.512 | 1.089 | -0.058* |
| Family (Fam) | -3.259 | 1.095 | -0.07** |
| Fantasy (Fan) | 0.603 | 0.962 | 0.015 |
| Foreign (For) | -1.315 | 0.774 | -0.045 |
| Horror (Hor) | -1.4 | 1.007 | -0.041 |
| Romance (Rom) | -4.251 | 1.651 | -0.109** |
| Science-Fiction (Sci-Fi) | 0.422 | 0.776 | 0.013 |
| Thriller (Thr) | 1.071 | 0.818 | 0.033 |
| Com x Hor | 3.603 | 1.794 | 0.055* |
| Com x Rom | 4.201 | 1.921 | 0.088* |
| For x Cri | 5.901 | 2.97 | 0.05* |
| For x Rom | 3.964 | 2.405 | 0.045 |
| For x Com x Act | 1.777 | 1.968 | 0.024 |
| Intercept -0.307 | 0.755 | 0 | |
Note: Sample consists of 1140 feature-length films seen by patient RD, and for which RD can recollect something about the viewing conditions, film content, etc. R2 = ..656. * p < .05. ** p < .01. *** p < .001
Multiple Regression Analyses
The multiple regression model accounted for 43% of the variance in the data (R2 = .43, F(20,1119) = 42.197, p < .001). The large amount of unexplained variance could be used as evidence against amovia because it suggests that RD's movie taste is idiosyncratic of the movie, and not necessarily due to higher order aspects of films such as genre, year, or foreignness. Indeed, he likely is not biased towards a film a priori. Rather, he may judge it based on individual characteristics or a higher-level factor not modeled in our regression equation, e.g., director, actor, or presence of talking farm animals.
There was only one major outlier, The Jade Pussycat, which had a standardized residual of -3.267, not surprising since this was a John Holmes film. It received an IMDB rating of 8.7/10 and 3/40 by RD. Since only seven people scored this movie for the IMDB, the discrepancy between ratings is likely due to measurement bias by the IMDB. Its leverage could not be calculated using the statistical package used by the authors, but exploratory analysis of removing the variable did not severely change any of the major results.
Continuous variables
As suggested by the zero-order correlations, IMDB rating was a strong predictor of RD’s movie rankings (B = 5.401, β = .589, t = 14.836, p < .001). The shape of the IMDB raw scores led us to test for an exponential term in the regression equation, but this was only marginally significant (B = -.415, β = .068, t = 1.711, p = .087). Additionally, RD tends to rank older films more highly than recent ones (B = -.109, β = .025, t = -4.349, p < .001). A year x IMDB interaction suggests that his nostalgic bias dissipates with recent movies that the IMDB ranks highly, for instance, Memento and Amelie.
Genre analysis
As suggested by the zero-order correlations, a categorization as a family film predicts a negative rating by RD (B = -3.259, β = 1.095, t = -2.977, p = .003). Romance also appears to be a negative predictor of movie preference (B = -4.251, β = -.11, t = -2.575, p = .01). This was consistent with RD’s self-report disgust of what he terms “simple-minded and manipulative love stories” (Ex-girlfriend et al., 1995), but did not represent his tendency to like Woody Allen films and romantic foreign films. This led us to test interaction terms for Comedy x Romance (CR) and Foreign x Romance (FORROM). Only the CR term proved significant at the .05 level (B = 4.201, β = .088, t = 2.187, p = .029).
Approximately 5% of the sample consisted of Kung-Fu movies, and the authors were surprised to see no patterns revealing a preference for the genres with which these films are associated. RD is known to rent many Jackie Chan films, so we tested for a three way interaction between Foreign x Action x Comedy. This revealed no significant results (t = .903, p = .367). There was no main effect for action (t = -.346, p = .729), but crime films tended to be rated lower than the comparison group (B =-2.5121, β = 1.089, t = -2.306, p = .021). A Foreign x Crime interaction did exist (B = 5.901, β = .05, t = 1.987, p = .047), however, suggesting that while crime films are rated low overall, crime films made outside of America are rated positively. An overview of the sample’s Foreign Crime films revealed that 1/3 starred Jackie Chan. The other 2/3 consisted mainly of films where ingenious plans went awry (e.g., Elevator to the Gallows or Lock Stock & Two Smoking Barrels).
Although there were no main effects for comedy or horror (all p’s > .15), we suspected that an interaction between the two might explain RD’s tendency to rent B-horror movies, a telltale sign of amovia. That interaction was indeed significant (B = -.109, β = .025, t = -4.349, p < .001), suggesting that funny horror movies are highly favored by RD, but comedy and horror, alone, have no predictive power. This interpretation could explain why Blood Feast and Dead Alive ranked highly on RD's list.
General Discussion
Presented above is the first quantitative evidence of amovia in a human subject. Patient RD showed favoritism for certain genres, some of which (e.g., comedic horror films) are of questionable taste. He also tends to voluntarily watch action, comedy, horror, and sci-fi movies that receive low ratings on the IMDB, strong evidence for a Video Store Navigation Deficit. However, any conclusion of abnormality should be considered with caution. First, the high correlation between IMDB norms and patient RD’s rankings suggests only a mild form of amovia. RD seems to have fairly well rounded movie tastes as evidenced by the small b weights for all items but IMDB ranking. In fact, were we to include the gender of IMDB voters as a predictor, we may find that his tendencies to dislike romance and family films, and to favor comedy horror movies, is consistent with male behavior, and not related to damage to area μV per se. Second, the combination of retrospective and projective techniques makes the analysis subject to demand characteristics and various other biases. Taken together, these findings suggest that RD’s deficit is mostly related to selecting, but not judging films.
References
Covey, S. (1997). Seven rabbits with highly effective breeding. Chicago: Pizza & Sons.
Dewey, A, Cheatem, A. & Howe, A. (1995) The success of Brendan Fraser: Mass hypnosis or chemical warfare? Journal of Experimental Psychology, Learning, Memory, and Tiger Beat, 1, 2.
Ex-Girlfriend, N., Noying-Friend, A, & Migod, O. (1995). The magically short relationship: Seven plus or minus two weeks. Nature, 7, 102.
Internet Movie Database (n.d.) Accessed from http://us.imdb.com.
Ivory, D., & Phant, L. E. (1944). Julia Roberts induced brain trauma. Brain Lesion Monthly, 8, 1-4.
Smacks, O. N. E.. (1999). The man who mistook “Tomb Raider” for a good movie. New York: Apple.
Slackter, D. (1988). Memories: The Way We Were or Biodome: Evidence for functional segregation in area µV. Neuroimage, 99, 44-48.
Appendix A
Most reviled 56 films as rated by patient RD
|
|
Appendix B:
Favorite 59 films as rated by patient RD
|
|