Experiment. shows the materials used in the study and, for each item, the percentage of choices for the matching cause.

Similar documents
Prototypes in the Mist: The Early Epochs of Category Learning

Encoding processes, in memory scanning tasks

Appendix for. Institutions and Behavior: Experimental Evidence on the Effects of Democracy

Non-linear Multiple-Cue Judgment Tasks

Incorrect Beliefs. Overconfidence. Types of Overconfidence. Outline. Overprecision 4/22/2015. Econ 1820: Behavioral Economics Mark Dean Spring 2015

ALMALAUREA WORKING PAPERS no. 9

Using Past Queries for Resource Selection in Distributed Information Retrieval

N-back Training Task Performance: Analysis and Model

Clinging to Beliefs: A Constraint-satisfaction Model

International Journal of Emerging Technologies in Computational and Applied Sciences (IJETCAS)

(From the Gastroenterology Division, Cornell University Medical College, New York 10021)

Physical Model for the Evolution of the Genetic Code

Inverted-U and Inverted-J Effects in Self-Referenced Decisions

ARTICLE IN PRESS Neuropsychologia xxx (2010) xxx xxx

An Introduction to Modern Measurement Theory

Maize Varieties Combination Model of Multi-factor. and Implement

A Linear Regression Model to Detect User Emotion for Touch Input Interactive Systems

Study and Comparison of Various Techniques of Image Edge Detection

Toward a Unified Model of Attention in Associative Learning

Reconciling Simplicity and Likelihood Principles in Perceptual Organization

Copy Number Variation Methods and Data

The High way code. the guide to safer, more enjoyable drug use [GHB] Who developed it?

Richard Williams Notre Dame Sociology Meetings of the European Survey Research Association Ljubljana,

Bimodal Bidding in Experimental All-Pay Auctions

Does Context Matter More for Hypothetical Than for Actual Contributions?

Parameter Estimates of a Random Regression Test Day Model for First Three Lactation Somatic Cell Scores

Integration of sensory information within touch and across modalities

THE NATURAL HISTORY AND THE EFFECT OF PIVMECILLINAM IN LOWER URINARY TRACT INFECTION.

Were the babies switched? The Genetics of Blood Types i

Journal of Economic Behavior & Organization

Desperation or Desire? The Role of Risk Aversion in Marriage. Christy Spivey, Ph.D. * forthcoming, Economic Inquiry. Abstract

Sparse Representation of HCP Grayordinate Data Reveals. Novel Functional Architecture of Cerebral Cortex

What Determines Attitude Improvements? Does Religiosity Help?

EVALUATION OF BULK MODULUS AND RING DIAMETER OF SOME TELLURITE GLASS SYSTEMS

HIV/AIDS-related Expectations and Risky Sexual Behavior in Malawi

Machine Understanding - a new area of research aimed at building thinking/understanding machines

EFFECTS OF FEEDBACK CONTROL ON SLOW CORTICAL POTENTIALS AND RANDOM EVENTS

The Limits of Individual Identification from Sample Allele Frequencies: Theory and Statistical Analysis

The High way code. the guide to safer, more enjoyable drug use. (alcohol)

Modeling Multi Layer Feed-forward Neural. Network Model on the Influence of Hypertension. and Diabetes Mellitus on Family History of

Economic crisis and follow-up of the conditions that define metabolic syndrome in a cohort of Catalonia,

A MIXTURE OF EXPERTS FOR CATARACT DIAGNOSIS IN HOSPITAL SCREENING DATA

Introduction ORIGINAL RESEARCH

Project title: Mathematical Models of Fish Populations in Marine Reserves

THE NORMAL DISTRIBUTION AND Z-SCORES COMMON CORE ALGEBRA II

310 Int'l Conf. Par. and Dist. Proc. Tech. and Appl. PDPTA'16

Do norms and procedures speak louder than outcomes? An explorative analysis of an exclusion game. Timo Tammi

Gene Selection Based on Mutual Information for the Classification of Multi-class Cancer

Optimal Planning of Charging Station for Phased Electric Vehicle *

Bimodal Score Distributions and the MBTI: Fact or Artifact?

Bonsai Trees in Your Head: How the Pavlovian System Sculpts Goal-Directed Choices by Pruning Decision Trees

Alma Mater Studiorum Università di Bologna DOTTORATO DI RICERCA IN METODOLOGIA STATISTICA PER LA RICERCA SCIENTIFICA

Using the Perpendicular Distance to the Nearest Fracture as a Proxy for Conventional Fracture Spacing Measures

II. Key stimuli in avoidance learning

A GEOGRAPHICAL AND STATISTICAL ANALYSIS OF LEUKEMIA DEATHS RELATING TO NUCLEAR POWER PLANTS. Whitney Thompson, Sarah McGinnis, Darius McDaniel,

Single-Case Designs and Clinical Biofeedback Experimentation

Human development is deeply embedded in social

HIV/AIDS-related Expectations and Risky Sexual Behavior in Malawi

The Influence of the Isomerization Reactions on the Soybean Oil Hydrogenation Process

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

Economists are increasingly analyzing data on subjective well-being. Since 2000, 157

Biased Perceptions of Income Distribution and Preferences for Redistribution: Evidence from a Survey Experiment

National Polyp Study data: evidence for regression of adenomas

Psychological Correlates of Dogmatism and Heterogeneity of Power in Small Groups

arxiv: v1 [cs.cy] 9 Nov 2018

The Reliability of Subjective Well-Being Measures

Delving Beneath the Covers: Examining Children s Literature

Price linkages in value chains: methodology

Using a Wavelet Representation for Classification of Movement in Bed

HYPEIIGLTCAEMIA AS A MENDELIAN P~ECESSIVE CHAI~ACTEP~ IN MICE.

Child abuse and domestic abuse: Factors in reunification

Estimation for Pavement Performance Curve based on Kyoto Model : A Case Study for Highway in the State of Sao Paulo

Survival Rate of Patients of Ovarian Cancer: Rough Set Approach

CLUSTERING is always popular in modern technology

J. H. Rohrer, S. H. Baron, E. L. Hoffman, D. V. Swander

Are Drinkers Prone to Engage in Risky Sexual Behaviors?

Hierarchical Prediction Errors in Midbrain and Basal Forebrain during Sensory Learning

Perceptual image quality: Effects of tone characteristics

Evaluation of Literature-based Discovery Systems

Intact Perceptual Memory in the Absence of Conscious Memory

Fitsum Zewdu, Junior Research Fellow. Working Paper No 3/ 2010

NUMERICAL COMPARISONS OF BIOASSAY METHODS IN ESTIMATING LC50 TIANHONG ZHOU

A Mathematical Model of the Cerebellar-Olivary System II: Motor Adaptation Through Systematic Disruption of Climbing Fiber Equilibrium

ME Abstract. Keywords: multidimensional reliability, instrument of students satisfaction as an internal costumer, confirmatory factor analysis

Can Subjective Questions on Economic Welfare Be Trusted?

4.2 Scheduling to Minimize Maximum Lateness

Addressing empirical challenges related to the incentive compatibility of stated preference methods

Investigation of zinc oxide thin film by spectroscopic ellipsometry

The Effect of Fish Farmers Association on Technical Efficiency: An Application of Propensity Score Matching Analysis

The High way code. the guide to safer, more enjoyable drug use. [cannabis] Who developed it?

CORRUPTION PERCEPTIONS IN RUSSIA: ECONOMIC OR SOCIAL ISSUE?

Prediction of Total Pressure Drop in Stenotic Coronary Arteries with Their Geometric Parameters

Lymphoma Cancer Classification Using Genetic Programming with SNR Features

I T L S. WORKING PAPER ITLS-WP Social exclusion and the value of mobility. INSTITUTE of TRANSPORT and LOGISTICS STUDIES

HERMAN AGUINIS University of Colorado at Denver. SCOTT A. PETERSEN U.S. Military Academy at West Point. CHARLES A. PIERCE Montana State University

A comparison of statistical methods in interrupted time series analysis to estimate an intervention effect

The High way code. the guide to safer, more enjoyable drug use. (lsd / magic mushrooms)

Estimation Comparison of Multidimensional Reliability Coefficients Measurement of Senior High School Students Affection towards Mathematics

Evaluation of the generalized gamma as a tool for treatment planning optimization

KTu DISTRIBUTED BY: National Technical Information Service U. S. DEPARTMENT OF COMMERCE 5285 Port Royal Road, Springfield Va.

Transcription:

J ameson,j.,& Gent ner,d.( 2008).Causalst at usandexpl anat or ygoodness ncat egor zat on.i nb.c.love,k.mcrae,& V.M.Sl out sky( Eds. ), Pr oceed ngsoft he30t hannualconf er enceoft hecogn t vesc encesoc et y( pp.291296).aust n,tx:cogn t vesc encesoc et y.

magnng a robn wth, say, zebra DNA, would make many other typcal propertes of robns mplausble (Wlson & Kel, 1998). There s a long hstory of research on categorzaton demonstratng that some features assocated wth category members are partcularly mportant for categorzaton judgments (e.g., Ahn, et al., 2000; Medn & Shoben, 1988; Rosch & Mervs, 1975; Tversky, 1977). One promnent perspectve on centralty s based on the theory- or knowledge-based vew of conceptual structure (e.g., Carey, 1985; Kel, 1989; Murphy & Medn, 1985; Rps, 1989). Accordng to ths framework, knowledge of a category s, n mportant ways, lke a scentfc theory, comprsng a host of mental explanatons (Murphy & Medn, 1985). In ths vew, category knowledge s seen, not n terms of a prototypcal member or collecton of exemplars, but rather n terms of an explanatory prncple common to category members (p. 298, Murphy & Medn, 1985). Accordngly, categorzaton s typcally vewed as an nference to the best explanaton. Murphy and Medn llustrate ths dea wth a well-known example: a man at a party who jumps nto a pool fully clothed n a busness sut would probably be classfed as ntoxcated, not because he s smlar to the prototype or to nstances of a drunken behavor category, but because beng ntoxcated s the best explanaton for hs behavor (see also Rps, 1989, for a clearly artculated sketch of ths dea). Despte the mportance of the theory-based vew for orentng research on categorzaton and centralty, the framework leaves unspecfed the specfc constrants related to background knowledge that nfluence classfcaton decsons. One mportant response to ths challenge s the dea that nformaton that partcpates n causal relatons s of greatest centralty (Ahn, et al., 2000; Rehder & Haste, 2001; Rehder & Burnett, 2004). In Ahn s (1998) causal status hypothess, causes are weghted more heavly n categorzaton decsons than are the correspondng effects. For example, n the smple case of a sngle causal relatonshp between two features, such as havng wngs and flyng (wth wngs enablng flyng), havng wngs would be gven greater weght than flyng, and would more greatly nfluence categorzaton decsons. Why should causes play a more mportant role than the correspondng effects n the underlyng prncples of a category? Accordng to Ahn, et al. (2000), causal propertes may be seen as generatng other features: e.g., DNA produces external features such as har color. Features such as DNA may thus be regarded psychologcally as most defnng or dagnostc of category membershp, because they form part of the essence or core of a concept. It follows, Ahn, et al. clam, that possessng the most central features would provde better evdence for category membershp than more superfcal features. Whle not dsputng the emprcal valdty of the causal status effect, and the crtcal role of causal nformaton for many f not most knowledge-dependent categorzaton decsons, we worry that the exclusve focus on causal knowledge rsks obscurng the supportng role that causal nformaton plays n explanatory processes. As Lombrozo (2006) notes, explanatons can t be reduced to just the supportng causal nformaton, as explanatons ental a set of factors that go beyond the causal nformaton per se (see also Read & Marcus-Newhall, 1993). For one thng, there are many knds of causal structures (Forbus & Gentner, 1986; Rottmann & Gentner, 2006) e.g., causal chans versus feedback systems that draw on dfferent knds of doman knowledge. An addtonal factor s the preference for smple or parsmonous causal explanatons over those that would nvoke more assumptons (Lombrozo & Carey, 2006; Read & Marcus-Newhall, 1993; Rps, 1989). In a recent study that llumnates the relatonshp between the causal status effect and explanatory structure, Lombrozo (2007b) has shown that the strength of the causal status effect depends partly on the type of explanaton mechanstc (an appeal to drect, proxmate causes) or teleologcal (an appeal to functon) n whch the causal knowledge s embedded. In one experment, partcpants were presented wth a novel category characterzed by two features and were told that one feature (e.g., eats blueberres) causes the other (e.g., has blue fur). Partcpants were then asked to explan why the anmal has the second feature (.e., blue fur), and to estmate the probablty that an object mssng one of the features was a member of the category. Partcpants who provded a mechanstc explanaton (e.g., mentonng eatng blueberres as the proxmate cause) showed a larger causal status effect than those who provded a teleologcal explanaton (e.g., blue fur serves as camouflage). In addton to the type of explanaton, the causal status effect may also depend on the qualty of an explanaton. We suggest that some of the puzzlng results obtaned n studes of the causal status hypothess mght be best explaned n terms of explanatory goodness. In one such study (Experment 4), Ahn, Km, Lassalne, and Denns (2000) gave partcpants a standard (a sngle sentence descrbng a causal relatonshp between a cause and two propertes treated as effects) and two alternatves. Partcpants were nstructed to choose whch of the two alternatves should be categorzed wth the standard. One standard read: Ths object has a hgh-ntensty lght bulb and a pouch that can contan lqud because t was desgned to kll bugs. The frst alternatve shared the cause, but not the effects (.e., Ths object has a sweet, smelly patch and an x-ray generator because t was desgned to kll bugs. ); the other alternatve shared the effects, but not the cause (.e., Ths object has a hgh-ntensty lght bulb and a pouch that can contan lqud because t was desgned to be used n a photograph studo. ). Ahn, et al. predcted that people would prefer to match an alternatve to the standard on the bass of the shared cause, rather than the shared effects. Although the overall results confrmed the predcton, nspecton of the results reveals that only half the tems conformed to the predcton. Table 1 292

shows the materals used n the study and, for each tem, the percentage of choces for the matchng cause. Table 1: Experment 4 standards, from Ahn, et al., (2000). Standard % Ths object has a hgh-ntensty lght bulb and a pouch that can contan lqud because t was desgned to kll bugs. Ths sculpture s made of metal and conssts of sx cubes stacked up because the sculptor ntended t to symbolze polluton. Ths object has a rubber platform and vbrates smoothly because t was desgned to relax pregnant mares durng labor. Ths plant has needle leaves, and produces tny pnk flowers n the sprng because t has a DNA structure called Valva. Ths anmal has a block-shaped head, s red, and has 13 teeth because ths anmal has a genetc code, XB12. Ths pantng has four pllars and s red because the panter ntended to draw a dog. Whle there s a notceable causal status effect for three of the tems, two tems exhbt chance respondng, and one even suggests a preference for the shared effect alternatve. We suggest that the tem varablty can best be understood n terms of the explanatory role of the causal property beng asserted for each tem. One hallmark of a good functonal explanaton s that the ntended functon s causally connected to the effects to be explaned. Ideally, the facts can then be understood as subsumed under a general causal law (Lombrozo & Carey, 2006). In addton, good explanatons exhbt breadth the extent to whch an explanaton accounts for most f not all of the avalable facts (Pennngton & Haste, 1988; Read & Marcus-Newhall, 1993; Thagard, 1989); and depth the extent to whch the local explanaton fts wthn a larger explanatory framework, and can tself be explaned (Read & Marcus- Newhall, 1993; Thagard, 1989). These consderatons can shed lght on why some tems n Ahn et al. s study conformed to ther predcton, whle others dd not. Consder the tem wth the strongest causal status effect, the frst tem ( Ths object has a hgh-ntensty lght bulb and a pouch that can contan lqud because t was desgned to kll bugs. ). In ths case, a clear, general causal relatonshp lnks the ntenton wth the effects: t s generally known that bugs are attracted to lght, and that 80 72 60 52 52 40 bug-kllng agents typcally take the form of lqud pestcdes. Second, the explanaton s broad (ntenton to kll bugs explans every stated effect) and deep (the ntenton to kll bugs fts wthn a larger explanatory framework: bugs are generally annoyng). Now, consder the worst-performng tem: Ths pantng has four pllars and s red because the panter ntended to draw a dog. The causal relaton between four pllars and the ntenton to draw a dog s extremely weak, and there s no clear explanatory framework that would lnk the ntenton to pant a dog wth any of the facts. Of course, these are after-the-fact suggestons. To make ths account plausble, what s needed s a manpulaton of explanatory goodness. In the current study, we test the hypothess that the magntude of the causal status effect depends on explanatory qualty: we predct that the better the explanaton n whch causal nformaton s embedded, the stronger the causal status effect. Experment To test the hypothess that the magntude of the causal status effect depends on explanatory qualty, we manpulated the qualty of explanatons n a two-alternatve forced choce categorzaton task, as n Ahn, et al. s (2000) study. Thus, partcpants had to choose whch of two alternatves should be categorzed wth a gven standard. As a manpulaton check, we also collected judgments of explanatory qualty for fact-explanaton pars: pars of sentences that descrbe facts and a potental explanaton for those facts. The sentences were smple and clearly nterpretable, patterned after the materals used n Ahn, et al., (2000). Method Partcpants Twenty undergraduate students from Northwestern Unversty partcpated for partal credt n an ntroductory psychology course. Desgn, materals, and procedure The desgn was a 2 2 mxed factoral wth explanatory qualty (good or poor), manpulated wthn-subjects and task order (categorzaton task frst or second), manpulated as a between-subjects counterbalancng varable. The materals conssted of 24 stmulus sets, each based on a standard fact-explanaton par, whch comprsed a set of facts joned wth a potental explanaton for those facts. In the explanaton-ratng task the facts were presented as short sentences descrbng a partcular stuaton, whch partcpants were nstructed to assume as true; and the explanaton was presented as a separate sentence descrbng a potental explanatory account of those facts. For example, one set of facts read: Ths object has a hgh-ntensty lght bulb and a pouch that can contan lqud ; and the assocated explanaton read: Ths object was desgned to kll bugs. For the categorzaton task, the facts and explanaton were combned nto a sngle sentence, followng Ahn, et al. (2000)(e.g., Ths object has a hgh-ntensty lght bulb and a pouch that can contan lqud because t was 293

desgned to kll bugs. ). Of the 24 standards, 16 were test tems, and 8 were fllers. The fllers were constructed to mnmze the lkelhood of demand characterstcs n the categorzaton task (and wll be descrbed n greater detal n the context of that task). Of the 16 test tems, 6 standards were taken wthout modfcaton from the materals used n the Ahn, et al. (2000) study descrbed above. In addton to the sx standards from Ahn et al. s study, we constructed an addtonal ten tems, fve of whch were desgned to express good explanatons; and fve to express poor explanatons. The good-explanaton tems were desgned so that the explanaton specfed a general and plausble causal relaton that accounted for all of the stated facts (as verfed n the explanaton-ratng task descrbed below). (See Table 2 for sample tems). These addtonal ten standards were adapted from materals used n studes of category-based nference by Sloman (1994) and by Patalano, Chn-Parker, and Ross (2006). The experment conssted of two paper-and-pencl tasks, counterbalanced for order across partcpants: evaluatng explanatory qualty and carryng out a two-alternatve forced-choce categorzaton task. For the explanatonratng task, partcpants were gven all 24 standards, each arranged on a separate page of the test booklet. The facts were presented at the top of each page n a separate sentence, wth the canddate explanaton presented mmedately below the facts. Partcpants were nstructed to read the facts and the explanaton, and then to provde a ratng of how good or satsfyng they found the explanaton to be on a scale from 1 (very poor) to 7 (very good). They were then nstructed to brefly explan ther response. The categorzaton task was dentcal to that of Experment 4 n Ahn, et al. (2000) as descrbed earler. Specfcally, t was a forced-choce task between two alternatves, n whch partcpants were nstructed to choose the one alternatve that should be categorzed wth the target. Each trad a standard together wth both alternatves was presented on a separate page of the test booklet, wth the standard at the top of the page, and the alternatves below t. For the test tems, one alternatve shared the cause wth the standard, but not the effects; the other alternatve shared the effects, but not the cause. Sample tems are presented n Table 2. The eght fller tems were constructed to vary the pattern so as to mnmze the possblty of task demands. To ths end, for four fllers, the alternatves both shared the cause wth the standard, but dffered on the effects; and for the remanng four fllers, both alternatves shared the effects, but dffered on the cause. For each partcpant, tem order was randomzed n both tasks; for the forced choce task, left-rght presentaton of alternatves was also randomzed. Partcpants completed the tasks at ther own pace: the categorzaton task requred roughly 15 mnutes to complete; the explanaton-ratng task, 25 mnutes. Results We submtted the proporton of responses for matchng on the bass of a shared cause to a 2 2 analyss of varance (ANOVA) for a mxed-factoral desgn. There was no relable nteracton (p =.28), nor any relable man effect of task order (p =.25): subsequent analyses are collapsed over ths varable. Table 2: Sample tems from the study. Standard Shared Cause Shared Effect Ths person makes a good salesperson and persuasve publc speaker, because she used to be a lawyer. (GOOD) Ths person s estranged from hs chldren and has few close frends, because he s a travelng salesman. (GOOD) Ths man has a sxfgure ncome and works out regularly, because he was homeschooled as a chld. (POOR) Ths person s a baseball fan and lkes to read poltcal magaznes, because he s a nght watchman. (POOR) Ths person has a strong vocabulary and several enemes, because she used to be a lawyer. Ths person s outgong and has many busness contacts, because he s a travelng salesman. Ths man has been dvorced twce and s an avd movegoer, because he was homeschooled as a chld. Ths person lkes to play racquetball and watch documentares, because he s a nght watchman. Ths person makes a good salesperson and persuasve publc speaker, because she lkes to please people. Ths person s estranged from hs chldren and has few frends, because he has a drnkng problem. Ths man has a sxfgure ncome and works out regularly, because he was the eldest of three sblngs. Ths person s a baseball fan and lkes to read poltcal magaznes, because he s an actor. In support of our man clam that the sze of the causal status effect depends on the goodness of the explanaton n whch the causal nformaton s embedded we found a relable man effect of explanatory goodness, F(1, 18) = 19.68, p <.001. Planned comparsons reveal that the good explanatons (M =.77, SD =.19) ndeed exhbted a stronger causal status effect than the poor explanatons (M =.52, SD =.24), t(19) = 4.41, p <.001, d = 1.13. That s, partcpants preferred to match on the bass of a shared cause to a much greater extent f that nformaton was part of a good explanaton. The explanatory goodness ratngs were n accord wth the ntended desgn: When partcpants rated the goodness of the explanatons, the a pror good explanatons were gven hgher goodness ratngs (M = 4.46, SD =.74) than those desgned to be poor explanatons (M = 2.86, SD =.75), F(1, 18) = 425.86, p <.001, d = 2.31. (We have snce 294

confrmed ths fndng wth a separate group of partcpants who only rated the qualty of the explanatons). Dscusson The results of ths study provde support for the man clam of ths paper: that the causal status effect derves from the role that causal nformaton plays n an explanaton; and specfcally, that the magntude of the causal status effect depends on the goodness of the supportng explanaton. Ths fndng adds support to the poston that explanatory structure s mportant n categorzaton (e.g., Kel, 1989; Lombrozo, 2006, 2007a; Murphy & Medn, 1985; Rps, 1989). Accordng to ths explanaton-based perspectve, categorzaton s treated as an nference to the best explanaton. Ths perspectve suggests that a prmary research focus should be on nvestgatng propertes that characterze the qualty of explanatons, and the role that those propertes play n cogntve tasks, such as categorzaton. A possble objecton to our study would be that we smply constructed explanatons that were nconsstent wth partcpants pror knowledge. Ahn et al., (2000) found that the causal status effect can be elmnated f the stated causal relatonshp contradcts background knowledge. However, ths consderaton does not apply here, because we were careful to ensure that the poor explanatons n our study dd not contradct known facts. Rather, the poor explanatons were desgned to lack explanatory goodness. In some cases, there was no general causal connecton between the facts and the explanaton; n others, the explanaton faled to explan all of the facts; and n others, there was more than one avalable explanaton for the facts. Conclusons The explanatory approach s compatble wth, but broader than, the recent hghly nfluental approaches that have focused strongly on causal knowledge such as the causal status hypothess (e.g., Ahn, 1998) and causal model theory (e.g., Rehder & Haste, 2001). Causalty-based theores tend to focus chefly on characterzng causal knowledge and the causal reasonng processes that operate over t to support classfcaton decsons. In contrast, the explanaton-based approach focuses on characterzng the structure and qualty of explanatons and how they nfluence category processng. The explanatory approach often nvolves varous knds of causal structures, but t can also draw on other knds of hgher-order explanatory structures that do not rely on causalty (e.g., nformaton about perceptual or mathematcal concepts, such as symmetry or ntegral doman)(wlson & Kel, 1998). However, because causal nformaton s crtcal for many knds of mechanstc and functonal explanatons, the two approaches may turn out to be fellow travelers for most purposes. (See Lombrozo & Carey, 2006, for a recent proposal along these lnes the Explanaton for Export hypothess whch assgns a crtcal role to the nature of the causal relatonshp bndng a cause to an effect). At ths pont, t s far to ask whether our efforts to explcate explanatory goodness are overkll. Arguably, a drect focus on causal knowledge would be more parsmonous and would render the problem more tractable. For several reasons, we beleve that a more drect focus on explanatory knowledge s needed. Frst, and most mportantly, as Lombrozo (2006) notes, there are factors related to explanatory goodness that cannot be reduced to propertes of causal knowledge, but whch may be psychologcally relevant for categorzaton. Second, although t mght seem that a drect focus on causal knowledge would render the problem of centralty more tractable, recent theoretcal and emprcal work suggests that a consderaton of explanatory goodness and explanatonbased processes helps to focus research on the rght questons (e.g., Lombrozo, 2006, 2007a; Lombrozo & Carey, 2006; Thagard, 1989; Pennngton & Haste, 1988). For example, usng ther nfluental explanaton-based model of juror decson-makng, Pennngton and Haste (1988) have shown that jurors arrve at a decson to acqut or convct a defendant on the bass of the most coherent explanaton constructed for the evdence presented at tral. The success of ther model demonstrates how a complex reasonng process can be proftably understood n explanaton-based terms. A thrd pont s that a focus on categorzaton as nference to the best explanaton may brng together areas of research that have been typcally treated as separate. For example, t may be useful to thnk of analogy as another knd of nference to the best explanaton. An analogy typcally nvolves a mappng n whch nferences are mported from a well-understood doman (the base) to one that s more poorly understood (the target)(gentner, 1983). People ntutvely prefer mappngs that yeld a good explanaton, as evdenced by a tact preference for systematcty that s, a preference for depth and hgher-order constranng relatons n the resultng relatonal structure (e.g., Clement & Gentner, 1991). Though there have been arguments for a connecton between analogy and categorzaton (e.g., Kuehne, et al., 2000; Ramscar & Pan, 1996), the relatonshp between them has not receved much attenton. Focusng more explctly on the nature of explanatory processes may help to draw out the mportant commonaltes, as well as meanngful dfferences, between categorzaton and other explanatory processes. In summary, the results of our study offer evdence that the causal status effect depends on the qualty of the explanaton n whch the causal nformaton s embedded. These fndngs nvte further exploraton of the propertes of explanaton that nfluence categorzaton. Do smpler explanatons (.e., those that appeal to fewer causes; Lombrozo & Carey, 2006) lead to a stronger causal status effect than explanatons that are more broad (.e., account for more facts), or deeper? How mportant s depth of explanaton n moderatng the causal status effect? Do the same propertes of explanatory goodness operate across a varety of cogntve processes (such as categorzaton and analogy), or to the same extent? Fnally, focusng on explanatory structure may help to shed lght on long- 295

standng debates over the specfc role that background knowledge plays n categorzaton. Acknowledgments Ths research was supported by ONR Grant N00014-02-1-0078. We thank the Smlarty and Cognton group for ther valuable nput. References Ahn, W. (1998). Why are dfferent features central for natural knds and artfacts? Cognton, 69, 135-178. Ahn, W., & Km, N. S. (2000). The causal status effect n categorzaton: An overvew. In D. L. Medn (Ed.), The psychology of learnng and motvaton (pp. 23 65). San Dego, CA: Academc Press. Ahn, W., Km, N. S., Lassalne, M. E., & Denns, M. J. (2000). Causal status as a determnant of featre centralty, Cogntve Psychology, 41, 361-416. Carey, S. (1985). Conceptual change n chldhood. Cambrdge, MA: Plenum. Clement, C. A., & Gentner, D. (1991). Systematcty as a selecton constrant n analogcal mappng. Cogntve Scence, 15, 89-132. Forbus, K. D., & Gentner, D. (1986). Causal reasonng about quanttes. Proceedngs of the Eghth Annual Meetng of the Cogntve Scence Socety, 196-207. Gentner, D. (1983). Structure-mappng: A theoretcal framework for analogy. Cogntve Scence, 7, 155-170. Kel, F. (1989). Concepts, Knds, and Cogntve Development. Cambrdge, MA: MIT Press. Kuehne, S. E., Forbus, K. D., Gentner, D., & Qunn, B. (2000). SEQL - Category learnng as progressve abstracton usng structure mappng. Proceedngs of the 22nd Annual Conference of the Cogntve Scence Socety. Lombrozo, T. (2007a). Smplcty and probablty n causal explanaton. Cogntve Psychology, 55, 232-257. Lombrozo (2007b). The effects of mechanstc and functonal explanatons on categorzaton. Proceedngs of the 29th Annual Conference of the Cogntve Scence Socety. Lombrozo, T. (2006). The structure and functon of explanatons. Trends n Cogntve Scences, 10, 464-470. Lombrozo, T. & Carey, S. (2006). Functonal explanaton and the functon of explanaton. Cognton, 99, 167-204. Marsh, J. K., & Ahn, W. (2006). The role of causal status versus nter-feature lnks n feature weghtng. Proceedngs of the 28th Annual Conference of the Cogntve Scence Socety. Medn, D. L. & Shoben, E. J. (1988). Context and structure n conceptual combnaton. Cogntve Psychology, 20, 158-190. Medn, D. L., & Ortony, A. (1989). Psychologcal essentalsm. In S.Vosnadou & A. Ortony (Eds.), Smlarty and analogcal reasonng (pp. 179-196). New York, NY: Cambrdge Unversty Press. Murphy, G. L., & Medn, D. L. (1985). The role of theores n conceptual coherence. Psychologcal Revew, 92, 289-316. Patalano, A. L., Chn-Parker, S., & Ross, B. H. (2006). The mportance of beng coherent: Category coherence, crossclassfcaton, and reasonng. Journal of Memory and Language, 54, 407-424. Pennngton, N. and Haste, R. (1988). Explanaton-based decson makng: The effects of memory structure on judgment. Journal of Expermental Psychology: Learnng, Memory, and Cognton, 14, 521-533. Ramscar, M., & Pan, H. (1996). Can a real dstncton be made between cogntve theores of analogy and categorzaton? Proceedngs of the 18th Annual Conference of the Cogntve Scence Socety. Read, S. J., & Marcus-Newhall, A. R. (1993). Explanatory coherence n socal explanatons: A parallel dstrbuted processng account. Journal of Personalty and Socal Psychology, 65, 429-447. Rehder, B. & Burnett, R. (2005). Feature nference and the causal structure of categores. Cogntve Psychology, 50, 264-314. Rehder, B. & Haste, R. (2001). Causal knowledge and categores: The effects of causal belefs on categorzaton, nducton, and smlarty. Journal of Expermental Psychology: General, 130, 323-360. Rps, L. J. (1989). Smlarty, typcalty, and categorzaton. In S. Vosnadou & A. Ortony (Eds.), Smlarty and analogcal reasonng (pp. 21-59). New York: Cambrdge Unversty Press. Rosch, E., & Mervs, C. B. (1975). Famly resemblances: studes n the nternal structure of categores. Cogntve Psychology, 7, 573-605. Rottman, B. M., & Gentner, D. (2006). Expertse effects on sortng strateges of causal phenomena. Proceedngs of the 28th Annual Meetng of the Cogntve Scence Socety. Salmon, W. (1989) Four Decades of Scentfc Explanaton. Unversty of Mnnesota Press Sloman, S. A. (1994). When explanatons compete: The role of explanatory coherence on judgments of lkelhood. Cognton, 52, 1-21. Thagard, P. (1989). Explanatory coherence. Behavoral and Bran Scences, 12, 435-467. Tversky, A. (1977). Features of smlarty. Psychologcal Revew, 84, 327-352. Wlson, R.A. and Kel, F.C. (1998). The shadows and shallows of explanaton. Mnds and Machnes, 8, 137-159. 296