How Automatic Are Social Judgments? - Uleman Lab

JournaJ of Personality and Social ftychology1985, Vol. 49, No. 4, 904-917

Copyright 1985 by the American Psychological Association, Inc.002M514/85/S00.75

How Automatic Are Social Judgments?

Laraine Winter, James S. Uleman, and Cathryn CunniffNew York University

Do people infer personality dispositions automatically when they encode behavior?TUlving's encoding-specificity paradigm was adapted to test three operational in-dicants of automatism: absence of intention, of interference from other mentalactivity, and of awareness. Recruited for a digit-recall study, subjects read sentencesdescribing actions during the retention interval of either an easy or a difficult digitrecall task. Later, sentence recall was cued by (a) disposition cues, (b) strong semanticassociates to the sentence actor, or (c) words representing the gist of the sentence,or (d) sentence recall was not cued. Awareness was measured immediately afterthe last sentence was read. Disposition-cued recall was higher than (b) or (d) andwas unaffected by digit recall difficulty. Awareness of making dispositional inferenceswas only weakly correlated with disposition-cued recall. Results suggest that dis-position inferences occurred at encoding, without intention, without interferenceby differential drain on processing capacity, and with little awareness. Thus, makingdispositional inferences seems to be largely, but not entirely, automatic.

Although inferences about people form thecore of attributions and social cognition, socialpsychologists have only recently begun to in-quire into how general or frequent these in-ferences actually are. Do they occur sponta-neously and effortlessly, or must they be spe-cially motivated? To what extent are peopleaware of making social inferences? The liter-ature on person perception and social judg-ment contains contradictory assumptionsabout the cognitive nature of social inferencesin general. One position presents such judg-ments as conscious and deliberate, the otheras swift, sure, and even automatic.

For example, seminal theorists in personperception assumed the ease, even the ines-capability, of such social inferences as impres-sion formation and evaluation and assertedtheir ubiquity and centrality in everyday psy-

This research was supported by National Institute ofMental Health Grant MH 08573 to the first author.

We wish to thank John Bargh and David Doifman for

their suggestions on the design of the study; Robert L.Simpson for his help in preparing slides: and Susan T. fiste,E. Tory Higgins, Richard ShiBHn, and one anonymousreviewer for their helpful criticisms of earlier drafts of themanuscript.

Requests for reprints should be sent to James S. Uleman,Department of Psychology, New York University, 6 Wash-ington Place, Room 753, New York, New \brk 10003.

chological functioning (Asch, 1946; Heider,1944; Tagiuri, 1958). Some recent approachesto social inference have argued for the spon-taneity of attributional thinking (Smith &Miller, 1983; Winter & Uleman, 1984).

In contrast, others imply or assert that socialinferences are deliberate and effortful opera-tions, requiring special motivating circum-stances. Langer (1978) has described this viewas picturing the individual "primarily as [an]information processor w h o . . . is cognitivelyaware most of the time and who consciously,constantly, and systematically applies rules toincoming information" (p. 35). KeDey andMichela (1980) discussed "motivation to makeattributions" and asserted that "a person's in-terests . . . determine when he will becomemotivated to make attributions at all" (p. 473).Enzel and Schopflocher (1978) have criticizedattribution research for using methods that in-stigate attributions that might not occur in reallife, thereby inflating impressions of how com-mon they are. Berscheid. Graaano, Monson,and Dermer (1976) asked, "Why is the per-ceiver sometimes willing to spend his time andenergy in an effort to arrive at a causal under-standing of another's behavior?" (p. 979). Also,Pyszczynski and Greenberg (1981) noted that"current conceptualizations . . . imply ahighly rational inferential process by which the

904

SOCIAL JUDGMENTS 905

attributor systematically samples and weightspotentially relevant information before arriv-ing at an explanation for an event" (p. 31).

Surely social judgments may be consciousand occupy attention. However, a model of or-dinary social judgments has developed thatimplicitly assumes that they usually occupyattention, require eifort, and give rise to con-scious awareness (at least of their final prod-ucts). We suggest that this model exaggerateshow intentional, effortful, and conscious at-tributional processes are.

Automatic Processes

The topic of attentional or conscious pro-cesses has itself been the focus of recent re-search in information processing. Cognitiveprocesses may be conceptualized along a con-tinuum of attentional requirements. Thoseoperations requiring substantial processingcapacity are called conscious strategies (Posner& Snyder, 1975) or controlled processes(Schneider & Shiffrin, 1977; Shiffrin &Schneider, 1977). Although researchers havesomewhat different characterizations of au-tomatic processes, three central criteria pre-sented by Posner and Snyder are representedin each treatment. Such processes occur with-out intention, without giving rise to awareness,and without interfering with other ongoingmental activity. They also identified consciousprocessing with a limited capacity mechanismthat may be directed toward different types ofactivity. In their view, most cognitive taskscombine automatic activation with consciousstrategies. Schneider and Shiffrin (1977; Shif-frin & Schneider, 1977) have presented a uni-fied model for selective attention, short-termsearch, and the detection and recognition oftargets. Logan (1979) has studied the devel-opment of automatism using a concurrentmemory load technique.

A few social cognition researchers have usedthe terms automatic and passive to character-ize the psychological processes under their in-vestigation (Higgins & King, 1981; Langer,1978; Taylor, Crocker, Fiske, Sprinzen, &Winkler, 1979). Bargh and his associates haveprovided the most direct demonstrations ofautomatic social information processing, usingauditory and visual stimuli of which their sub-jects were unaware (Bargh, 1982,1984; Bargh,

Bond, Lombardi, & Tola, 1985; Bargh & Pie-tromonaco, 1982; Bargh & Thein, in press).

This suggests that social inferences based onstimuli of which people are aware may alsosatisfy one or more automatism criteria. Inparticular, people may make such inferenceswithout intentions to do so, without specialmotivating conditions, and without awarenessof doing so or of the cognitive results. Tworecent studies by Winter and Uleman (1984)indicate that this may be the case. The presentresearch involves a closer examination of theseencoding-specific inferences and their inten-tionality and awareness and a test of whetherthese inferences are interfered with by con-current processing demands.

Spontaneous Encoding-Specific Inferences

The present research used a theoreticalframework and a design developed by Winter(1980; Winter & Uleman, 1984). Based on thenotion of episodic memory organization(Tulving, 1972) and an adaptation of Tulving'sencoding specificity paradigm (Thomson &Tulving, 1970; Tulving & Osier, 1968; Tulving& Pearlstone, 1966; Tulving & Thomson,1973), that research indicated that inferencesabout personality can be part and parcel of theencoding of behavioral information, engagedin spontaneously (i.e., without instructions orother unusual motivating conditions), ratherthan discrete operations dependent on infor-mation retrieval.

The encoding specificity principle postulatesthe time of an input's encoding as the primarydeterminant of its storage format and retriev-ability (Tulving & Thomson, 1973, p. 369).Any information that is encoded at the sametime as the to-be-recalled input should be aneffective retrieval cue. Tulving's subjects stud-ied lists of target words, like CHAIR, that werepaired with weak semantic associates, like glue,and later were asked to recall the target wordsin the presence of either the input cue (glue),a strong semantic associate of the target (e.g.,table), or no cue. Recall was best when theinput cue was present.

If people make trait inferences when theyobserve behavior, those inferred traits shouldbe stored with the behavioral information.Therefore, as part of the encoding context ofthe behavioral information, the attributed trait

906 L. WINTER, J. S. ULEMAN, AND C. CUNNIFF

itself should serve as a self-generated input cue.One can explore whether trait inferences occurat encoding by investigating the retrieval cueeffectiveness, for the behavioral information,of subjects' most likely trait inferences. If uponreading "The accountant takes the orphans tothe circus" subjects infer that the accountantis kindhearted, the word kindhearted shouldbe an especially effective retrieval cue for thesentence. Disposition words' effectiveness asrecall cues was established by comparison withnoncued recall and also by comparison withthe effectiveness of strong semantic associatesto important sentence parts (e.g., numbers, astrong associate to ACCOUNTANT, and fun, astrong associate to TAKES TO THE CIRCUS).These semantic associates were included as acontrol for the possibility that the dispositioncues' effectiveness might be due to a priori se-mantic associations to sentence words ratherthan to dispositional inferences.

Thus, under only general memory instruc-tions, subjects in the previous research (Winter& Uleman, 1984) were shown sentences de-scribing human actions, each implying a par-ticular personality trait. Later they recalled thesentences, cued by either the personality wordsrepresenting a probable trait inference (kind-hearted), a strong semantic associate to an im-portant sentence word, or no cue. In the firstexperiment, semantic associates were to actors(ACCOVtfTMlT-numbers); in the second, theywere to verbs or verb phrases (TAKES TO THEClRCUS-fun). Results showed the dispositioncues to be strongest, nonsignificantly strongerthan the semantic associates to actors, signif-icantly stronger than semantic verb associates,and much stronger than no cue.

These findings suggested that as each sen-tence was read, a social judgment was madespontaneously and stored in relation to thesentence information. Thus, at recall the per-sonality word proved an effective retrieval cuebecause it promoted access to the memory lo-cation at which the behavioral information wasstored. Correlations between disposition-cuedrecall and self-reports of making personalityjudgments, used as measures of how muchsubjects knew about their own thought pro-cesses, were all nonsignificant. Subjects werenot aware of their own judgments.

Although these studies had not been de-signed as tests of automatism, these findingssuggest that dispositional inferences may be

instigated in the absence of intention andawareness. The issue is compelling becauseautomatic processes possess characteristicsthat strongly argue for their generality and im-portance—characteristics such as unsuppres-sibility and unmodifiability. Hence, the spon-taneous encoding paradigm was refined for thepresent study to address the three major in-dicators of automatism: absence of intention,interference, and awareness.

The Present Study

The fact that Winter and Uleman's (1984)subjects apparently made social inferenceswithout instructions to do so suggests that theinferences were unintentional, but subjectsmay also have made dispositional inferencesas a mnemonic device (even though they didnot report using that strategy later). To ruleout this possibility, we introduced a manipu-lation suggested by Logan's (1979) concurrentmemory load technique. Subjects performedan irrelevant task, digit recall, which demon-strably involves the use of controlled strategiessuch as rehearsal. The sentences were pre-sented as "distractors" in the interval betweenpresentation of the numbers and their recall.Hence, subjects had no intention even to re-member them.

The memory load technique is also usefulfor testing the absence of interference fromother ongoing mental operations. This requiredvarying the demands on processing capacity.Because the sentences were encoded in the re-tention interval of a digit recall task, this couldbe done by varying the difficulty of the digits.Half of the subjects were presented with moredifficult sequences of multiple rather than sin-gle digits. Digit recall and ratings of task dif-ficulty provided manipulation checks thatthese digit conditions produced differentialdemands on processing capacity.

Probing for awareness of making trait in-ferences required a more immediate andstringent test than had been used in Winterand Uleman's (1984) experiments, in whichsubjects' awareness was assessed after sentencerecall. The interval between the process (i.e.,trait inference) and report probe was at least10 min. This gap would allow relevant infor-mation to be forgotten, confounding lack ofawareness with failure to remember. The delaycould also have permitted subjects to fill in


their memories of process with a priori theories(Nisbett & Wilson, 1977). Finally, the aware-ness probe questions were too general, askingsubjects to generalize what they thought acrossall the sentences. To meet Ericsson and Si-mon's (1980) conditions for accurate intro-spective reports about cognitive processes, weelicited reports about a single sentence (the lastin the set) immediately after subjects read itand before they were asked to recall the lastsequence of numbers. Self-report accuracy wasassessed by correlating subjects' self-reportswith actual disposition-cued recall.

The present experiment also introducedanother kind of cue, an associate to the entiresentence. These gist cues summarized eachsentence's action but were neutral with regardto personality.

Overview and Predictions

The experiment was ostensibly a study ofrecall for numbers in which subjects believedthat the sentences served as distractors. Eachof the 16 trials consisted of the presentationof a series of five numbers, shown one at atime on slides, with subjects instructed to readeach aloud. These were followed by a singlesentence, also shown on a slide and read aloudby the subject, and then immediate recall ofthe numbers. Half of the subjects received se-quences of single digits, and the rest receivedsequences of multiple digits.

If subjects make trait inferences while tryingto retain digit sequences, then the inferenceprocess must be unintentional. The hypothesisthat disposition inferences are made withoutintention predicts that recall cued by the dis-position words will be superior to noncued re-call and at least as strong as recall cued byeither the actor associates or the gist cues. Thelatter two comparisons are essential for elim-inating the possibility that the disposition cues'effectiveness is actually due to strong a prioriassociations to sentences.

The prediction that trait inferences do notrequire awareness was tested by correlatingself-reports of making trait judgments aboutthe actors, elicited immediately after the lastsentence was read, with actual disposition-cuedrecall.

The noninterference criterion for automa-tism predicts that memory load should haveno effect on the extent to which subjects

make dispositional inferences. Therefore, eventhough recall of the more difficult numbersshould be worse than the single numbers, sub-jects at both levels of difficulty should showcomparable rates of sentence recall cued bydisposition words, because making a trait in-ference automatically should require little orno processing capacity. In effect, an interactionbetween digit recall and sentence recall waspredicted, with digit recall declining in themore difficult condition and sentence recallremaining about the same.

Method

The present research comprised two parts: (a) pretests

of the sentences and digit sequences and (b) the recall ex-periment.

Pretests

Pretest 1 identified sentences in a set of 28 that mostreliably produced dispositional rather than situational at-

tributions. Thirty-seven introductory psychology studentsdescribed "what probably caused the event" in each. Aresponse was scored 1 if it was totally dispositional and 3if it was totally situational.1 The 18 sentences that were

most dispositional were chosen. The mean attribution rat-ing was 1.63. The range was from 1.25 to 2.23. Two sen-tences were eliminated later on the basis of Pretest 4.

Pretest 2 obtained the personality traits attributed mostfrequently to sentence actors. Thirty-two additional intro-ductory psychology students answered the question "Whatkind of person is this?" with as many as three words for

each of the 18 sentences. The most frequently given wordwas selected as the dispositional cue for that sentence.

Pretest 3 obtained gist cues. Sixty introductory psy-chology students were asked to "think of a word (for eachof 18 sentences) which expresses the kind of activity thatthe sentence exemplifies. You might try to think in termsof a good title for a paragraph in which the sentence ap-pears." The most frequently given word for each sentencewas selected as the gist cue. Table 1 presents the 16 sen-

tences finally selected and their gist cues, as well as thedispositional cues and the semantic associates to actors.

Pretest 4 estimated the associative strength between cuesand sentences in a paired associates test in order to matchthe disposition and gist cues' associative strengths with the

sentences. Pairs consisted of sentences (the targets) andeither disposition cues or the gist cues derived inPretest 3. Tested individually, 31 undergraduates heard acue (e.g., helpful or assisting) on a cassette tape recorder,followed 5 s later by the sentence to which it pertained

1 Even though Miller, Smith, and Uleman (1981) showedthat dispositional versus situational is not a single contin-uum, it served our purpose here, which was to distinguishdispositional from nondispositional attributions. A similarprocedure in Winter and Uleman's (1984) Pretest 2 yieldedan interjudge reliability of .89.

(text continues on page 910)


•H•Si

|6•~

73

S

1

)le 1

Hen

ce S

tim

uli

, R

eca

ll C

£$

J

'

I

1

U

\

1*

I!!

a

at,

S £

Hi

3u

118f- - a

u

Sen

tence

fN

O«

OO

"8

12

'-O

tN

r-

'*3g

Enjo

yab

le i

rs

|E

1

The

acco

unta

nt

takes

-

. :OS VI < N ~ - O O — t^ *t *S)

O <N O — ' — « O ^ O ^

S £ 9 3 S 3 $ S 8c - i f N <N cj ( N csi o i r v i f s

[-. (N — ' C-4 V*i (N

* * & 1I >. I s . * " "& i B ». §• § - g ' 1 1l a I I J 1 1 'E 1

« ^ i O ^ r f ( N f T ) ' * t w ^ ^( N ( N f N * ^ ( N < S ( N ( N ( N

. , , . -- , - . ,

g

c ^ S

i d I I I s s 1 1

— o o r ~ - r o ( N — ^ O > * ~ >—. — ( (N *N •'T

•I

ii|. .B ^ :1 s . G

I I I s 1 1 i l l

1 ili^llil! 11 if!1!!.

r j m ^ 10 *& t** o d o > O


<N r-i

•5b

'a<

1

I 0

b vi2 o."s 5

'ssi VIS o.


("The tailor carries the old woman's groceries across thestreet"). About 5 s later the next cue was heard. Subjects

were instructed to learn each cue-sentence pair in prep-aration for a recall test in a single pass through the set ofcue-sentence pairs.

Sentences were presented in the same order to all sub-jects. Half of them heard the even-numbered sentencespreceded by the disposition cues, and half heard the odd-numbered sentences. During recall, cues were presentedin a random order with a dilferent random order on eachtrial and unlimited time after each cue. When a subjectcould not remember the sentence, it was read aloud. Thesession terminated when the subject had recalled each sen-tence predicate at least twice consecutively. The trials tothis criterion for each sentence were averaged over all sub-jects and are presented in Table 1. Differences betweendisposition and gist cues' strength of association with eachsentence were assessed via r tests (see Table 1). For sixsentences, the gist cue was significantly stronger; for onesentence the disposition cue was stronger. Thus, we couldnot eliminate all sentences with unmatched cues, but sta-tistical adjustments were planned, taking this discrepancyinto account.

Pretest 5 established that recall of the two sets of digitsdiffered in their demands on processing capacity. We alsotested the credibility of verbal distractors inserted between

the digit sequence. On each of eight trials, individuallytested subjects heard a sequence of five to-be-rememberedtape-recorded numbers, followed by three unrelated im-agible nouns (from Winters & Brzoska, 1975). They re-peated each number and noun aloud and recalled thenumber sequence immediately after the last noun. Half of

the subjects received a series of five single digits, and halfreceived a sequence of three double and two triple digits.Two min after the last digit sequence was recalled, recallfor the words was requested. Of the 40 numbers, subjectsin the single-digit condition recalled 37.38, or 93.45%, andmultiple-digit subjects recalled 14.33, or 35.83%, ((26) =22.25, p < .001. Recall of the 24 nouns did not differ (3.92vs. 3.20, respectively). All subjects found the digit recalltask perfectly plausible and assumed that the nouns weredistractors. None were suspicious of the task or reportedtrying to remember the nouns, and virtually all expressedsurprise (and dismay) when asked to recall the nouns.

Recall Experiment

Subjects. Ninety-five volunteers from New York Uni-versity's Introduction to Psychology course signed up foran experiment called "Memory for Digits" in partial ful-fillment of course requirements.

Materials. Sixteen sentences were selected on the basisof the pretests described earlier. In each sentence, an actordenoted by an occupation (e.g., carpenter) performs anaction that is described without referring to internal stateslike intention or affect. Digit sequences from Pretest 5,sentences, and the anagrams used in the distractor taskwere shown on slides projected by a Kodak Carousel SlideProjector. Recall sheets given to subjects after the distractortask presented recall instructions, the list of retrieval cues,and space in which to write recalled sentences.

Procedure. Subjects were run individually and wereinformed that they were participating in a study of memoryfor digits. Written instructions described the procedure up

to the point of sentence recall. On each of 16 trials, asequence of five numbers was shown. After the subjectread each number aloud, the projector advanced to thenext one. Following the fifth number, a single sentence wasshown. The subject read the sentence aloud and then re-peated it, looking away from the screen. Then the projector

advanced to a blank slide, and the subject recalled the digitsaloud. Ten s were allotted for recalling each number se-

quence.All 16 trials were conducted this way except the last. As

soon as the last sentence was spoken twice, the subject wasasked to report what he or she thought about while readingit and then any thoughts about the actor or the action inthe sentence. The subject then rated how much he or she

had thought about visual images, word associations, whocaused the event in the last sentence, and the actor's per-sonality or personal qualities on 11-point scales. Recall forthe last number sequence was then elicited, and a 2-min

anagram distractor task was then performed. Twelve an-agrams were presented, three on a slide, with a maximumof 1 min for each slide. Finally, the lights were turned upand the recall sheet was given to the subject, who wrotedown all the sentences or parts of sentences that he or shecould recall, using the cues provided. After 10 min, therecall sheet was collected and the subject was debriefedand thanked.

Results

Because each sentence consisted of an actor,a verb, an object, and a prepositional phrase(or in one case, a second object), recall wasscored by giving up to 4 points per sentence.We used lenient scoring, with no extra creditfor verbatim recall. A second coder scored thefirst 60 recall protocols; interrater agreementwas 98%.

To determine recall cues' relative effective-ness and whether differential memory de-mands affected sentence recall, recall data weresubjected to a split-plot factorial analysis ofvariance (ANOVA). With the four cuing con-ditions, each block of four sentences was cuedby a different type of recall cue for one quarterof the subjects in a Latin square. Thus, eachsentence was cued by each type of cue an equalnumber of times, producing a four-level be-tween-subjects factor called block-cue pairing.Block-cue pairing and memory load were be-tween-subjects factors. Cue type and sentencepart (actor, verb, object, or prepositionalphrase) were within-subjects factors. Thisyielded a 2 X 4 X 2 X 4 X 4 (Memory Load XBlock-Cue Pairing X Sex X Cue Type X Sen-tence Part) design.

This initial ANOVA revealed main effects forcue type, F(3, 258) = 8.77, p < .001, and sen-tence part, F(3, 258) = 8.04, p < .001, and


interactions between cue type and sentencepart, F(9,774) = 23.44, p < .001, and betweenblock-cue pairing and cue type, F[9, 258) =4.30, p < .001. Memory load had no generaleffect on sentence recall.2

Main effects and interactions were evaluatedby tests of simple effects and protected / testsbetween pairs of means. As predicted, dispo-sition cues (M = 0.99) were superior to se-mantic (M = 0.59) and no-cue (M = 0.50)recall, ft(94) = 8.71 and 7.86, ps. < .001. Inaddition, gist-cued recall (M = 0.90) was su-perior to semantic and no-cue recall, fs(94) =3.15, p = .002, and 3.51, p = .001. Dispositionand gist cues did not differ in retrieval effec-tiveness, nor did semantic cues and free recall.Thus, both disposition and gist cues were su-perior to the other cuing conditions.

The main effect for sentence part was dueto better recall of objects (M = 0.84) than ac-tors (M = 0.68), verbs (M = 0.76), and prep-ositional phrases (M = 0.71), /s(94) = 4.02,p = .001; 4.05, p = .001; and 4.57, p = .001,respectively. Verb recall was marginally betterthan actor recall, t(94) = 1.82, p = .07.

Figure 1 shows the Cue Type X SentencePart interaction and the results of significancetests. For actor recall, semantic cues were mosteffective. For verbs, objects, and prepositionalphrases, dispositional and gist cues were mosteffective. Analyses within each cue type showeddifferences among sentence parts for all butnoncued recall.

Analyses of the interaction between block-cue pairing and cue type revealed that dispo-sition cues were particularly effective for thesentences that happened to be in Block 3, F(3,91) = 7.17, p < .001, and gist cues were rela-tively ineffective for these sentences F(3,91) =2.96, p = .04.

Disposition and Gist Cues

In the above analyses, dispositional retrievalcues were nonsignificantly superior to gist cues,and both were generally more effective thanother cue types. In previous research, dispo-sition cues were significantly more effective incuing entire sentence predicates (verb + ob-ject + preposition) than were semantic asso-ciates to either the sentence actors (Winter &Uleman, 1984, Experiment 1) or brief verbphrases (Experiment 2). Furthermore, dispo-

Verb— Sentence

ObjectPart —

Prep.

Figure 1. Recall of sentence parts as a function of cuingconditions. {Differences between means are shown withasterisks: *p <S .05, "p «E .01, and ***p s; .001. Verticalarrows point to significant differences between cuing con-

ditions. Disp. = disposition; sem. = semantic.)

sition cues retrieved entire sentences more ef-fectively than verb semantic associates. Thiswas interpreted as evidence that dispositionswere inferred (at encoding) from the entiresentence, rather than being mere semantic as-sociates to sentence parts. The present study'sgist cues offer some control for dispositioncues' possible advantage in cuing whole sen-tences, because they too were based on wholesentences.

2 There was a significant three-way interaction betweenmemory load, cue type, and sentence part, F(9, 774) =

1.94, p = .044. Separate ANOVAS examined the Cue Type X

Memory Load interactions for each sentence part and theSentence Part X Memory Load interactions for each cuetype. Interactions with memory load were significant only

in the ANOVA for semantic-cued recall, F(3, 261) = 2.90,p = .036.

There were also marked differences in recall across sen-tences, and in the effectiveness of cues across sentences.

To control for the variance due to sentences, an additionalANOVA was performed that included sentences as a 16-level factor. This analysis revealed that sentences did con-tribute significant variance to recall, F(15,1290) = 12.65,p < .001, implying that cued recall effects are not the same

across all sentences. Nevertheless, cue type remained sig-nificant in this analysis, F(3, 258) = 8.75, p < .001. Thesame relations between recall means were obtained in thisanalysis too. Recall cued by disposition words (M = 3.98)and recalled cued by gist words (M = 3.58), though notdiffering significantly from each other, were higher thanrecall cued by either semantic actor associates (M = 2.37),M95) = 3.39. p < .001, and 3.03, p = .003, or no cues(M = 1.89), ;s(95) = 3.69, p < .001, and 3.73, p < .001.Hence, although sentences varied in recallability, cue typeremained a powerful independent source of variance.

912 L, WINTER, J. S. ULEMAN, AND C. CUNNIFF

The above analyses assumed that dispositionand gist cues were equally matched in strengthof association with the sentences. From Pretest5, however, we know that gist cues had strongerassociations to some sentences (see Table 1).Although paired associates trials to criterionor probability of recall and free associationstrengths are not strictly comparable, they areclosely related (Bahrick, 1969; Watkins &Gardiner, 1982). Therefore, we performed arough adjustment of the disposition and gistrecall scores to take into account the differ-ences in these cues' strength of association withsentences from Pretest 5. Disposition-cued re-call scores were weighted by the ratio of dis-position cues' trials to criterion to gist cues'trials to criterion (1.09). Gist-cued recall wasweighted by the inverse.

Using these weighted scores, the differencebetween gist- and disposition-cued recall forboth total sentences and predicates only wasassessed by matched (tests. The adjusted dis-position-cued recall of the total sentence (i.e.,all four parts; M = 1.11) was marginally higherthan that of adjusted gist-cued recall (M =0.86), 1(94) = 1.84, p = .069. For predicatesonly, recall for disposition cues (0.92) was re-liably higher than for gist cues (0.67), 1(94) =2.30, p = .024. Hence, when a priori strengthof association between sentences and cues istaken into account, the superiority of dispo-sition cues emerges as significant in the caseof the predicates only and as marginally sig-nificant for recall of total sentences.

Interference

The effectiveness of the memory load ma-nipulation was assessed by I tests on digit recalland on subjects' ratings of the ease of thenumber-recall task. Difficulty level strongly af-fected the number of digits recalled, as Figure2 shows. Of a possible 80 numbers, subjectsin the single-digit condition recalled 68.89, or86.11%, and those with multiple digits recalledonly 29.21, or 36.51%, t(93) = 26.42, p < .001.Single-digit recall (5.73) was rated easier thanmultiple-digit recall (6.86) on a 10-point scale,/(91) = 3.21,p = .002.

The automatism criterion of noninterfer-ence from controlled processing was addressedby examining the effect of differential short-

single multiple— Digit Condition—

Figure 2. Percentage of sentences (left axis) and digits (right

axis) recalled as a function of digit difficulty, and cue type

for sentences. (Disp. = disposition; sem. = semantic.)

term memory load on dispositions' effective-ness in cuing sentence recall. The initial overallANOVA described above showed that memoryload made no difference in cued or noncuedrecall of sentences, F(l, 86) < 1. Interactionswith both cue type and sentence part were alsoless than 1. Figure 2 depicts cued and noncuedrecall in the two memory load conditions, withthe digit recall superimposed and scaled onthe right-hand ordinate. It shows that differ-ential memory load had no effect on subjects'spontaneous trait inferences, as measured bydisposition-cued recall.3

Note that none of the other three cuing con-ditions showed any effect of the memory loadmanipulation either. This might indicate eitherthat all the association and inference processeswere automatic or that no effective manipu-lation of processing capacity actually tookplace. Unfortunately, the design provides nodirect measure of spare processing capacity,so it is not possible to exclude the latter pos-sibility.

Awareness

Subjects' awareness of making trait infer-ences was assessed immediately after reading

3 This acceptance of the null hypothesis, although un-

usual, is justifiable (Greenwald, 1975). The present com-

parison would seem to possess sufficient power—.66, as-

suming a moderate effect size with N = 95 (Cohen, 1969)—

and a sufficiently high p value (greater than .50) on which

to base speculation that the absence of a difference reflects

the absence of an effect.


the final sentence. On the initial open-endedquestion, only 18 subjects (19%) mentionedanything at all about the actor. Responses tothe closed-ended questions about personalityand causality-related thoughts were made ona scale on which 0 meant that the subject hadhad no such thoughts while reading the finalsentence and 10 indicated that such thoughtshad occurred a great deal. Low means (3.03and 2.67, respectively) suggest that the inci-dence of such conscious thoughts was minimal.More important, however, responses to thesequestions were correlated with subjects' actualdisposition-cued recall, with the recall datafrom the final sentence for each subject deleted.(A split-plot ANOVA and analyses of simple ef-fects were also performed on these abridgedrecall data, with the final sentence deleted,corroborating the original results.) Disposi-tion-cued recall correlated .18 (p = .09) withthe open-ended response and .23 (p = .03) withthe closed-ended question about personality.Thus, there was marginally significant evidenceof some accurate introspective awareness ofmaking dispositional inferences.4

Discussion

The central purpose of this study was to ex-amine the extent to which spontaneous en-coding-specific inferences meet each of threecharacteristics of automatism: occurrencewithout intentions, without interference byother mental activity demanding processingcapacity, and without awareness. Before dis-cussing these, however, it is essential to estab-lish that encoding-specific inferences did occur.

Dispositions were more effective sentenceretrieval cues than were strong a priori se-mantic associates to key sentence words, eventhough the disposition cues' a priori associa-tions to the sentence were weak or absent. Inreplicating previous findings of Winter andUleman (1984), this again suggests that sub-jects made inferences about actors' personal-ities upon reading behavioral descriptions ofactors, thus linking the attributed traits inmemory with the behavioral information onwhich the inferences were based. These newlinks, which pretesting had shown to be eitherweak or absent before the sentences were read,

facilitated recall of sentences by promoting ac-cess to the behavioral information in episodicmemory.

We have emphasized the importance of an"automatic encoding process [that] produces(over a short period of time, perhaps severalhundred milliseconds) a large number of dif-ferent types of informational features, orga-nized by content," including category codesand concept codes (Gillund & Shiffrin, 1984,pp. 58-59). Of course, retrieval processes alsoaffect recall, including inferences made fromretrieval cues. At retrieval, cues narrow thesearch and provide a test for evaluatingwhether the retrieved item had been presentedearlier. If recall depended primarily on suchretrieval processes, though, semantic-cued re-call in our studies would be superior to dis-position-cued recall because of the semanticcues' stronger a priori associations with sen-tence parts (established in pretesting). Thatthey were not points to the importance of en-coding processes.

An analogy may be useful. Imagine lookingfor The Castle (the to-be-retrieved item) in theLibrary of Congress (long-term memory store)with and without the card catalogue system(encoding processes). Retrieval cues (fiction,by F. Kafka) could narrow the search set, de-pending on how the library was organized, andcould confirm that the right book had beenfound. However, the cues would be much moreeffective if they were linked to the book whenit was shelved (encoded) and referred to thebook's address in the library. Without such alink at encoding, one would have to argue thatour dispositional cues are as good as or betterthan the semantic cues at narrowing the searchand testing its results. The pretests challengethe plausibility of this view. Surely inferences

4 We also examined possible sex differences in awareness.Women (3.32) reported more causal thoughts than men(2.00), r(94) = 2.35, p --=- .022, but there was no differencein personality thoughts (3.34 and 2.72, respectively). Cor-relations between disposition-cued recall, deleting the lastsentence, and the open-ended measure were .02 for menand .32 (p = .03) for women. Correlations with thoughtsabout causality were -.17 for men and .25 (p = .10) forwomen; with personality thoughts, they were .13 and 32(p - .03), respectively. Thus, accurate introspective aware-ness was limited to women.


made from recall cues at retrieval do play arole in recall, and in our findings, but it seemsunlikely that they can account for the superi-ority or even equivalence of dispositions as re-call cues in our paradigm.

How well did these encoding-specific traitinferences satisfy the three criteria of auto-matic processes? The criterion of absence ofintention was clearly met without the ambi-guity of previous studies. Subjects believed thatthe study concerned memory for digits and thatthe sentences were merely abstractors. Thus,they had no intent to remember the sentencesand therefore no reason to engage in elabo-rative rehearsal that might generate disposi-tional inferences.

The results also appear to satisfy the crite-rion of absence of interference from ongoingmental operations, because disposition-cuedrecall was unaffected by the level of simulta-neous demands on processing capacity, oper-ationalized by the digit-retention task. Thesedemand levels did seem to differ markedly be-tween conditions according to the digit recalldata and subjects' difficulty ratings.

An alternative interpretation of this pre-dicted absence of effect on sentence recall isthat processing capacity was not effectivelymanipulated. Spare processing capacity wasnot directly measured, and the single-digit se-quences may have placed enough demands onprocessing capacity to create a ceiling effect.Another possibility is that subjects in the mul-tiple condition, perceiving the task's difficultyearly in the procedure, did not attempt to re-member all five numbers in each sequence butcontented themselves with recalling two orthree. These possibilities temper our interpre-tation of these results and compromise thedigit-retention task, by itself, as a completelyconvincing manipulation of processing capac-ity. Studies currently underway are attemptingto address these issues.

The evidence for awareness of making traitinferences at encoding showed, unlike previousstudies, some weak evidence of accurateawareness (but only among women; see foot-note 4). Previous studies' awareness measureswere obtained at least 10 min after the sen-tences had been read, after distractor anagramsand cued recall, rather than immediately afterthe last sentence was read. They also elicited

awareness of inferences about the sentences ingeneral rather than about a particular sentence.Thus, the present evidence indicates thatwhatever awareness exists of making trait in-ferences is relatively fleeting and that, onlyamong women, it accounts for about 10% ofthe variance in the effectiveness of disposition-cued recall.

Overall, our results provide evidence thatmaking one fundamental kind of social judg-ment, inferences about traits, can be initiatedin a largely automatic way. They are certainlymade without intentions while other inten-tional processing occurs. They seem to be atleast difficult to interfere with and are unaf-fected by simultaneous variations in processingdemands and capacity (though this evidenceis open to alternative interpretations), and theyoccur with only weak and fleeting awareness,as shown by a measure that is both more im-mediate and more specific than those usedpreviously. Initiating the process of makingtrait inferences is more automatic than hasbeen demonstrated heretofore.

.It is also important to remember that mostcognitive processes combine automatic andcontrolled processes. Our subjects devotedenough attention to the sentences to read themaloud twice. Our evidence suggests that thesentences, once read aloud and repeated, au-tomatically engaged processes resulting in dis-positional inferences. Such automatically en-gaged dispositional inference processing maythen require some controlled processing for itsconsolidation and storage, even when it is un-intentional. More effective interference meth-ods and measures of cognitive capacity duringsentence encoding will be required to assessthe degree to which controlled processes mayalso be involved in spontaneous dispositionalinferences.

The near-equivalence of the disposition andgist cues in recall raises a question about whatthe gist cues actually represent. Intended assemantic associates to whole sentences, theywere derived from a free-associate style pretest,but they are more than ordinary semantic as-sociates. Many represent the actor's intention(e.g., dieting, giving, welcoming, assisting), it-self a kind of social judgment. It is not clear,then, that the comparison between gist anddisposition cues represents the distinction be-


tween semantic associates and episodicallybased inferences. They may represent twokinds of social judgments.

Another issue for future research is whetherdisposition-cued recall indicates the occur-rence of personality inferences, as hypothe-sized, or just categorizations of actions. Ourtrait terms may characterize actions as muchas, or more than, actors. Srull and Wyer (1979)have shown that specific action descriptionsmay prime such abstract categories as hostileand kind even without instructions to infersuch categories from the descriptions. If dis-position words also served as action categories,like our gist cues, this might imply that per-sonal or disposition terms are simply goodsummary representations of sentences aboutpersons. Our gist cues could not rule out thispossibility, and it may not be possible with thepresent paradigm. However, one might sub-stitute pictures effaces for the common nounsrepresenting actors and measure recognitionaccuracy for faces paired with trait versus gistcues derived from the same actions.

Other interesting directions for future re-search are suggested by other criteria for au-tomatism. For example, automatic processescannot be suppressed, modified, or improvedby instructions (e.g., Shiffrin & Dumais, 1981),and our paradigm is being extended to asksimilar sorts of questions about other socialjudgments, such as causal attributions.

What are the implications of the evidencethat traits are inferred unintentionally, withoutinterference, with little or no awareness, andare linked with the memory representations ofthe behavior on which they were formed?There is a literature that shows that the me-morial status of information can bias subse-quent social judgments as well as memory forevents (Carlston, 1980; Higgins & Rholes,1976; Ostrom, Lingle, Pryor, & Geva, 1980).Our results suggest that such effects may occureven without inference instructions and per-haps even without subjects' awareness that theyhave made inferences. Thus, behavioral evi-dence that implies clear dispositions mayspontaneously affect subsequent judgmentsmore than other, more diffuse or ambiguousevidence, because the inference itself is storedin memory and becomes an additional basisfor subsequent judgments (Carlston's, 1980,

model 3. Id, p. 90). The present paradigm pro-vides a way of testing this possibility directly.

In these models, social inferences serve bothas "themes" (Ostrom et al., 1980) that organizememory and bias recall and as bases for sub-sequent social judgments themselves. The lat-ter role may account for the poor correspon-dence often found between judgments of othersand recall of their behavior (e.g., Anderson &Hubert, 1963;Dreben,Fiske,&Hastie, 1979).There is also a poor correspondence betweeninferences and memory for their bases in theliterature on attitudes and persuasion (e.g.,Eagly & Himmelfarb, 1978;Greenwald, 1968;McGuire, 1969). Here, too, spontaneous socialinferences may occur about both the attitudeobject and the communicator (e.g., Eagly &Chaiken, 1975). If these decay more slowlythan memory for the specifics of the messageand the source's behavior or features, then at-titudes and recall of specific message contentscould easily become uncorrelated.

On the whole, early theorists' intuitions thatsocial inferences may be ubiquitous are sup-ported, as are the more recent suggestions bySmith and Miller (1983) and Anderson (1983).Certainly trait inferences are not always ef-fortful, attention-demanding cognitive opera-tions or strategies. They can also occur withoutinstructions or other special motivations suchas the needs to explain novelty or failure (seeWeiner, 1985). As Jones (1979) suggested, thecognitive road from acts to dispositions is notso rocky after all. In fact, like a superhighway,it may often be traveled "on automatic."

References

Anderson, C. A. (1983). Abstract and concrete data in theperseverance of social theories: When weak data lead to

unshakable beliefs. Journal of Experimental Social Psy-chology, 19, 93-108.

Anderson, N. H., & Hubert, S. (1963). Effects of concom-

mitant verbal recall on order effects in personalityimpression formation. Journal of Verbal Learning andVerbal Behavior, 2, 379-391.

Asch, S. E. (1946). Forming impressions of personality.

Journal of Abnormal and Social Psychology, 41, 258-290.

Bahrick, H. P. (1969). Measurement of memory byprompted recall. Journal of Experimental Psychology,79,213-219.

Bargh, J. A. (1982). Attention and automaticity in the pro-cessing of self-relevant information. Journal of Person-

ality and Social Psychology, 43, 425-436.


Baigh, J. A. (1984). Automatic and conscious processing

of social information. In R. S. Wyer& T. K. Srull (Eds.),Handbook of social cognition (Vol. 3, pp. 1-43). Hills-dale, NJ: Erlbaum.

Bargh, J. A., Bond, R. N., Lombard!, W. J., & Tota,

M. E. (1985). Category accessibility and impression for-

mation: The long and the short of it. Unpublished manu-script, New York University.

Bargh, J. A., & Pietromonaco, P. (1982). Automatic in-formation processing and social perception: The influ-

ence of trait information presented outside of consciousawareness on impression formation. Journal of Person-

ality and Social Psychology, 43, 437-449.Bargh, J. A., & Thein, R. (in press). Individual construct

accessibility, person memory, and the recall-judgmentlink: The case of information overload. Journal of Per-

sonality and Social Psychology.Berscheid, E., Graziano, W., Monson, T., &. Dermer, M.

(1976). Outcome dependency: Attention, attribution,

and attraction. Journal of Personality and Social Psy-

chology, 34, 978-989.Carlston, D. E. (1980). Events, inferences, and impression

formation. In R. Hastie et al. (Eds.), Person memory:

The cognitive basis of social perception (pp. 89-119).

Hillsdale, NJ: Erlbaum.Cohen, J. (1969). Statistical power analysis for the behav-

ioral sciences. New \brk: Academic Press.

Dreben, E. K., Fiske, S. T, & Hastie, R. (1979). The in-dependence of item and evaluative information:Impression and recall order effects in behavior-basedimpression formation. Journal of Personality and Social

Psychology, 37, 1758-1768.Eagly, A. H., & Chaiken, S. (1975). An attribution analysis

of the effect of communicator characteristics on opinion

change: The case of communicator attractiveness. Jour-

nal of Personality and Social Psychology, 32, 135-144.Eagly, A. H., & Himmelfarb, S. (1978). Attitudes and

opinions. In M. R. Rosenzweig & L. W. Porter (Eds.),Annual review of psychology (Vol. 29, pp. 517-554). Palo

Alto, CA: Annual Reviews,Enzel, M. E., & Schopflocher, D. (1978). Instigation of

attribution processes by attributional questions. Per-

sonality and Social Psychology Bulletin, 4, 595-599.Ericsson, J. A., & Simon, H. A. (1980). Verbal reports as

data. Psychological Review, 87, 215-251.Gillund, G., & Shiffrin, R. M. (1984). A retrieval model

for both recognition and recall. Psychological Review

91, 1-67.Greenwald, A. G. (1968). Cognitive learning, cognitive re-

sponse to persuasion, and attitude change. In A. G.Greenwald, T. C. Brock, & T. M. Ostrom (Eds.), Psy-

chological foundations of attitudes (pp. 147-171). NewYork: Academic Press.

Greenwald, A. G. (1975). Consequences of prejudiceagainst the null hypothesis. Psychological Bulletin, 82,

1-20.Heider, F. (1944). Social perception and phenomenal cau-

sality. Psychological Review, SI, 358-374.Higgins, E. T, & King, G. (1981). Accessibility of social

constructs: Information processing consequences of in-dividual and contextual variability. In N. Cantor &J. F. Kihlstrom (Eds.), Personality, cognition, and socialinteraction (pp. 69-121). Hillsdale, NJ: Erlbaum.

Higgins, E. T, & Rholes, W. S. (1976). Impression for-

mation and role fulfillment: A "holistic reference" ap-

proach. Journal of Experimental Social Psychology, 12,422-435.

Jones, E. E. (1979). The rocky road from acts to disposi-

tions. American Psychologist, 34, 107-117.Kelley, H. H., & Michela, J. L. (1980). Attribution theory

and research. In Annual review of psychology (Vol. 31,pp. 457-501). Lincoln: University of Nebraska Press.

Langer, E. J. (1978). Rethinking the role of thought insocial interaction. In J. H. Harvey et al. (Eds.), New

directions in attribution research (Vol. 2, pp. 35-58).Hillsdale, NJ: Erlbaum.

Logan, G. D. (1979). On the use of a concurrent memory

load to measure attention and automaticity. Journal ofExperimental Psychology: Human Perception and Per-

formance, 5, 189-207.McGuire, W. J. (1969). The nature of attitudes and attitude

change. In G. Lindzey & E. Aronson (Eds.), The hand-

book of social psychology (Vol. 3, pp. 136-314). Reading,MA: Addison-Wesley.

Miller, F. D., Smith, E. R., & Uleman, J. S. (1981). Mea-surement and interpretation of situational and dispo-

sitional attributions. Journal of Experimental SocialPsychology, 17, 80-95.

Nisbett, R. E., & Wilson, T. D. (1977). Telling more thanwe can know: Verbal reports on mental processes. Psy-

chological Review, 84. 231-259.Ostrom, T. M., Lingle, J. H., Pryor, J. B., & Geva, N.

(1980). Cognitive organization of person impressions.

In R. Hastie et al. (Eds.), Person memory: The cognitive

basis of social perception (pp. 55-88). Hillsdale, NJ:Erlbaum.

Posner, M. T, & Snyder, C. R. R. (1975). Attention andcognitive control. In R. L. Solso (Ed.), Information pro-

cessing and cognition: The Loyola Symposium (pp. 55-85). Hillsdale, NJ: Erlbaum.

Pyszczynski, T. A., & Greenberg, J. (1981). Role of dis-

confirmed expectancies in the instigation of attributionalprocessing. Journal of Personality and Social Psychology,

40, 31-38.Schneider, W, & Shiflrin, R. M. (1977). Controlled and

automatic human information processing: I. Detection,search, and attention. Psychological Review, 84, 1-61.

Shiffrin, R. M., & Dumais, S. T. (1981). The developmentof automatism. In J. R. Anderson (Ed.), Cognitive skills

and their acquisition (pp. 111-140). Hillsdale, NJ: Erl-baum.

Shiffrin, R. M., & Schneider, W. (1977). Controlled andautomatic human information processing: II. Perceptuallearning, automatic attending, and a general theory.Psychological Review, 84, 127-190.

Smith, E. R., & Miller, F. D. (1983). Mediation among theattributional inferences and comprehension processes:Initial findings and a general method. Journal of Per-

sonality and Social Psychology, 44, 492-505.Srull, T. K., & Wyer, R. S. (1979). The role of category

accessibility in the interpretation of information aboutpersons: Some determinants and implications. Journal

of Personality and Social Psychology, 37, 1660-1672.

Tagiuri, R. (1958). Introduction. In R. Tagiuri & L. Petrullo(Eds.), Person perception and interpersonal behavior.Stanford, CA: Stanford University Press.

Taylor, S. E., Crocker, J., Fiske, S. T, Sprinzen, M., &Winkler, J. P. (1979). The generalizability of salience


effects. Journal of Personality and Social Psychology,37, 357-369.

Thomson, D., & Tulving, E. (1970). Associative encoding

and retrieval: Weak and strong cues. Journal of Exper-

imental Psychology, 86, 255-262.Tulving, E. E. (1972). Semantic and episodic memory. In

E. Tulving & W. Donaldson (Eds.), Organization and

memory (pp. 381-403). New York: Academic Press.Tulving, E. E., & Osier, S. (1968). Effectiveness of retrieval

cues in memory for words. Journal of ExperimentalPsychology, 77, 593-601.

Tulving, E. E., & Pearlstone, Z. (1966). Availability versusaccessibility of information in memory for words. Jour-

nal of Verbal Learning and Verbal Behavior, 5.381 -391.Tulving, E. E., & Thomson, D. M. (1973). Encoding spec-

ificity and retrieval processes in episodic memory. Psy-

chological Bulletin, 30, 352-373.Watkins, M. J., & Gardiner, J. M. (1982). Cued recall. In

C. Richard Puff (Ed.), Handbook of research methods

in human memory and cognition (pp. 173-195). New

York: Academic Press.Weiner, B. (1985). "Spontaneous" causal thinking. Psy-

chological Bulletin, 97, 74-84.Winter, L. (1980). Social judgments as cognitive processes:

A test of some basic assumptions in social cognition.Unpublished master's thesis, New York University.

Winter, L., & Uleman, J. S. (1984). When are social judg-ments made? Evidence for the spontaneousness of trait

inferences. Journal of Personality and Social Psychology,

47, 237-252.Winters, J. J., & Brzoska, M. A. (1975). Development of

lexicon in normal and retarded persons. Psychological

Reports, 37, 391-402.

Received November 26, 1984Revision received May 3, 1985 •

Special Call for Papers on the Personality and Social Psychology of Aging

The Psychology and Aging journal is gearing up for its first year of publication in 1986.

Manuscripts have been received in the editorial office for a number of months, but

more than 50% of the original submissions have been experimental. The Editor,

M. Powell Lawton, and the Associate Editor, Donald H. Kausler, wish to emphasize

that Psychology and Aging will be a broad-ranging publication, and manuscripts from

all areas of psychology are desired.

Papers on personality and social issues related to aging are encouraged. As the

proposed editorial policy statement outlines:

Psychology and Aging publishes original articles on adult development and aging. Such original

articles include reports of research, which may be applied, biobehavioral, clinical, educational,experimental (laboratory, field, or naturalistic studies), methodological, or psychosocial. Whilethe emphasis is on original research investigations, occasional theoretical analyses of researchissues, practical clinical problems, or policy may appear, as well as critical reviews of a content

area in adult development and aging. Clinical case studies that have theoretical significanceare also appropriate. Brief reports are acceptable with the author's agreement not to submit afull report to another journal; a 75-100 word abstract plus 48-space lines of text and referencesconstitute absolute limitations on space for such brief reports.

Manuscripts should be directed to:

M. Powell Lawton

Philadelphia Geriatric Center

5301 Old York Road

Philadelphia, Pennsylvania 19141

Documents

How Automatic Are Social Judgments? - Uleman Lab