Journal of Experimental Psychology : General Speeded Naming or Naming Speed ? The Automatic Effect of Object Speed on Performance

We test the hypothesis that naming an object depicted in a picture and reading aloud an object's name are affected by the object's speed. We contend that the mental representations of everyday objects and situations include their speed, and that the latter influences behavior in instantaneous and systematic ways. An important corollary is that high-speed objects are named faster than low-speed objects, although object speed is irrelevant to the naming task at hand. The results of a series of 7 studies with pictures and words support these predictions.

Everyday objects evoke mental constructs and dispositions beyond mere perception or recognition.These constructs are activated in a fast and automatic fashion, often outside of conscious awareness (Bargh, 1994;Higgins, 1996;Hassin, 2013).Arguably the best-known and most-researched property of objects is their valence.Valence is present not only with obviously threatening or appealing stimuli such as a snake or a piece of food, but a modicum of valence or "microvalence" (Lebrecht, Bar, Barret, & Tarr, 2012) is present in such an innocuous object as your morning coffee mug.The prevalence of valence is easily understood considering its role in evolution and its role in shaping online motivations, emotions, and decisions.
In the present study, we turn the spotlight to another high-level property-speed-which has largely been overlooked in the existing literature.We argue that the mental representations of everyday objects often include a value of speed, which can influence people's actions in a systematic fashion.In a series of seven studies, we show that object speed influences performance in such simple and instantaneous tasks as naming a picture of these objects or reading aloud their names.This influence is all the more impressive when one recognizes that object speed is irrelevant to the explicit task set at hand.Speed is a continuous variable.Some objects are not associated with speed or merely have low speed (what one might term, following Lebrecht et al., 2012, "microspeed," e.g., a plant).Other objects prompt slow to moderate speed (e.g., turtle, snail), and still other objects evoke high velocity (e.g., airplane, train).Of course, the relevant speed value is dependent on context.An airplane parking on the ground is likely to activate a different speed than one that is currently in flight.
The activation of an object's speed might be also vital for survival.A threatening attack dog is of extremely negative valence, yet the proverbial decision of fight or flight is resolved by the assessment of speed and proximity (Fanselow, 1994;Maren, 2007;Mobbs et al., 2007).In the same manner, regardless of the affect associated with cars, the decision to cross or not cross a busy road should depend on speed and direction of movement-and an error can be very costly.It is partly for these reasons that movement and speed, subsumed under "activity," were found to be a fundamental dimension of meaning in the classic research by Osgood, Suci, and Tannenbaum (1957).
The idea that object speed may affect simple and early cognitive processing was inspired by two leading schools of thought in contemporary experimental psychology.Embodiment, or grounded cognition, holds that when "people perceive visual objects, simulations of potential actions become active in preparation for situated action" (Barsalou, 2008, p. 624).Thus, one way that words and pictures convey meaning is grounded in the bodily activity associated with them (e.g., Glenberg & Kaschak, 2002;Schubert, 2005;Williams, Huang, & Bargh, 2009).When one sees sharks, or the word "shark," an implication is that things can happen really fast; therefore, speedy action might be needed.
A similar line of thinking originates from considering the multitudinous effects of priming (for recent reviews, see Bargh, Schwader, Hailey, Dyer, & Boothby, 2012;Hassin, 2013).For example, in the most relevant line of research, Bargh, Chen, and Burrows (1996) have shown that priming old age leads to slower walking (but see Doyen, Klein, Pichon, & Cleeremans, 2012 for a different take on this effect).Building on these findings, Cesario, Plaks, and Higgins (2006) have shown that this effect is motiva-tional in nature: Participants who implicitly like older adults indeed walk more slowly after priming whereas those who like them less actually walk faster.In another study, Matlock (2004) found that it took participants longer to semantically process a sentence entailing fictive motion (e.g., "The road runs through the valley") when this sentence followed a story involving slow motion (vs.one that implied fast motion).
Thus, the priming and the embodiment accounts suggest a link between an object's speed and subsequent behaviors (e.g., Bargh et al., 1996;Barsalou, 2008).Note that both accounts distinguish between the stimulus that brings about the priming (or the simulation) and the process that it changes.To take just one example, Bargh and colleagues (1996) have primed the notion of slowness via the reading of words related to old age and shown that this priming phase subsequently slowed participants' walking (see also Matlock, 2004).
But why wait?If we assume that the act of speeding cognition toward faster objects serves a function, then it makes sense to speed cognition as fast as one can.Therefore, the hypothesis examined here is that the effect of stimulus speed is inherent in processing to the extent that it affects the performance with respect to the stimulus itself.To test this hypothesis, we examine the process of reading a word or naming a picture, hypothesizing that faster objects are processed more quickly than slower ones.

The Present Study
The latency of naming words and pictures is influenced by a wealth of well-known variables, including word frequency, word length, phonetic structure, orthographic neighborhood, age of acquisition, picture complexity, goodness of depiction, and name agreement (e.g., Balota et al., 2007;Bates et al., 2003;Székely et al., 2004).Nevertheless, a relatively large portion of the variance in naming latency is still unexplained by these attributes.In the present study, we suggest the variable of object speed as a potent predictor of word-and picture-naming latencies, even when one controls for all previously mentioned variables.
The hypothesis that we examine here is that the aforementioned determinants are not the only systematic determinants of cognitive speed, and that high-level features also influence this process.In the studies we conduct and report here, the participants' task was to name objects depicted as pictures or to read them as words.The presented pictures and words were drawn from standardized, internationally recognized pools of stimuli (pictures: the International Picture Naming Project [IPNP, Székely et al., 2004]; words: the English Lexicon Project [ELP; Balota et al., 2007]).Three features of these pools of stimuli are noteworthy.The first refers to their sheer size.For instance, the ELP includes more than 40,000 English words.Second, the pictures (and the words) transcend several categories, from household items to foods to natural phenomena.Third, and perhaps most important, the data entail behavioral norms, notably mean latencies to name the pictures and the words.These behavioral data are also broadly based (e.g., latencies in the ELP are based on responses by over 400 people).
Of the large population of pictures (and words), we focused on the subcategory of vehicles.Our goal was to sample stimuli for which speed is almost invariably relevant, spanning a large range of values of speed.In Study 1 (IPNP pictures) and Study 4 (ELP words), our subjects rated the stimuli for speed.Subsequently, we correlated the mean latencies for naming available in the international norms with the rating of speed by our participants.Studies 2 and 5 looked at correlations between rated speed and the speed of naming/reading in our laboratories, controlling for known lexical predictors.Study 3 further demonstrated the effect of object speed in a larger set of 275 IPNP pictures of common actions.Finally, in two dedicated experiments probing causality, Experiment 1 and Experiment 2, we manipulated the context of the objects such that in one context the objects were faster than in another (e.g., a car driving uphill or downhill).This context manipulation allowed tight control over virtually all confounding variables while testing a fully causal account.

Study 1
The stimuli presented in this study were pictures of vehicles drawn from the IPNP (Székely et al., 2003(Székely et al., , 2004)).Mean latency norms to name each picture were also obtained from the IPNP.We independently collected (nonspeeded) ratings of speed for each picture by a group of Israeli students.Does the time needed to name the stimulus-a task of picture recognition-depend on the speed inherent in the referent object?

Method
Participants.The participants were 44 Open University undergraduates (33 women; mean age 27 years).The participants rated the apparent speed of the objects depicted in each picture.All participants had normal or corrected-to-normal vision, and they received course credit.
We tested this relatively large group of 44 participants to produce reliable assessments of object speed.In subsequent (betweensubject) Studies 3 and 4, we similarly tested large groups of at least 40 participants to achieve the same goal.The precise number of participants depended on availability (via volunteer enrollment) before the study.
Apparatus and stimuli.We selected all 35 items in the IPNP category of vehicles.Because some items were stationary (e.g., slide), we used the 29 items depicting vehicles of locomotion (e.g., wheelchair, bicycle, motor car, rocket).For each picture, we recorded the mean latency to name the referent object that had been collected within the IPNP.1 Pilot measurements: Ratings of valence, threat, and arousal of pictures in Study 1 and 2. A group of 42 Open University undergraduates (31 females; mean age 29 years), none of whom participated in the current studies, judged the pictures on valence, on threat, or on arousal.Each judge rated all randomly presented pictures in a different order on one of three Likert scales: 1 (good) to 7 (bad), 1 (not threatening) to 7 (threatening), and 1 (not exciting) to 7 (exciting).One item from Study1 (stroller #019) was missing because of technical reasons.Because the category of vehicles included two pictures of baby strollers, the pertinent missing values were subsequently replaced with those of the second stroller in the list (stroller#428).
Procedure.The participants were tested individually in a dimly lit room.Presented with a single picture on the computer screen, they judged the referent vehicle's speed on a 1 (slow) to 7 (fast) scale.The participants typed in their response on the keyboard, after which the next picture appeared.These ratings of speed were not timed.Each participant received the set of pictures in a random and different order.

Results and Discussion
A glimpse at Figure 1 reveals a remarkable association between the two independent sets of data.The Pearson correlation between the naming reaction times (RTs) and the average ratings of speed amounted to r(27) ϭ Ϫ.62 (p Ͻ .001). 2 To assess the unique contribution of speed to naming RT and to control residual shared variance with other known higher-order variables, we additionally correlated the ratings of speed with naming latency after partialing out the ratings of valence, threat, and arousal: r(26) ϭ Ϫ.62, p Ͻ .001after partialing out valence; r(26) ϭ Ϫ.63, p Ͻ .001after partialing out threat; and r(26) ϭ Ϫ.66, p Ͻ .001after partialing out arousal.Removing valence, threat, or arousal clearly left the association of naming RT and object speed intact.
What about other known determinants of recognition latency?When we included the lexical features provided by the IPNP (number of alternative names, percent name agreement, 3 length in syllables, length in characters, frequency, 4 age of acquisition, 5 and picture visual complexity 6 ; Székely et al., 2003) and three known semantic variables (valence, threat, and arousal; see pretest) in a stepwise multiple regression, object speed proved an important predictor of naming latency.In fact, object speed (␤ ϭ Ϫ.37, p ϭ .006)and the number of alternative names (␤ ϭ .60,p Ͻ .001)proved the only reliable predictors of naming performance.It is interesting to note that together these two variables explained over 65% of the variance in naming latency (F(2, 26) ϭ 27.37, p Ͻ .0001for adjusted R 2 ), whereas speed alone explained over 36% in an independent model (F(1, 27) ϭ 16.72, p Ͻ .001).
In conclusion, ratings of object speed by Israeli participants reliably correlated with the time needed to recognize the same objects by U.S. participants.

Study 2
In Study 2, we tested our hypothesis in a more powerful withinsubject design in which the same group of participants performed both speeded naming of objects and nonspeeded rating of the speed of those objects in separate blocks of presentations.

Method
Participants.The participants were 20 Open University undergraduates (15 women; mean age 26 years).All participants had normal or corrected-to-normal vision, and they received course credit.One participant whose responses were more than 50% invalid (multiple microphone failures and object misidentifications; see data analysis below) was omitted from the analysis.
In Studies 2 and 5 that entailed a single group of participants (performing both in ratings of speed and object naming), we collected data from approximately 20 participants.The precise number depended on enrollment before the study.
Apparatus and stimuli.The stimuli were the same set of pictures of vehicles used in Study 1. Two of the pictures were dropped because they have the same name in Hebrew (stroller [#19], wagon [#488]).The set of 27 pictures was presented three times in a random fashion, making for 81 experimental trials in all.
2 Performing the same analysis on all 35 items in the IPNP category, including the 6 nonvehicle stationary items (slide, tire, steering-wheel, wheel, seesaw, and swing), keeps the speed association significant at r(33) ϭ Ϫ.60, p Ͻ .001.
3 Measures the proportion of all valid trials on which participants produced the dominant target name. 4Frequency counts based on the CELEX Lexical database (Baayen, Piepenbrock, & Gulikers, 1995).
5 Taken from published norms of the U.S. version of the MacArthur Communicative Development Inventories (Fenson et al., 1994).This measure includes a three-point scale of parental assessments: 1 ϭ word was acquired between 8 and 16 months; 2 ϭ between 17 and 30 months; 3 ϭ words that are not acquired in infancy (Ͼ30 months).
6 Estimates based on the size of the digitized JPEG stimuli picture files set at a resolution of 300 ϫ 300 pixels. 7We used the models suggested by Minitab, v. 17.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
All of the pictures were presented via a Dell computer and displayed on a 17-in.color monitor set at a resolution of 1,024 ϫ 768 pixels (the resolution of the pictures was set at 300 ϫ 300 pixels).
The participants performed eight practice trials with a set of nonvehicle objects.
Procedure.The participants were tested individually in a dimly lit room.Their first task was a speeded naming of the objects depicted in the pictures.Presented with a picture on the computer screen, the participant was asked to name it as quickly and accurately as possible by saying the name into the microphone headset (Teac HPX-8 brand).DirectRT software (Version 2008.1.0.11) recorded the time until the participant began to pronounce a response.Stimulus exposure was response-terminated.The interval between the participant's response and the appearance of the next stimulus was 500 ms.
The second task was a nonspeeded rating of the same object depicted in the picture.The participants judged speed on a 1 (slow) to 7 (fast) scale.Each participant received the set of pictures in a different random order in the two tasks.
Data analysis.For the first speeded task, we used the criteria offered by Székely et al. (2003) for classification of valid responses (e.g., removing verbalizations such as "that's a ball," hesitations, or noncodable names).In addition, responses shorter than 250 ms or longer than 2,250 ms (2.8% of valid responses) were not analyzed.As performed by Székely et al. (2003), the number of alternative names for each picture was determined by "number of types" (i.e., number of different names provided on valid trials, including the target name).The percentage name agreement was defined as the proportion of all valid trials in which participants produced the dominant target name.

Results
Object speed and speed of naming.The correlation between the time needed to name the object and the rating of object speed amounted to an appreciable r(25) ϭ Ϫ.44, (p ϭ .022;see Table 2 for all correlations).To further control for shared residual variance of other semantic variables with speed, we correlated the ratings of speed with naming latency after partialing out ratings of valence, threat, or arousal (see Study1, pretest).Although valence had a significant association with naming latency (see Table 2), this association seemed independent from the association of speed ratings because the correlation of speed ratings with naming RTs remained highly reliable at r(24) ϭ Ϫ.42, p ϭ .034after partialing out valence.Arousal and threat did not correlate with naming latency, but similarly partialing out arousal or threat did not affect the association of speed with naming latency (r(24) ϭ Ϫ.44, p ϭ .025after partialing out arousal; r(24) ϭ Ϫ.47, p ϭ .017after partialing out threat).Thus, it seems that the documented effects of speed cannot be attributed to these variables.In a stepwise multiple regression that included all available lexical features,8 and the three semantic variables-valence, threat, and arousal-speed was selected as a significant predictor of naming latency in a threefactor model (␤ ϭ Ϫ.25, p ϭ .047)including name agreement (␤ ϭ Ϫ.51, p ϭ .001)and valence (␤ ϭ .37,p ϭ .012),adjusted R 2 ϭ 51.9%, F(3, 23) ϭ 10.37, p Ͻ .001.Speed alone explained 16% of the adjusted variance in an independent model, F(1, 25) ϭ 5.92, p ϭ .022.

A Multilevel Within-Participant Analysis
Because in this study we collected in the laboratory RTs from individual participants, we could additionally conduct a more  This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

4
BEN-HAIM, CHAJUT, HASSIN, AND ALGOM powerful multilevel analysis that incorporates interparticipant variability.In this analysis, speed was similarly found a highly significant predictor of naming times, 9 B ϭ Ϫ26.12 (SE ϭ 5.91), t(502.2) ϭ Ϫ4.42, p Ͻ .0001.The estimate represents a predicted acceleration rate of 26.12 ms in RT per single point in speed rating, and a cumulative acceleration of ϳ157 ms of an item with a rating score of 7 compared with an item with a rated speed of 1.In a multilevel analysis that included the predictors selected by the stepwise regression, speed was a reliable predictor of naming latency, B ϭ Ϫ16.18 (SE ϭ 5.73), t(499.9)ϭ Ϫ2.82, p Ͻ .01,along with valence and name agreement, p Ͻ .0001.
In another test of the effects of lexical features, we selected five of the fastest and five of the slowest rated pictures matched on the three most contributing variables (alternative names, t(4) ϭ .17p ϭ .87;frequency, t( 4

Discussion
The results of Studies 1 and 2 support the notion of activation of the higher-level property of speed when people recognize everyday objects.Crucially, object speed was never mentioned in the instructions of the naming task, nor was it an explicit part of the task description.
A possible reservation with respect to the results of Studies 1 and 2 is that speed plays a role only when speed is blatantly expressed.To address this concern, in Study 3 we examined stimuli that are not as clearly associated with speed or movement.We made use of the large category of pictures from the IPNP (Székely et al., 2005) depicting people's actions.The category of actions includes 275 pictures of everyday behaviors (e.g., writing, tooth brushing, painting, fishing, drinking).Many of the stimuli that appear in these scenes move very slowly, at best.Nevertheless, if speed is a feature that is activated in an automatic fashion, one should still find an association between naming latency and subjective speed.
In Study 3, we used the tactic of Study 1: RT norms for naming each of the 275 pictures in the IPNP were pit against ratings of speed of the same objects obtained from a local group of Israeli participants.

Study 3 Method
Participants.The participants were 44 Open University undergraduates (34 women; mean age 27 years).They rated the apparent speed of each object depicted in the picture.All participants had normal or corrected-to-normal vision, and they received course credit.
We tested this relatively large group of 44 participants to produce reliable assessments of object speed in this between-subject design.The precise number of participants depended on availability (via volunteer enrollment) before the study.
Apparatus and stimuli.The stimuli were the 275 drawings of everyday actions included in the IPNP database.For each picture, we selected the mean latency to name the referent object from the IPNP norms (Székely et al., 2005).
Pretest: Ratings of valence, threat, and arousal.Forty-five Open University undergraduates (37 females; mean age 29 years), none of whom participated in the current or previous studies, judged the pictures on either valence, threat, or arousal.Each judge rated all 275 randomly presented pictures in two sessions separated by a break of ϳ30 min in a different order on one of three Likert scales: 1 (good) to 7 (bad), 1 (not threatening) to 7 (threatening), and 1 (not exciting) to 7 (exciting).
Procedure.The participants were tested individually in a dimly lit room.Presented with a single picture on the computer screen, they judged the referent's speed on a 1 (slow) to 7 (fast) scale.The participants typed their response on the computer keyboard.The ratings of speed were not timed.Each participant received the set of pictures in a random and different order.A short break separated the first and second halves of the stimuli.

Results and Discussion
We found an association between the time needed to name each picture, as it was obtained from the U.S. participants in the IPNP, and the ratings of each object's speed, given by Israeli participants.The correlation was relatively small, yet reliable, at r(273) ϭ Ϫ.14, p ϭ .016.To further control for shared residual variance of these variables with speed, we correlated the ratings of speed with naming latency after partialing out ratings of valence, threat, or arousal (see pretest).Valence had a significant association with naming latency and was also correlated with speed ratings (see Table 3).However, this association seemed independent from the association of speed ratings because the correlation of speed ratings with naming RTs increased to r(272) ϭ Ϫ.19, p ϭ .001after partialing out valence.Likewise, partialing out threat, which was also correlated with speed ratings, or arousal had a refining effect on the association of speed with naming latency: r(272) ϭ Ϫ.20, p ϭ .001after partialing out threat; r(272) ϭ Ϫ.16, p ϭ .008after partialing out arousal.
Notably the contribution of object speed to picture recognition remained reliable (␤ ϭ Ϫ.11, p ϭ .008) in a stepwise multiple regression that included all available lexical, perceptual, and semantic factors (Székely et al., 2005). 10Along with speed, included in the stepwise solution were alternative names (␤ ϭ .44,p Ͻ .0001),name agreement (␤ ϭ Ϫ.33, p Ͻ .0001),visual complexity (␤ ϭ .08,p ϭ .042),and valence (␤ ϭ .13,p ϭ .002),R 2 adjusted ϭ 58.7%, F(5, 269) ϭ 78.75, p Ͻ .0001.The best-subset analysis (for criteria, see Study 1) indicated that there was a model 9 The multilevel analysis included ratings of speed as a fixed covariate as well as a fixed intercept using objects as a repeated measure.The analysis was performed in SPSS v. 18.We used a model of restricted maximum likelihood estimation, entailing compound symmetry as a repeated covariance type. 10The lexical predictors used were the same set of factors provided within the IPNP (Székely et al., 2005).These included the number of alternative names, percentage name agreement, length in syllables, length in characters, frequency, and age of acquisition.The perceptual factors included were visual complexity and ratings of valence, threat, and arousal (see pretest).This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.3 for all of the correlations with naming latency.Note that the speed ratings effect was comparable in size with that of valence in this large pool of everyday actions.It seems that because speed was less consequential for the tested actions, its effect was weaker, yet present.
The results of Study 3 show that the variable of subjective speed is correlated with object naming in a wide variety of situations and actions.In Studies 1-3, naming performance was influenced by the task-irrelevant property of object speed.Would the same effect be obtained for word reading?Studies 4 and 5 examined the effect of speed implied in the meaning of the word against the time needed to read that word.
Study 4 was a conceptual replication of Study1, except that here the stimuli were words denoting the same objects.The words along with their RT norms were drawn from the large database included in the ELP (Balota et al., 2007).Ratings of the speed of the objects conveyed by the words were obtained from a local group of participants.

Study 4 Method
Participants.The participants were 43 Open University undergraduates (31 women; mean age 27 years).They rated the speed of the referent vehicles denoted by the words.All participants had normal or corrected-to-normal vision, and they received course credit.
We tested this relatively large group of 43 participants to produce reliable assessments of object speed in this between-subject design.The precise number of participants depended on availability (via volunteer enrollment) before the study.
Apparatus and stimuli.The stimuli were the same set of vehicles used in Study 1.Because the words unicycle, fire truck, and roller skate do not have an RT norm in the ELP, we presented only 24 items with available RT norms for reading.For items with several equivocal dictionary translations in Hebrew, we presented all alternative names on the screen as describing the item to be rated.
Pilot measurements: Ratings of valence, threat, and arousal of the words in Study 4 and 5.A group of 48 Open University undergraduates (36 females; mean age 31 years), none of whom participated in the current or previous studies, judged the Hebrew words on valence, threat, or arousal.Each judge rated all 42 randomly presented words (including alternative Hebrew translations of the vehicles) in a different order on one of three Likert scales: 1 (good) to 7 (bad), 1 (not threatening) to 7 (threatening), and 1 (not exciting) to 7 (exciting).Because the Hebrew equivalent of a given English word can be translated into several alternative names of an object (e.g., either plane or airplane is a legitimate translation of airplane in Hebrew), the ratings of the (English) vehicle names in the pilot of Study 4 were calculated based on the mean rating of all of its alternative Hebrew translations.
Procedure.The participants were tested individually in a dimly lit room.Presented with a single word, they judged the speed of the referent item on a 1 (slow) to 7 (fast) scale.The ratings of speed were not timed.Each participant received the set of words in a random and different order.

Results and Discussion
Figure 2 shows that the correlation between the independent sets of data is appreciable.The Pearson correlation amounted to r(22) ϭ Ϫ.55 (p ϭ .005).This result shows that the time needed to read a word is correlated with the speed of the object that the word names.To further control for any shared residual variance of these variables with speed in this study, we correlated the ratings of speed with naming latency after partialing out ratings of valence, threat, or arousal (see pretest).Although threat was highly correlated with speed ratings and reading latency (see Table 4), the correlation of speed with reading latency remained reliable after partialing out threat, r(21) ϭ Ϫ.37, p ϭ .041.Likewise, partialing out arousal or valence did not harm the association of speed ratings with reading times, r(21) ϭ Ϫ.60, p ϭ .001,after partialing out arousal; r(21) ϭ Ϫ.55, p ϭ .003after partialing out valence.Thus, it seems that speed has a unique and independent contribution in predicting reading latency.
The ELP reports the values of several lexical factors for each word.In a stepwise multiple regression including all available perceptual, semantic, and lexical factors, 11 speed turned out to be the strongest predictor of reading time (␤ ϭ Ϫ.54, p ϭ .001),along with pronunciation (␤ ϭ Ϫ.52, p ϭ .002)and syllables (␤ ϭ .33,p ϭ .03).These three variables explained over 55.5% of the 11 The lexical factors drawn from the ELP were length in characters, Hyperspace Analogue to Language (HAL) frequency, orthographic neighbors, number of syllables, and pronunciation naming accuracy (Balota et al., 2007).The semantic factors were the ratings of valence, threat, and arousal (see pretest).This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.4 for the correlation with naming latency of the individual predictors.

Study 5
We deemed the results of Study 4 worthy of replication in a laboratory context.In Study 5, a group of participants performed both speeded reading and nonspeeded rating of the same items for a more powerful within-participant design.In addition, we also included alternative names for items that have alternative names in Hebrew (e.g., plane-airplane).This addition allows us to measure the correlation between the reading latency of two different words that denote the same object.

Method
Participants.The participants were 18 Open University undergraduates (14 women; mean age 23 years).All participants had normal or corrected-to-normal vision, and they received course credit.
In this study and Study 2, which entailed a single group of participants (performing both in ratings of speed and object nam-ing), we collected data from ϳ20 participants.The precise number depended on enrollment before the study.
Apparatus and stimuli.The stimuli were the same set of vehicles from Study 1.This set was extended in the present study by including alternative names for the original pictures (e.g., the words "car" and "automobile" were both included).Consequently, the list of stimuli included 42 words.The set of 42 words was presented twice in a random fashion.All of the words were presented via a Dell computer and displayed on a 17-in.monitor set at a resolution of 1,024 ϫ 768 pixels.The words were presented in black, in bold Ariel font, size 20, on the white background of the screen.
Procedure.The participants were tested individually in a dimly lit room.The first task for the participants was speeded reading of the words.Presented with a word on the computer screen, the participant was asked to read it as quickly and accurately as possible by saying its name out loud into the microphone headset (Teac HPX-8 brand).DirectRT software (Version 2008.1.0.11) recorded the time until participants began to pronounce a response.Stimulus exposure was response-terminated.The interval between response and the appearance of the next stimulus was 1,000 ms.
The second task was nonspeeded rating of the speed of the object depicted by the word.The participants judged speed on a 1 (slow) to 7 (fast) scale.Each participant received the set of words in a different random order.
Data analysis.In the speeded task, correctly articulated responses shorter than 1,500 ms and longer than 250 ms were analyzed (97.1% of the responses; including all responses leaves the RT-rating correlation significant at .34).Overall, invalid pronunciations were rare (1.3% on average); however, one item, Zeppelin, was inaccurately articulated 14% of the time and was removed from the analysis.
In the comparison of alternative name latencies, the items were 10 pairs with an alternative name in Hebrew (e.g., helicopterchopper; stimuli appear in Appendix A).One outlier (carautomobile) was removed from the analysis because its frequency scores were above 3 standard deviations of the frequency mean (which spuriously generated a correlation in frequency scores-  This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Results
The correlation between reading time and judgment of speed was r(39) ϭ Ϫ.41 (p ϭ .008).Performing the same calculation on the original set of 29 items from Study 1 yielded a correlation of r( 27) ϭ Ϫ.44, p ϭ .017.To further control for shared residual variance of these variables with speed in this study, we correlated the ratings of speed with naming latency after partialing out ratings of valence, threat, or arousal (see Study 4 pretest).Although threat and arousal were correlated with speed ratings (see Table 5 note), this association seemed independent from the association of speed ratings because the correlation of speed ratings with reading RTs remained highly reliable (r(38) ϭ Ϫ.41, p ϭ .009)after partialing out threat or arousal (r(38) ϭ Ϫ.45, p ϭ .003).Likewise, partialing out valence did not reduce the association of speed ratings with reading times (r(38) ϭ Ϫ.44, p ϭ .004).
In another test of the influence of lexical factors, we selected matched subsets of 15 words each, denoting objects that were rated as fastest and slowest, respectively.The items were matched on length (t(14) ϭ 0, p ϭ 1) and average frequency (t(14) ϭ 0.9, p ϭ .38).The difference in reading time between the matched slow items (M ϭ 616 ms, SD ϭ 80) and fast items (M ϭ 584 ms, SD ϭ 83) also remained appreciable in this analysis, t( 17 The addition of alternative names allows an interesting comparison between the reading times of two words that denote that same object.Because these words denote the same object, there is, of course, a strong correlation between judgments of speed (r(8) ϭ .90,p ϭ .0001).However, it is important to note that in our sample there is no correlation in the lexical factors of the words (length, r(8) ϭ Ϫ.08, p ϭ .82;syllables, r(8) ϭ Ϫ.14, p ϭ .70;frequency, r(8) ϭ Ϫ.25, p ϭ .48;or pronunciation, r(8) ϭ Ϫ.15, p ϭ .67).Crucially, even with no clear lexical similarity, there is still a remarkable correlation between the reading times of alternative names, r(8) ϭ .76,p ϭ .01(see Figure 3).In the absence of correlation in the lexical factors, this item-specific correlation is a powerful demonstration of the role of semantic factors, and particularly object speed, which is predominantly relevant in the category of vehicles.

Experiments on Causation
The results of Studies 1-5 are systematic, but they are correlational in nature; the speed of objects was never manipulated.The following Experiments 1 and 2 test the causal claim we made in the introduction by manipulating speed.We presented each object twice.In one context, the to-be-named object was presented in a "slow" situation, whereas in a second context the same object was presented in a "fast" situation.We hypothesize that the "faster" objects will be named faster than the "slower" objects, although the objects are the same.To reduce effects of task sets, the instructions to these experiments did not mention speed of response (although we did measure latency, of course).

Experiment 1
Twenty pictures of objects were presented in settings that implied slow or fast motion.For example, the same car appeared once on an upward slope and once on a downward slope (see examples in Figure 4).Again, the task was to name the object (i.e., to say "car" in both cases).This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Method
Participants.Forty-six students (34 females; mean age 29 years) from the Open University performed the experiment for course credit.In the absence of prior information, in context Experiments 1-2, we collected data in multiplies of ϳ20 participants.The precise number depended on prior enrollment.
Apparatus stimuli and design.There were 10 "fast" and 10 "slow" pictures of the same 10 items.The pictures were drawn from the IPNP; we merely altered the context to create impressions of "fast" and "slow" movement (see Figure 4 and Appendix B).We randomly selected 5 "fast" pictures and their corresponding "slow" pictures to make one block of 10 objects.The remaining 10 stimuli comprised the other block.The stimuli within each block were randomly intermixed.Each stimulus in each block was presented five times.Thus, the resulting block has 50 trials.There was a break of 1 min between the two blocks.The apparatus was the same as in Study 2.
Procedure.The participants were asked to name the object in the picture into a microphone headset (Teac HPX-8 brand).Of note, participants were not instructed to be fast.DirectRT software (Version 2008.1.0.11) recorded the time until the participant began to pronounce a response.Stimulus exposure was response-terminated.The interval between the participant's response and the appearance of the next stimulus was 2,000 ms.
Data analysis.Responses were analyzed using the same criteria as in Study 2. If one member of the slow/fast-context pair was removed by these criteria, then we removed its counterpart to allow for a valid comparison in averaging RTs.

Results and Discussion
As hypothesized, the mean naming latency for the same set of objects was longer in a context suggesting slow movement than in a context suggesting fast movement (see Figure 5).The respective means were 1,053 (SD ϭ 172) and 1,030 (SD ϭ 162) ms, t( 45 To examine whether the learning of the task and its structure make a difference, we compared the first presentations of an object as "slow" and "fast."The effect was nominally bigger, amounting to 50 ms (t(1,45) ϭ 2.25, p ϭ .03).Considering the remaining data (i.e., repetitions 2-5), object speed again made a difference.The speedier context yielded an advantage of 17 ms in naming latency (t(45) ϭ 2.38, p ϭ .022).
In sum, manipulating object speed in Experiment 1 yielded results that were qualitatively the same as those obtained in Studies 1-5: Objects with a faster (implied) motion were named more quickly.In the next and final experiment, our goal was to replicate the results of the previous experiment with an extended set of objects.

Method
Participants.Sixty participants from the Open University (47 females; mean age 28 years) participated in this experiment for course credit.In context Experiments 1-2 we collected data in multiplies of ϳ20 participants.The precise number depended on prior enrollment.
Apparatus stimuli and design.The design, apparatus, and stimuli were similar to those of Experiment 1, with the following exceptions.First, we avoided repeated presentations of items; a   This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

9
ACTIVATION OF OBJECT SPEED given item was depicted only once as "slow" and once as "fast."Second, we presented a larger set of items, again drawn from the IPNP.There were 36 pictures: 18 "slow" ones and their 18 "fast" counterparts (see Appendix C).In all other respects the procedures followed those of Experiment 1.
Manipulation check: Context speed judgment.Fifteen Open University undergraduates (9 females; mean age 28 years) performed a forced choice judgment on the speed of the picture pairs.Each judge rated all 18 randomly presented pairs in a different order and was requested to select the fastest of each pair of pictures.The location of the picture contexts on the screen (left/right) was also randomly presented.
To establish that participants tend to perceive our designated contexts as slower and faster, we asked an independent group of judges to choose which of the contexts is faster in each pair (see Manipulation Check above).Sixteen of 18 of our "fast" stimuli were rated as faster by more than 80% of judges (p Յ .018 of the binomial test, mean 89.6% of participants), one item (airplane) was rated as such by only 60% of participants (p ϭ .30),and another (row) by 33% (p ϭ .15).Clearly, our fast and slow categories are explicitly recognized as such by most participants.Nonetheless, if one removes the less distinctive items (airplane and row), the effect is augmented slightly to 31 ms (p ϭ .04) in Experiment 2 and to 27 ms (p ϭ .002) in Experiment 1.
The results of Experiments 1-2 collectively rule out stimulusspecific explanations.Presenting the same object once as "fast" and once as "slow" serves as a radical control for virtually all confounding variables, especially those that refer to object-specific properties (including semantic and linguistic features).

Conclusion
In the seven studies, we found that object speed-irrespective of the explicit task set-influenced performance such that "fastmoving" objects were named faster than "slow-moving" objects.This difference was even observed for the same objects in "fast" and "slow" contexts.These results suggest that people are disposed to act swiftly with speedy objects, regardless of whether swift action is explicitly demanded by the task or not.
The effect documented here is instantaneous: It is caused by object speed, and it affects the naming/reading of that same object's picture/name.Unlike the lexical features that affect reading/ naming times, speed is probably a high-order semantic feature.As such, it should not be stored in the lexicon.
Speed turned out to be a highly reliable predictor of naming latency, at times with higher (or comparable) effect sizes than the well-established higher order variable of valence (and threat; e.g., see Algom, Chajut, & Lev, 2004;Chen, & Bargh, 1999).Valence or threat correlated with naming latency in all studies.Although most of our studies involved the sampling of vehicles of locomotion, it is noteworthy that in Study 3, which sampled common everyday objects with little movement, speed and valence carried comparable effect sizes.Of note, the effects of speed and valence were independent, and they remained equally strong (or were even augmented) after clearing of the shared variance.The importance of valence is easily understood considering its role in evolution and its role in shaping online motivation, emotions, and decisions.The activation of object speed can be vital for survival because the proverbial decision of fight or flight is often resolved by the assessment of speed and proximity (Fanselow, 1994;Maren, 2007;Mobbs et al., 2007).Our study revealed for the first time that the higher order variable of speed can be a strong predictor of reading and naming latency in the simplest of tasks.
These results establish a novel phenomenon, but they do not shed much light on the underlying process.It seems likely to us that once semantic understanding is reached, it can act swiftly to affect online processing.One possible way in which this can be achieved is through embodied/grounded cognition.According to Barsalou (2008, p. 633), "As people comprehend a text, they construct simulations to represent its perceptual, motor, and affective content.Simulations appear central to the representation of meaning."In turn, these stimulations affect behavior, broadly defined, and hence they may also affect the performance in our tasks.
The influence of mental simulations and top-down cognitive expectations are easily detected when considering moving objects.In the phenomenon termed "representational momentum" (Freyd & Finke, 1984), people often view the halt position of a moving object as lying further away along its trajectory than it really is.Although representational momentum typically involves movement (implicit and explicit), it may also play a role in generating expectations and simulations with still pictures of objects that possess or imply movement.In a noteworthy observation with still pictures entailing implied motion (very similar to those used in our study), it was shown that the pictures induced activation of brain regions associated with the processing of active visual motion (i.e., the medial temporal/ medial superior temporal cortex; Kourtzi & Kanwisher, 2000).
The present results can also be understood as an online example of priming.A remarkable aspect of priming is the access it affords to unconscious or implicit information stored in the cognitive system.Our results document the effect of priming in the simplest of tasks.In an often-cited study, Bargh et al. (1996) recorded sluggish walking after priming with elderly stereotypes.In another study, reading stories entailing slow motion induced slower decisions of fictive motion sentences (Matlock, 2004).Our results go beyond the studies of Bargh, Matlock, and others in identifying object speed as an important property of perception of objects that is instantaneously processed to influence performance with the primedstimulus itself.
Final Experiments 1 and 2 teach us one more thing about the underlying process.The documented effects of object speed cannot be attributed solely to long-term semantic knowledge.As these experiments show, at least in naming pictures, the context quickly changed the implied speed, and with it the speed of naming the pictures.This seems consistent with the idea that unconscious, automatic processes have effects that are far more pervasive than This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Our study also invites a brief look at the so called "flash lag" effect (MacKay, 1958;Nijhawan, 1994).The cognitive system perceives a moving object aligned with a flashed still object as displaced further along his trajectory than it really is.One of the explanations offered is that the brain adjusts the position of moving objects to account for the lag time it takes for it to reach consciousness (e.g., Khurana & Nijhawan, 1995;Nijhawan, 1994).
In the tenth of a second it takes the brain to perceive an object, the object has already moved, and the brain adjusts for that in the representation conveyed to consciousness.Another explanation offers that moving objects are perceived faster than flashed objects (e.g., Baldo & Klein, 1995;Purushothaman, Patel, Bedell, & Ogmen, 1998;Whitney & Murakami, 1998).Both of these explanations contend that moving objects are subject to unique processing in the visual system, making them an important feature to tag early.Our results indicate that this early unique processing may extend to still pictures and words that carry information on movement and speed.
Let us conclude with a pragmatic caveat.Current experimentation in cognitive and social psychology is largely based on speeded responses.Given the present results, investigators should watch out for possible confounding of RT by the irrelevant variable of the implicit speed inherent in the presented stimuli.

Figure 1 .
Figure 1.Naming latencies to pictures of vehicles from the IPNP norms plotted against ratings of speed of the same objects by a group of Israeli participants.The color version of this figure appears in the online article only.This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Figure 2 .
Figure 2. Reading latencies for words (drawn from the ELP norms) plotted against the ratings of speed for the named objects by an independent group of local participants.The color version of this figure appears in the online article only.

Figure 3 .
Figure 3. Reading latencies of individual items plotted against the latencies of each item's alternative name (e.g., plane-airplane; rocket-missile).The color version of this figure appears in the online article only.

Figure 4 .
Figure 4.A sample of the stimuli used in Experiments 1 and 2. Each object is presented once in a "fast" context (upper panel) and once in a "slower" context (lower panel).The pictures, drawn from the IPNP, were modified slightly to create the different contexts.

Figure 5 .
Figure 5. Mean RT to name the same 10 objects in a context suggesting impression of slow motion and in a context suggesting impression of fast motion.The bars depict 1 standard error around the mean.This document is copyrighted by the American Psychological Association or one of its allied publishers.This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Table 1
Correlation Coefficients of the Predictors Used in Study 1 With Naming Latency Note.Lexical predictors were drawn from the IPNP.None of these predictors correlated with ratings of object speed (p Ͼ .05,multiple comparisons Bonferroni corrected).‫ءءء‬ p Յ .001.

Table 2
Correlation Coefficients of the Predictors Used in Study 2 With Naming LatencyThe lexical predictors were calculated based on the Hebrew norms of the participants' dominant response.Picture visual complexity was drawn from the IPNP.None of these predictors correlated with ratings of object speed (p Ͼ .05,Bonferroni corrected).

Table 3
Correlation Coefficients of the Predictors Used in Study 3 With Naming Latency p Ͻ .001)after partialing out threat or valence (r ϭ -.19, p Ͻ .001).None of the remaining predictors correlated with the ratings of object speed (p Ͼ .05,Bonferroni corrected).‫ء‬ p Ͻ .05. ‫ءء‬ p Ͻ .01. ‫ءءء‬ p Յ .001.

Table 4
Correlation Coefficients of the Predictors Used in Study 4 With Reading Latency Note.Lexical predictors were drawn from the ELP.Threat ratings tended to correlate with speed ratings (r ϭ .60,p ϭ .002).Nonetheless, speed remained a reliable predictor after partialing out the shared variance (r ϭ -.37, p ϭ .041).None of the remaining predictors correlated with the ratings of object speed (p Ͼ .05,Bonferroni corrected).‫ء‬ p Ͻ .05. ‫ءء‬ p Ͻ .01.

Table 5
Correlation Coefficients of the Predictors Used in Study 5 WithReading Latency Lexical predictors were calculated based on the Hebrew norms.Ratings of threat and arousal tended to correlate with ratings of speed (r ϭ .68;r ϭ .47respectively, p Յ .01).Nonetheless, speed remained a reliable predictor after partialing out the shared variance with threat (r ϭ -.41, p Ͻ .01)or arousal (r ϭ -.45, p Ͻ .01).None of the remaining predictors correlated with the ratings of object speed (p Ͼ .05,Bonferroni corrected).‫ءء‬ p Ͻ .01. ‫ءءء‬ p Յ .001.This document is copyrighted by the American Psychological Association or one of its allied publishers.