01 Castro PC

In: Advances in Psychology Research, Volume 53 ISBN: 978-1-60021-924-5 Editor: Alexandra M. Columbus, pp. 17-46 © 2007 Nova Science Publishers, Inc.

Chapter 1

WORDED AND SYMBOLIC TRAFFIC SIGN STIMULI ANALYSIS USING REPETITION PRIMING

AND SEMANTIC PRIMING EFFECTS1

Cándida Castro1*, Francisco J. Tornay1, Tim Horberry2, Carlos Martínez, Alastair Gale3

and Francisco J. Martos1 1Facultad de Psicologia, University of Granada, Spain

2Centre for the Human Factors and Human Performance, the University of Queensland, Brisbane, Australia

3UVRC, Applied Vision Research Centre, University of Loughborough, UK

Abstract

Three experiments are presented in this paper. They study repetition and semantic priming effects in the recognition of traffic signs. Both indication and warning signs were used as worded and symbolic signs. We obtained a clear repetition priming effect, showing that the repetition of identical signs results in faster and more accurate responses. However, no reliable semantic priming effect was found when two signs of the same category – indication or warning – were presented. We suggest possible ways of applying the results to the improvement of traffic signalling.

Unexpectedly, and challenging the theory about picture superiority, a worded format improves the performance of indication signs. However, a pictorial format of warning signs produces better performances.

We propose procedures for pre-signalling traffic signs taking these findings into account and argue that traffic sign categories should be revised in order to make them more psychologically relevant. This way, semantic relationships between them and their format appropriateness could be used to enhance their effectiveness.

Keywords: Semantic Priming, Repetition Priming, Traffic Signs, Worded and Symbolic Signs, Warning Signs, Indication Signs.

1 This study was subsidised by the grant of the Dirección General de Investigación Científica (Directorate General

for Scientific Research) of the Spanish Ministry of Education, DIGYCIT PB97-0802. Some of the preliminary results of this work were presented at the VIII Conference on Vision in Vehicles, held in Boston en 1999.

* E-mail address: [email protected]. Telephone: +34 958 246240. Fax: +34 958 246239. Corresponding author: Cándida Castro; Contact Information: Facultad de Psicología., Universidad de Granada. Campus Cartuja, s/n. 18701 Granada. (SPAIN)

Priming Effects with Traffic Signs 18

Repetition Priming but no Semantic Priming Effect Using Worded and Symbolic Traffic Signs as Stimuli

Since the 1960s, psychologists have devoted much effort to understanding the reading processes of printed words. Such efforts have specifically dealt with semantic and repetition priming effects (see Neely (1991) for a review of literature about this effect). The semantic priming effect was already studied by Meyer and Schvaneveldt (1971). The participants were presented two stimuli consecutively and asked to press a key when the two stimuli presented were words, and a different key when one of the stimuli was a non-word, that is, an item which is syllabically correct but has no meaning. The task carried out by the participants was therefore a lexical decision task. The participants responded faster and more accurately when the two words presented were semantically related (e.g. bread and butter) than when they were not (e.g. doctor and butter).

In spite of the simplicity of the procedure, this semantic priming effect was found by a growing number of researchers, and the amount of publications showing how word recognition is influenced by semantic context increased as well.

In repetition priming, two words are also presented consecutively. The participants also carry out the same lexical decision task. They have to press a key if both stimuli are words and a different key if they are not. The participants respond faster and more accurately when both words are the same. This is due to the semantic relatedness of the word with itself, and also to its morphological and phonological identity (Warren and Morton, 1986).

The robustness of priming effects has been widely proven since then, as evidenced by the review carried out by Neely in 1991.

When exploring semantic and repetition priming effects, the conventional procedure to present the stimuli is the following: First, a stimuli (Prime) is presented, and then a second stimuli (Probe) is presented to the participants, who must give a response or carry out some kind of task, such as naming it, deciding whether it is a word, or something else.

The repetition priming effect is calculated by taking the behavioural measures (reaction time or RT and percentage of errors) obtained in the condition where the same word is presented twice (Prime and Probe), and subtracting the behavioural measures obtained in the condition where two unrelated words are presented. The semantic priming effect can be calculated by taking the behavioural measures (RT and percentage of errors) obtained in the condition where both words (Prime and Probe) are semantically related, and subtracting the behavioural measures obtained in the condition where two unrelated words are presented. This difference shows the priming effect. Thus, it is possible to obtain a priming effect when the difference is positive or an inhibition effect if the difference is negative.

We may say that, regardless of their explanation, priming effects are robust effects. The basic idea of this research is to explore their possible existence in an applied environment.

Semantic and Repetition Priming with Traffic Signs Repetition priming and semantic priming effects have rarely been studied in the context of Road Safety. On the few occasions when the effect of presenting two signs that are repeated or semantically related has been dealt with, the authors have only indirectly referred to priming effects as such, because their objective was not to analyse those effects directly. This


explains why they did not follow the classic stimuli manipulating procedure used when studying priming effects and chose very reduced samples of stimuli instead.

We are aware of the limitations of these studies, and of the great leap needed to extrapolate their conclusions to repetition and semantic priming effects. However, we shall review the procedure followed by their authors in their research as well as their success in obtaining benefits by presenting either the same sign repeatedly or two semantically related signs.

The study carried out by Milosevic and Gajic in 1986 is one of the few which have documented the effect of the repeated presentation of the same sign. They chose a reduced sample of traffic signs and situations. They used two warning signs: 'uneven road', 'road work in progress', which are both triangular, and the obligation sign 'speed limit 20 km/hour', which is round-shaped. In the first, second and third experimental situations, they explored driver recall after one of the signs mentioned above had been presented only once. In the fourth situation, they studied driver recall after the signs 'speed limit 20 km/hour' and 'road work in progress' had been presented simultaneously. The fifth and last experimental situation involved the repeated presentation of the same sign: 'speed limit 20 km/hour'. In this experimental situation, the distance between the test and the pre-test sign was 100 metres.

These authors only manipulated the repeated presentation of one of the five traffic situations they included in their experiment. In other words, they presented the sign 'speed limit 20 km/hour' twice.

The drivers drove normally along the road where the experimental arrangements had been made. After a bend, the police would stop them and a regular person would ask them which was the last sign they had seen.

Milosevic and Gajic found levels of recall that ranged from 2% to 20% when the sign was presented only once. These levels increased to about 34% when the sign was presented repeatedly. They proved that the repeated presentation of the same sign in a short time interval considerably increased sign registration. The authors explained these results by assuming that, in this case, the drivers had two chances to register the sign in the environment. Therefore, the increase in the recall of the repeated presentation of the same sign can be explained by the sum of registration probabilities.

There are also few current studies in the area of Road Safety which have studied the successive presentation of semantically related signs. The closest study to the one presented here seems to be the research carried out by Avant, Thieman, Zang and Hsu (1996).

These authors analysed the effect of the simultaneous and sequential presentation of two signs that required that the driver carry out the following actions: 'stop', 'move right', 'move left' and 'slow down'.

Two signs were always presented, either simultaneously or sequentially. The physical coincidence of the signs, their semantic coincidence and the presentation of different signs were also manipulated. They coincided physically when the same sign was presented with the same format and meaning. They coincided semantically when the same message of the sign was presented once worded and once symbolically. The signs were different when they did not coincide in format or meaning.

The data from the three experimental conditions -identical, same meaning and different- were analysed separately. No statistical analysis was carried out to indicate whether the differences between the manipulated conditions were significant.


When the participants carried out a same-different categorisation task, that is, saying whether the two signs presented were the same or different, there were faster RT and a greater accuracy when the signs were presented sequentially than when they were presented simultaneously. Simultaneous presentation slowed responses and increased errors.

The presentation of physically identical signs produced faster and more accurate responses than the presentation of signs with an identical meaning (a symbolic sign and a worded sign with the same meaning). The authors argue that having to interpret two visual codes -picture and word- to activate the same meaning slows responses and decreases the accuracy of subjects.

When unrelated signs were presented and two semantic codes represented in a symbolic and a worded form had to be processed, decision times increased considerably and accuracy decreased dramatically.

The authors concluded by insisting upon the idea that traffic signs with clearly different formats should be redesigned and that the multiple ways of expressing left and right turns, for instance, should be eliminated.

We must also remember that in the above-mentioned study by Milosevic and Gajic (1986), two semantically related traffic signs were presented in the fourth experimental situation. More specifically, two warning signs were simultaneously shown, namely 'speed limit 20 km/hour' and 'road work in progress'. The simultaneous presentation of two related traffic signs did not produce a greater recall or percentage of registration with respect to the presentation of these signs in an isolated way.

This study attempts to explore these effects in a more comprehensive way. The aim is to test the existence of these priming effects by using traffic signs as stimuli.

Traffic Sign Categories: Warning and Indication According to the Classical Theory of Concepts, an item belongs to a category if and only if it meets all the features that define that category. The inclusion of items in a category is rigid and the limits between one category and another are perfectly defined. Yet, most natural categories do not completely match this strict organisation. Philosophers such as Wittgenstein suggested the new idea of family resemblance, as an alternative to the rigid classical conceptualisation. This is the basic idea behind the Prototype Theory (Collins and Quillian, 1969; Rosch, 1973; Smith, Shoben and Ripps, 1974; McCloskey and Glucksberg, 1978; Kintsch, 1980;). Concepts are formed by using a set of central features that most of the members of a given category have in common. The items that possess many of these features are the most representative ones of such a category. The inclusion of items in a category is flexible and the boundaries between a category and another are fuzzy. In fact, the same item may belong to more than one category. For instance, 'horse' belongs to the category 'animal' as a very representative item and to the category 'means of transport' as a less representative item.

The semantic network theory developed by Collins and Quillian (1969) assumes that concepts are associated by means of hierarchical relations. Hierarchy is expressed in this model as a network that links concepts to each other. Concepts represent nodes of the network and each node is associated to a set of properties. Cognitive economy is an important feature of this model. This means that the properties that are applied to a set of concepts are stored at the highest level of the hierarchy where they are usually applicable.


Later, Collins and Loftus (1975) proposed a revised network model based on Automatic Spreading Activation. They no longer assumed an associative hierarchy between semantic nodes and replaced it with a less rigid structure. The semantic distance between items had to be understood as an excitation flow from one node to the next. They also introduced the idea of different kinds of associative ties in their network of nodes.

If we extrapolate this theoretical discussion to the classification of different traffic signs into separate categories, we must take into account that this classification is closer to the one proposed by the Prototype Theory than to that suggested by the Classical Theory. This is because not all the items of each category are equally representative of it. Not all of them meet a set of defining attributes, and there are no clear and defined boundaries between one category and another. We could therefore say that the categorisation of traffic signs matches the proposal of the Prototype Theory, according to which the same item may belong to more than one category with different degrees of representativeness, depending on its family resemblance with the members of the category. For example, 'school' is included in the category 'warning signs', whereas 'hospital' is included in the category 'indication signs'. Yet, these two items which belong to different categories according to the categorisation of traffic signs could also be members of the same category, 'buildings of a city' for example. It is important to be aware that there may exist other semantic relations between the items that belong to the different categories established for traffic signs, which are the object of this study.

Each traffic sign belongs to a semantic category depending on the message it expresses. There are four main categories in the Highway Code: warning, information and direction, prohibition or obligation and advice. Some examples of traffic signs in the warning category are 'roadwork', 'bicycle lane', 'low beams', and 'pedestrian lane'. Some information and direction signs are 'restaurant', 'telephone', 'garage' and 'hotel'. Signs expressing prohibition or obligation are 'no overtaking', 'maximum speed', 'stop' and 'no parking' among others. Advice signs are 'recommended speed 60, 70 or 80 km/h'.

In this study we have used the warning and information categories. We chose these categories first because they are the most frequent ones in traffic environments -in urban and interurban driving- and other situations, such as public buildings. Second, the importance of the message expressed by warning signs and the large amount of research carried out on the subject, with sometimes contradictory results, encouraged us to study them from a new perspective. The need to choose another semantic category led us to select information and indication signs, which have a crucial role in navigation and orientation, that is, in guiding people towards their objective. For historic reasons related to the generation of the different traffic sign categories, we chose to study indication or direction signs, the first traffic signs used (Overweg (1999)), and warning signs, the second most ancient traffic sign category. The fact that both categories express messages that are different enough, and also have a different graphic expression and colour, was also suitable for the purposes of this study.

Warning signs are expressed with a triangle with a red border that includes a black picture or word on a white background. Some of them indicate a crossroads, roundabout, bend, steep hill, narrow road, pothole, bridge, level crossing, trains or tramways crossing, traffic lights, one-way traffic, two-way traffic, detour, school or children, pedestrians, signs ahead, animals, roadwork, temporary pavement, loose chippings, ice, slippery road, low flying aircraft, quayside or riverbank, wind or other dangers.


As mentioned above, several studies aimed at analysing driver recall of signs that have recently passed have shown that recall of warning signs is quite low (Johansson and Rumar, 1966; Johansson and Backlund, 1970; Sanderson, 1974; Aberg, 1981; Drory and Shinar, 1982; Milosevic and Gajic, 1986). Drory and Shinar (1982) also underlined the alarmingly low levels of attention paid to traffic signs.

Classic studies already indicated that there was a differential recall of the different traffic signs when the drivers were stopped immediately after passing the signs (Häkkinen, 1965; Johansson and Rumar, 1966; Johansson and Backlund, 1970; Sremec, 1973; Drory and Shinar, 1982; Shinar and Drory, 1983; Milosevic and Gajic, 1986). In fact, they reported recall variation levels that ranged from 17% for the warning sign 'pedestrians' to 78% for a 'speed limit' sign. Summala and Hietamäki (1984) found that drivers only responded equally to the signs 'danger', 'children' and 'speed limit 30' in that they lifted their foot from the accelerator when they saw the sign. Yet, they corrected their response to return to the previous speed, especially in the presence of the signs 'danger' and 'children', thus disobeying the warning message of the sign. Aberg (1981) reported levels as low as 40% of correct recall for moose warning signs. That same year, Zwahlen (1981) undertook some research to explore the differences in the eye pattern in exploring traffic signs. Again, he found significant differences depending on which warning sign was presented. For example, the first fixation took place closer to the sign and the number of fixations was lower for the sign 'narrow bridge' than for the sign 'dangerous bend'. In general, most recent studies (Lajunen, Hakkarainen, and Summala, 1996) have also shown better results in terms of speed reduction with speed limit signs than with other warning signs that refer to specific situations.

Indication signs are expressed with a blue rectangle with white pictures or words. They usually have a thin white border. These signs express direction or information and allow us to be informed about a location before getting there, when we are actually at that location and afterwards. In other words, they provide directions before getting to a place, to help drivers find their way before they reach a crossing. They also provide information about the specific location of the road where the vehicle is, and confirm information after a crossing so that the driver can identify the road. There are many examples of this kind of signs, some of which refer to roads and directions (e.g., M62, Nottingham 23 M1, Marton 3), whereas others refer to the different services that drivers may need on the way (e.g., petrol station, restaurant, telephone...)

Bhise and Rockwell (1972) indicated that direction signs are placed at a double or triple distance than the legibility distance. Shinar, Rockwell and Maleki (1980) found longer fixation times for indication signs than for warning signs.

Hall, MacDonald and Rutley (1991) ran a study where they combined the use of eye movement registration and driving simulation. They found a linear relation between the time required to read a traffic sign and the number of words written on the sign. Other recent studies such as that of Agg (1994) concluded that an information overload may be a problem that direction signs must overcome in the dense traffic systems where we drive. Information overload in direction signs occurs when a sign expresses more directions or destinations than it is possible to read in the time available. Acquiring information requires time and drivers must be able to read the information on the sign as fast as they can to continue performing the remaining crucial tasks involved in driving. It is therefore essential that drivers do not waste their time reading these messages so that they can choose the appropriate route and manoeuvre properly in the time available.


Finally, this research attempts to analyse the effect of the presentation of the different semantic categories -warning and indication- on the speed of response and the accuracy of the participants in charge of performing the lexical decision and semantic categorisation tasks.

Taking the previous literature into account, it can be hypothesised that faster responses will be found for warning signs than for indication signs.

Format of Traffic Signs: Worded or Symbolic It has been proven since early studies that icons are better than text in traffic signs. In other words, the assertion known as picture superiority has been corroborated by some studies on traffic signs. Jacobs, Johnston and Cole (1975) analysed the effect of presenting both alphabetic and symbolic traffic signs. Their results showed that symbolic signs are legible from a distance twice as great as alphabetic signs because a greater visual acuity is needed to read the latter. Ells and Dewar (1979) underlined the advantages of using symbolic signs instead of worded ones (e.g., 'steep hill') for the message to be more easily understood. This is related to the time used to perform different tasks. These authors also pointed out that when signs are partially degraded, there is a greater decrease in performance in the worded version than in the symbolic one. Jacobs, Johnston and Cole (1975), Kline, Ghali, Kline and Brown (1990), and Kline and Fuchs (1993) stressed that symbolic signs may be easier to identify than worded signs. Drivers of all ages can only identify a worded sign at half the distance than a symbolic one. That is to say, if a symbolic sign can be read from a distance of 200 m, a worded sign can only be read from 100 m. The advantages of symbolic signs are ever more crucial when lighting conditions are poor or low. MacDonald and Hoffmann (1991) proved the superiority of symbolic signs over worded signs as regards their conspicuity, legibility and comprehensibility in driving situations. Edworthy and Adams (1996a) also reported the usefulness of symbolic signs in that the symbolic versions of signs are more effective than those based on text. According to these authors, the main advantages of symbolic signs over worded ones are the following: First, pictures can be recognised by people who cannot read a given language. Second, symbolic signs can be recognised from a greater distance than worded signs. Third, pictures are recognised more easily and accurately than words. Fourth, pictures can endure a greater degradation and continue to be recognisable. Finally, if pictures are used in combination with a text on the same sign, the sign is more effective than just a worded sign.

Data from gerontology studies also support the symbolic presentation of traffic signs. Ageing causes a significant reduction in visual acuity, which entails serious difficulties when it comes to reading the message of road signs (Evans and Ginsburburg, 1985) and the need to increase contrast levels in order to read the message of signs (Owsley and Sloane, 1987). This disability becomes more acute when driving at night, dusk or dawn, in bad weather or in any other condition where light is reduced (Kline, Kline, Fozard, Kosnik, Schieber and Sekuler, 1992). Older drivers require more time to process information and make decisions (Lerner, 1994). Yet, the number of elderly people who drive in developed countries is increasing (Transportation Research, 1988). In fact, elderly drivers are involved in a greater number of accidents than young drivers (Federal Highway Administration, 1989). Incorrect acquisition of information is one of the main factors involved in these accidents. Therefore, it is important to find alternative ways to facilitate the acquisition of the information conveyed by traffic signs in these ageing societies.


In contrast to what happens with worded signs, the information provided by pictograms can be clearly and easily understood by the whole population. The importance of pictures has increased in our environments as an immediate source of information, though their use is sometimes problematic as well. Bruyas, Le Breton and Pauzie (1996) insisted upon the need to combine complex information -elements and codes- to configure a symbolic sign and reduce misunderstandings or ambiguity. Kline, Ghali, Kline and Brown (1990) found a considerable variation in comprehensibility between some icons and others. They found no differences in the comprehensibility of icons when comparing drivers of all ages. However, showing complex situations such as pictures or introducing new signs can diminish the power and efficiency of symbolic presentations. Edworthy and Adams (1996a) insisted upon the same idea that complex situations cannot be easily captured in a symbolic form without a proper learning process. Therefore, the comprehensibility of the symbolic and worded signs that are currently used on roads should be assessed.

The use of symbolic traffic signs also has some negative consequences. The main reasons for this are underlined and explained by certain studies which show unfavourable results for symbolic presentations. For example, Robertson (1977) used an unfamiliar symbolic sign and compared it with its worded counterparts. The sign used referred to danger caused by following the previous driver too close, and could be called 'danger caused by breaching the safety distance'. As regards understanding the meaning of the new sign, the drivers preferred the six text-based signs rather than those based on the symbolic representation.

Other studies which have also failed to obtain a better performance with pictures than with worded signs can be criticised for having important methodological errors. For example, the lack of success of Loo (1978) in finding faster RT for pictures than for worded signs is partially due to his inappropriate manipulation of the variables. Loo's study did not use the same kind of signs in the two experimental groups. He used the picture of a train with one group of participants and the word 'school' with another group. It is obvious that factors such as frequency of use, comprehensibility, familiarity and semantic category that the words belong to may interact with the manipulation of the symbolic or worded presentation between groups.

Several pieces of research agree on the effectiveness of pictures when combined with worded presentations, though this idea is not free of criticism either, since some studies suggest that adding a second code to a traffic sign can make its message more difficult to interpret. Moore and Christie (1963) underlined the advantage of using words in combination with abstract traffic signs. Adding words to symbolic traffic presentations without reducing their visibility may help drivers understand their symbolism. As we mentioned above, Edworthy and Adams (1996a) pointed out that the use of pictures and text on the same sign may be more effective than only text. Other studies such as Young's (1991), suggest that the fact of adding a pictogram or an icon to a warning sign reduces the time required to find and recognise the danger, and thus increases the salience of the information conveyed by the warning sign. Again, these results can be explained at least by two factors: the comprehensibility of the pictures and a proper learning process. In order to achieve a good performance of symbolic presentations, two aspects are important: the picture must be known and properly understood.

Our research also attempts to explore the effect of presenting traffic signs with words or pictures, starting from the hypothesis based on previous literature that a symbolic presentation is better than a worded presentation as regards the speed of responses from the participants.


Approach of the Research

Our intention is to apply the experimental findings about the priming effect to road safety and more specifically to the presentation of traffic signs. In theory, it is possible to facilitate the processing of signs by presenting other signs previously, that is, by pre-signalling them. This procedure is already in use in certain cases such as indications of motorway exits. Yet, it is not done systematically and has no empirical base whatsoever. In our opinion, it is imperative to be aware of the conditions and the degree to which the priming effect can be obtained when using the material present on the signs. Our aim is not to verify the existence of the priming effect but rather to explore whether it is possible or not to implement it in pre-signalling activities.

This research aims specifically at analysing the effect of consecutively presenting either the same sign twice or two semantically related signs and evaluating the effectiveness of the categorisation of signs (warning and indication). Our goal is also to analyse the effect of the category that the signs belong to, using worded and symbolic representations as Primes and Probes, while the participants carry out two different tasks: a lexical decision task with words, and a semantic categorisation task with words and pictures.

The stimuli chosen are warning and indication signs from the British Highway Code. The use of the chosen stimuli is crucial for the study to have an applied interest. 'Stimuli selection' is biased by the actual categorisation of traffic signs made by engineers. Therefore, our study may also be interpreted as an analysis of the existing categories.

There are two possible ways of pre-signalling that correspond to the two above-mentioned procedures to obtain priming. The same sign can be repeated, or several different but related signs can be presented. The first method implies repetition priming, whereas the second one involves semantic priming. Both possibilities are studied in this research.

Semantic priming requires a more detailed explanation. Our approach is to invert the classic procedure of studies on semantic priming. Traditionally, the attempt was to prove the existence of the effect, and therefore the most appropriate conditions and stimuli were selected to obtain it. Our experimental logic is the opposite. We consider that the semantic priming effect has been sufficiently proven and use it to verify whether the categorisation of traffic signs made by engineers is appropriate. If the priming effect is obtained, we can consider that the categories are relevant enough from the psychological point of view to be used as pre-signalling. If the effect is not obtained, we can assert that this kind of pre-signalling would not be useful. Therefore, we use priming to assess the stimuli, and not the stimuli to assess the existence of priming.

We are also interested in finding out whether the picture superiority effect is ratified in lexical decision and semantic categorisation tasks with traffic signs.

At the same time, we also explore whether the fact of belonging to a given semantic categorisation (warning and indication) has a differential effect on the speed of response and the accuracy of the participants in these tasks.

In order to meet these objectives, a series of three experiments have been carried out. In all three, two consecutive stimuli, -words or pictures- were presented after a fixation point. The first stimulus shown was called Prime whereas the second stimulus was called Probe. The subjects carried out the task only based on the Probe. The main differences between the experiments were based on two criteria: the task performed by the participants and the kind of stimulus presentation.


The participants' task can be considered as a Lexical Decision (LD) task (in Experiment 1) or a Semantic Categorisation (SC) task (in Experiments 2 and 3). The Lexical Decision task implied deciding whether the Probe, a chain of letters, was a word or a non-word, that is, a syllabically correct but meaningless entity. Semantic Categorisation consisted of deciding whether the Probe, a chain of letters or a picture, belonged to the semantic category of traffic signs.

Table 1. Experimental Sequence.

Prime/Probe RELATEDNESS

PRIME FORMAT

PROBE FORMAT

PROBE CATEGORY

TASK PERFORMED

EXP 1 I R NR Word Picture Word Warning Indication Lexical Decision

EXP 2 I R NR Word Picture Picture Warning Indication Semantic Categorisation

EXP 3 I R NR Word Picture Word Warning Indication Semantic Categorisation

* Key to the table:

♦ Framework: Same or Different. ♦ Prime-Probe Relatedness: Identical, Related or Unrelated. ♦ Prime Format: Word or Picture. ♦ Probe Format: Word or Picture. ♦ Probe Category: Warning or Indication. ♦ Task Performed: Lexical Decision or Semantic Categorisation.

In the three experiments, the stimuli used were pictures from the Highway Code and the

words that represent these signs. Following the standard traffic rules of the British Highway Code, the warning words and pictures were always shown inserted in a white triangle with a red border, whereas the words and pictures that suggested indication were always shown inserted in a blue rectangle with a white border. In these experiments, the first stimulus was always either a word or a picture. In Experiment 1, the Probe was always a word and the participants had to perform a Lexical Decision task. In Experiment 2, the Probe was always a picture and the participants had to perform a Semantic Categorisation task. In Experiment 3, the Probe was always a word and the participants had to perform a Semantic Categorisation task (See Table 1).

Experiment 1

We analysed the effect of consecutively presenting the same sign twice and also two semantically related or unrelated signs. While the participants performed a Lexical Decision task, we assessed the effectiveness of the Warning and Indication signs and analysed the effect of presenting the Prime in words or pictures. The presentation of the Probe was always worded.

The stimuli used -warning and indication signs- were chosen from the British Highway Code and expressed in accordance with its rules. The words or pictures that implied warning were shown in black and were inserted in a white triangle with a red border, whereas the


words or pictures that implied indication were shown in white and were inserted in a blue rectangle with a narrow white border.

In this experiment, three variables were manipulated: A. Prime-Probe Relatedness, which could be: Identical (when the same sign was presented twice), Related (when two words of the same semantic category were presented), or Unrelated (when two words of a different semantic category were presented). B. Probe Category, which could be Warning or Indication. C. Prime Format, which could be Worded or Symbolic. Probe Format was always Worded. The type of task performed was always a lexical decision task (See Summary on Table 1)

Method

Participants 20 participants took part in the experiment. They were all students and staff of the University of Derby. Their ages ranged from 18 to 50. They all had normal or corrected vision. Their vision was assessed with the Titmus II Vision Screening Equipment.

Stimuli The material used in this experiment consisted of 24 words and 24 pictures: 12 conveyed Warning and 12 conveyed Indication. The Format of the Prime was either a word or a picture, but the Probe was always a word.

The words used had an equal frequency of use, following the standards proposed by Kucera and Francis (1967), and an equal syllable length. The mean syllable length was 1.58 and 2.5 for those on the warning and the indication signs respectively, and the mean frequency of use was 61.91 and 61.75 for the warning and indication words respectively.

Besides, 24 non-words were created by altering only one vowel in the real word and creating a pronounceable non-word without a meaning.

The warning signs used were: 'bend', 'bridge', 'cattle', crossroads', 'children', 'deer', 'horses', 'humps', 'junction', 'roundabout', 'train' and 'tunnel'.

The indication signs used were: 'airport', 'bus', 'camping', 'caravans', 'hospital', 'information', 'motorway', 'parking', 'petrol', 'picnic', 'restaurant' and 'telephone'.

The pairs of related stimuli were constructed with the aim of maximising the degree of association between the words and the family resemblance between the stimuli with the aim of optimising semantic priming. The words with the greatest family resemblance were matched.

The pairs of related indication signs were: 'tunnel-humps', humps-bridge', 'bridge-train', 'train-tunnel', 'cattle-horses', 'horses-deer', 'deer-children', 'children-cattle', 'crossroads-junction', 'junction-bend', 'bend-roundabout', 'roundabout-crossroads'.

Some examples of pairs of related warning signs were: 'hospital-telephone', 'telephone-bus', 'bus-restaurant', 'restaurant-hospital', 'petrol-airport', 'airport-motorway', 'motorway-parking', 'parking-petrol', 'information-camping', 'camping-caravans', 'caravans-picnic', 'picnic-information'.


24 non-words were also created from the names of the traffic signs: 12 of them were shown as Warning non-words ('bund', 'bradge', 'cattla', 'crossriad', 'choldren', 'doer', 'horso', 'himps', 'jinction', 'reundabout', 'truin', 'tennel') and 12 were presented as Indication non-words ('oirport', 'bes', 'cimping', 'carovans', 'hespital', 'infirmation', 'matorway', 'parkeng', 'potrol', 'pucnic', 'rostorant', 'teluphone').

For the construction of the pairs of unrelated stimuli, each of the warning stimuli was randomly matched with each of the indication stimuli. The pairs of stimuli formed in this manner were presented to all the participants in the same way.

An effort was made to use the highest possible number of stimuli with the aim of increasing the likeliness of obtaining the semantic priming effect. It could be considered that, by reducing the number of stimuli from each category, items with a greater family resemblance could be selected, which would make them less likely to be very representative items of other categories. Yet, some studies have proven that the use of small samples of stimuli has a detrimental effect on the semantic priming effect. This effect is almost eliminated when a high proportion of the trials contains Primes and Probes which are the same word (Snow and Neely, 1987).

All the worded stimuli were created using the PaintBrush graphic software. All the symbolic stimuli were scanned from the British Highway Code and were later retouched using PaintBrush. The stimuli were shown on a 17-inch screen. All the stimuli which conveyed Warning were shown in black and inserted in a white triangle with a thick red border, whereas the Indication stimuli were shown in white and inserted in a blue rectangle with a thin white border. The size of the triangle was a constant height of 5.71º by a base of 6.27º degrees of visual angle. The size of the rectangle was a constant vertical of 5.71º by a horizontal of 6.27º, degrees of visual angle. These shapes always were shown in the centre of the computer screen and the words or pictures on them were also centred accordingly in the triangle or rectangle. A 24-point Universe font was used to construct the words. Only the first letter of the word was written in upper case. The distance between the eyes of the subject and the computer screen was 1 metre.

Procedure Each participant carried out the experiment individually. Every person sat on a chair in front of the computer screen and performed the lexical decision task.

In each trial, two consecutive stimuli always were always shown after the fixation point. This point was a small cross in the centre of the screen and was shown for 400 ms. After that, the Prime stimulus was shown for 500 ms. I.S.I. (Inter Stimulus Interval) was 800 ms. Therefore, stimulus onset asynchrony (SOA) was 1,300 ms. Next, a word (Probe) was shown for 40 ms. After that, the screen remained black until the participants' response was recorded. If there was no response after 1,500 ms, the next trial would appear. Both the presentation of stimuli and the collection of responses were controlled by the M.E.L (Micro Experimental Laboratory, Schneider, 1988) software. The sequence of events can be seen in the following figure:


Figure 1. Sequence of events in the trials of this experiment.

We used a relatively long SOA for several reasons. We wanted an SOA level where it had been widely proven that semantic priming occurs, but we also needed it to be plausible in the traffic context and supported by the driving experience. Neely (1991, pp. 273-274 and Table 4) reviewed the data on the time course of semantic priming and concluded that it is present in long SOAs up to 2000 ms. He mentioned that the fact ‘...that these effects have been obtained in four different laboratories, with different semantic categories and with variation in the manner in which SOA was manipulated attested to their generality’. (pp. 273-274).

These are important data from an applied perspective because pre-signalling is only possible if the effect lasts for a reasonable amount of time. There is no point in pre-signalling a sign in less than 1 s. The distance covered by a driver at 50 km/h for 1 s is 14 m. At 100 km/h, the distance covered is 28 m. In practical terms, it would not be feasible to present more than one sign in shorter distances than the above.

The participants' task was to pay attention to both stimuli, and press the key 'Yes' as fast as possible if the second stimuli was a correctly spelled word or press the key 'No' if the word was not properly spelled. The response was simple and involved pressing one of these two keys. The response hand was averaged across subjects. The experiment consisted of one practice block and six experimental blocks with breaks in between. The presentation order for the stimuli was determined randomly for each block and for each subject. 48 experimental trials were carried out in each block. 24 observations were collected for each experimental condition.

Results

In order to analyse the RT response, a 3x2x2 Repeated Measures ANOVA was carried out. The manipulated variables within subject were the following: Prime-Probe Relatedness


(Identical, Related, Unrelated) X Probe Category (Warning, Indication) X Prime Format (Picture, Word). The following results were obtained (See Table 2):

Table 2. Mean RT in all the experimental conditions of Experiment 1: 3x2x2: Prime-Probe Relatedness X Probe Category X Prime Format. Lexical Decision Task. The

Probe was always worded.

Prime Format

PICTURE WORD

Probe Category Probe Category

WARNING INDICATION WARNING INDICATION

IDENTICAL 595.61 93.87

568.77 88.42

560.67 86.54

552.46 66.72

569.38 83.89

RELATED 601.89 84.35

611.46 93.57

586.46 70.95

612.25 85.21

603.01 83.52

UNRELATED 623.86 74.8

589.35 79.35

614.93 80.73

586.77 89.92

603.75 81.2

607.12 84.34

589.86 87.11

587.35 79.41

583.82 80.62

598.49 85.73

585.59 80.02

* Numbers in bold = Means, in ms; Ordinary numbers = Standard Deviations, in ms. * Significant main effects are shown in the shaded areas.

Probe Category

WARNING INDICATION

597.23 81.88

586.84 83.86

The main effect of the factor Prime-Probe Relatedness was significant, F(2,38)=16.99;

p<.0001. The fastest RT were found in the condition Identical, 569.38 ms. The slowest RT were obtained in the conditions Related, 603.01 and Unrelated, 603.75 ms. The planned comparison between the conditions Identical and Related was significant, F(1.19)=46.71; p<.0001. The planned comparison between the conditions Identical and Unrelated was also significant, F(1.19)=2.10; p<.0003. These were the only comparisons that were found to be significant.

The main effect of the factor Prime Format was significant, F(1.19)=4.99; p<.0377. The RT in the condition Word, 585.59 ms, were faster than in the condition Picture, 598.49 ms.

The main effect of the factor Probe Category was not significant, F(1.19)=3.14; p<.0924. The only interaction that was found to be significant was a first order interaction between

the factors Prime-Probe Relatedness and Probe Category, F(2.38)=11.44; p<.0001 (See Figure 2). The following comparisons were found to be significant a posteriori: The comparison between the conditions Related and Unrelated in the category Warning,


F(1.19)=11.07; p<.0035; the comparison between the conditions Identical and Unrelated both in the categories Warning, F(1.19)=16.58; p<.0006, and Indication, F(1.19)=10.05; p<.005; the comparison between the conditions Identical and Related in the category Indication, F(1.19)=11.07; p<.0001. The a posteriori comparison between the categories Warning and Indication in the condition Unrelated, F(1.19)=20.32; p<.0002, was also significant.

2-way interaction: PRIME-PROBE RELATEDNESS X PROBE CATEGORY

F(2,38)=11.44; p<.0001

PRIME-PROBE RELATEDNESS

RE

AC

TIO

N T

IME

(mse

c)

460

480

500

520

540

560

580

600

620

640

660

IDENTICAL RELATED UNRELATED

WARNING SIGNS

INDICATION SIGNS

Figure 2. Mean RT for the six experimental conditions, product of the first order interaction between the variables Prime-Probe Relatedness and Probe Category manipulated in Experiment 5. Lexical Decision Task. The Probe was always worded.

In order to analyse the accuracy response, a 3x2x2 Repeated Measures ANOVA was carried out. The variables manipulated within subject were the following: Prime-Probe Relatedness (Identical, Related, Unrelated) X Probe Category (Warning, Indication) X Prime Format (Picture, Word). The following results were obtained:

The main effect of the factor Prime-Probe Relatedness was significant, F(2.76)=8.25; p<.0011. The most accurate condition was Identical, where 95.6% of the responses were correct. The least accurate conditions were Related, with 92.26% of correct responses, and Unrelated, with 92.22% of correct responses. The planned comparison between the conditions Identical and Related was significant, F(1.19)=9.58; p<.0059. The planned comparison between the conditions Identical and Unrelated was significant, F(1.19)=10.99; p<.0036. These were the only comparisons that were found to be significant and the only significant results obtained.

Conclusions

The Repetition Priming effect was significant. The fastest RT were obtained when the same sign was presented twice. The Semantic Priming effect was also significant, but only in the


category of Warning signs, since the interaction between the variables Prime-Probe Relatedness and Probe Category was significant. The results obtained when analysing accuracy prove that the Repetition Priming effect was marginally significant, whereas the Semantic Priming effect was not replicated. No positive correlation was found between the dependent measures collected in this experiment, namely RT and accuracy.

The most striking finding of this experiment may well be the fact that Semantic Priming was obtained only when words or pictures conveying Warning were used. In those cases, the semantic relation between the items in this category seems stronger. This Semantic Priming effect was not found when words conveying indication were used. Since warning signs are crucial in driving, practical consequences should be extracted from this finding. It would be advisable to place two warning signs consecutively to succeed at warning the drivers about the danger and minimise their reaction time when they read the message of the signs. However, it remains to be seen whether this Semantic Priming effect, which is specifically found in warning signs when the participants carry out a lexical decision task, also takes place when the participants carry out other tasks. This discussion, as well as the related practical consequences, is the subject of the following experiments 2 and 3, where the participants carried out another task, that of Semantic Categorisation.

The effect of the variable Probe Category that we found was only marginally significant, though the trend indicated by the data points in the same direction. Slower RT were found with words denoting Warning than with those denoting Indication. This finding is surprising, as it is logical to expect that words conveying warning should produce faster reactions than those conveying indication. It is common practice to represent warning and indication traffic signs in interurban areas mainly as pictures and not as words. This could place the worded representation of warning and indication signs at a disadvantage. Yet, indication signs are often worded in urban areas. This may be the reason for their better results.

The effect of the type of presentation of the Prime format was also significant. Faster RT were found in the worded presentation. Faster RT were obtained when the Prime and Probe stimuli were of the same type (word-word) than when they were of a different type (picture-word). Within the 'word-word' type, the fact that the meaning and format are the same increases the subjects' preparation and causes a decrease in RT. In the 'picture-word' types, equal meanings but different formats do not increase the subjects' preparation as much.

Since there is no significant interaction between the variables Prime-Probe Relatedness and Prime Format, the Repetition Priming effect is found both within types (word-word) and between types (picture-word), whereas, in general, the Semantic Priming effect is not found. This effect is only found in warning signs. The strong 'word-word' and weak 'picture-picture' semantic priming effects found within types by others (Sperber, McCauly, Ragain and Weil, 1979; Carr, McCauly, Sperber, Parmelee, 1982; Kroll and Potter, 1984) are not replicated in this research using traffic signs as stimuli. The apparent asymmetry found by other researchers (Durso and Johnson, 1979; Sperber et al., 1979; Scarbourg, Gerald and Cortese, 1979; Car et al., 1982), who did not find this effect in the 'picture-word' types, is replicated in this experiment indeed. According to these authors, when the task requires phonetic processing, no picture-word priming is found.


Experiment 2

This study analysed the effect of consecutively presenting the same sign twice and also two signs that are either semantically related or unrelated. While the participants carried out a Semantic Categorisation task, we also assessed the effectiveness of Warning and Indication signs, and analysed the effect of presenting the Prime in words or symbols. The presentation of the Probe was always symbolic.

As in Experiment 1, the stimuli used were warning and indication signs from the British Highway Code. They all were expressed in accordance to the rules spelled out in the code.

Three variables were therefore manipulated in this experiment: A. Prime-Probe Relatedness, which could be: Identical (when the same sign was presented twice), Related (when two words of the same semantic category were presented) or Unrelated (when two words of different semantic categories were presented). B. Probe Category, which could be Warning or Indication. C. Prime Format, which could be Worded or Symbolic. Whereas the presentation of the Probe was always symbolic, the type of task carried out was always semantic categorisation.

Method

Participants 20 participants carried out the experiment. All of them were students or staff of the University of Derby. Their ages ranged from 18 to 50. They all had normal or corrected vision. Their vision was assessed with the Titmus II Vision Screening Equipment.

Stimuli The material used in this experiment was 24 words and 24 pictures: 12 of them conveyed Warning and 12 conveyed Indication. The Format of the Prime was either a Word or a Picture, but the Probe was always a Picture.

The stimuli were elaborated in the same way explained above for Experiment 1. The warning and indication signs used were the same ones used in the previous experiment.

Besides, 24 traffic non-signs were created. They were simple pictures similar to those in the Highway Code: 12 of them were shown as Warning non-signs ('ball', 'cat', 'disk', 'fish', 'glasses', 'mouse', 'pound', 'sun', 'clover', 'ampersand', 'watch', 'window') and 12 were presented as Indication non-signs ('computer', 'dollar', 'hand', 'happy', 'heart', 'letter', 'moon', 'pencil', 'percent', 'scissors', 'shoes', 'time').

For the construction of the pairs of unrelated stimuli, each of the warning stimuli was randomly matched with each of the indication stimuli. The pairs of stimuli formed this way were presented to all the participants in the same way.


Procedure The procedure, the time sequence of events in each trial, the number of trials carried out, the number of observations per experimental condition and the experimental controls carried out were identical to those described in the previous experiment.

However, the participants' task was to perform a semantic categorisation task. They had to say whether the pictures that were shown on the screen as Probes were traffic signs or not.

Results

In order to analyse RT, a 3x2x2 Repeated Measures ANOVA was carried out. The variables manipulated within subject were the following: Prime-Probe Relatedness (Identical, Related, Unrelated) X Probe Category (Warning, Indication) X Prime format (Picture, Word). The following results were obtained (See Table 3):

The main effect of the factor Prime-Probe Relatedness was significant, F(2.38)=12.21; p<.0001. The fastest RT were found in the condition Identical, 479.41 ms. The slowest RT were obtained in the conditions Related, 502.05 ms, and Unrelated, 505.83 ms. The planned comparison between the conditions Identical and Related was significant, F(1.19)=10.06; p<.005. The planned comparison between the conditions Identical and Unrelated was also significant, F(1.19)=31.48; p<.0001. These were the only comparisons found to be significant.

The main effect of the factor Probe Category was significant, F(1.19)=14.11; p<.0013. The fastest RT, 485.19 ms, were found in the condition Warning, whereas the slowest RT, 506.33 ms, were obtained in the condition Indication.

The main effect of the factor Prime Format was significant, F(1.19)=8.22; p<.0099. The fastest RT, 488.76 ms, were found in the condition Picture, whereas the slowest RT, 502.77 ms, were found in the condition Word.

These were the only results found to be significant. To analyse the accuracy response, a 3x2x2 Repeated Measures ANOVA was carried out.

The variables manipulated within subject were the following: Prime-Probe Relatedness (Identical, Related, Unrelated) X Probe Category (Warning, Indication) X Prime format (Picture, Word). The results obtained were the following:

The main effect of the factor Probe Category was significant, F(1.19)=21.31; p<.0002. The most accurate condition was Warning, with 98.77% of correct responses. The least accurate condition was Indication, with 95.15% of correct responses.

These were the only results which were found to be significant.


Table 3. Mean RT in all the experimental conditions of Experiment 2: 3x2x2: Prime-Probe Relatedness X Probe Category X Prime Format. Semantic Categorisation Task.

The Probe was always symbolic.

Prime Format

PICTURE WORD




474.13 88.25

481.93 85.41

502.01 82.27

479.41 83.13

RELATED 483.32

85.5 505.16

69 502.63 83.99

517.11 70.19

502.05 77.17


518.38 65.49

491.72 76.42

521.24 73.08

505.83 72.87

478.3 79.53

499.22 74.24

492.09 81.94

513.45 75.18

488.76 76.89

502.77 78.56

* Numbers in bold = Means, in ms; Ordinary numbers = Standard Deviations, in ms. * Significant main effects are shown in shaded areas.

Probe Category

WARNING INDICATION

485.19 80.73

506.33 74.71

Conclusions

First of all, when analysing RT, the Repetition Priming effect was found again, but not the Semantic Priming effect. The fastest RT were found when the same sign was shown twice. These results are congruent with those of Experiment 4, and partially congruent with those of Experiment 5. Therefore, when the participants carry out the semantic categorisation task, no semantic priming effect is found. This suggests once again that the categories that traffic signs belong to are artificial and that the ties between their items are weak if compared to those in other categories these items may belong to.

Second, RT for the pictures conveying Warning was faster than for the pictures which conveyed Indication. Moreover, when accuracy was analysed, this effect found in RT was replicated. In other words, accuracy was greater in the case of pictures conveying Warning than for pictures conveying Indication. Therefore, pictures seem to be the best presentation form to maximise the warning message. If we compare this result with those obtained in the


previous experiments, we can conclude that the worded or symbolic presentation format seems to interact with the category of the stimulus: warning or indication. Worded presentation favours indication signs whereas a symbolic presentation reinforces warning signs. The fact that warning signs tend to be presented symbolically in interurban areas and that indication signs are usually worded in urban areas may explain this effect. At the same time, practical consequences should be extracted in order to minimise the reaction time of the participants and maximise their accuracy when confronted to these different signs.

Third, Prime format was also significant when we analysed RT. Faster RT were found for the symbolic presentation. When the Prime and Probe stimuli were of the same type (picture-picture) there were faster RT than when they were of different types (word-picture). These results replicate those found in the previous experiment in a different modality. In this case the Format of the Probe was always a word. Within the 'picture-picture' types, the fact that the meaning and format were the same increased the subjects' preparation and caused a decrease in RT. In the 'word-picture' types, equal meanings but different formats did not increase the subjects' preparation as much.

Since there was no significant interaction between the variables Prime-Probe Relatedness and Prime Format, the Repetition Priming effect was found both in the 'picture-picture' and in the 'word-picture' types. The Semantic Priming effect was not found in either case. The strong word-word and weaker picture-picture semantic priming effects that have been found within types by some authors (Sperber, McCauly, Ragain and Weil, 1979; Carr, McCauly, Sperber, Parmelee, 1982; Kroll and Potter, 1984) are not replicated in this research either using traffic signs as stimuli. The results obtained by other researchers (Durso and Johnson, 1979; Sperber et al., 1979; Scarbourg, Gerald and Cortese, 1979; Car et al., 1982) who found equivalent priming effects for both 'picture-word' and 'word-picture' stimuli types when the participants performed tasks that required semantic processing are not replicated either.

It remains to be explored whether a priming effect occurs between the 'picture-word' stimulus types in a situation in which the participants perform a semantic categorisation task with words. This is the idea that we tested in the last experiment of this series.

Experiment 3

This study analysed the effect of consecutively presenting the same sign twice and also two signs that are either semantically related or unrelated. While the participants carried out a Semantic Categorisation task, we also assessed the effectiveness of Warning and Indication signs, and analysed the effect of presenting the Prime in words or symbols. The presentation of the Probe was always worded.

As in Experiment 1, the stimuli used -warning and indication signs- were chosen from the British Highway Code and were expressed in accordance with the rules spelled out in the code.

Three variables were manipulated in this experiment: A. Prime-Probe Relatedness, which could be: Identical (when the same sign was presented twice), Related (when two words of the same semantic category were presented) or Unrelated (when two words of a different semantic category were presented). B. Prime Category, which could be Warning or Indication. C. Prime Format, which could be Worded or Symbolic. Probe Format was always Worded, whereas the type of task carried out was always semantic categorisation.


Method

Participants 20 participants carried out the experiment. All of them were students or staff of the University of Derby. Their ages ranged from 18 to 50. They all had normal or corrected vision. Their vision was assessed with the Titmus II Vision Screening Equipment.

Stimuli The material used in this experiment was 24 words and 24 pictures: 12 of them conveyed Warning and 12 conveyed Indication. The Format of the Prime stimulus was either a Word or a Picture, but the Probe stimulus was always a word. As in the previous experiments, the words had the same frequency of use, following the standards proposed by Kucera and Francis (1967), and also the same length.

24 more words were used to represent the traffic non-signs used in Experiment 2. The stimuli were elaborated in the same way as explained above for the previous

experiments.

Procedure The procedure, the time sequence of events in each trial, the number of trials carried out, the number of observations per experimental condition and the experimental controls carried out were similar to those described in the two previous experiments.

However, the participants had to perform a semantic categorisation task. They had to say whether the words that were shown on the screen as Probes were traffic signs or not.

Results

In order to analyse the RT, a 3x2x2 Repeated Measures ANOVA was carried out. The variables manipulated within subject were the following: Prime-Probe Relatedness (Identical, Related, Unrelated) X Probe Category (Warning, Indication) X Prime Format (Picture, Word). The following results were obtained (See Table 4):

The main effect of the factor Prime-Probe Relatedness was significant, F(2.38)=11.26; p<.0001. The fastest RT were found in the condition Identical, 581.82 ms. The slowest RT were obtained in the conditions Related, 605.3 ms, and Unrelated, 611.58 ms. The planned comparison between the conditions Identical and Related was significant, F(1.19)=14.99; p<.001. The planned comparison between the conditions Identical and Unrelated was also significant, F(1.19)=13.76; p<.0014. These were the only comparisons which were found to be significant.

The main effect of the factor Category of the stimulus was significant, F(1.19)=13.06; p<.0019. The fastest RT, 589.73 ms, were found in the condition Indication. The slowest RT, 610.71 ms, were found in the condition Warning.


The main effect of the factor Prime Format was significant, F(1.19)=6.95; p<.0163. The fastest RT, 595.12 ms, were found in the condition Word. The slowest RT, 605.35 ms, were obtained in the condition Picture.

These were the only results found to be significant. In order to analyse the accuracy response, a 3x2x2 Repeated Measures ANOVA was

carried out. The variables manipulated within subject were the following: Prime-Probe Relatedness (Identical, Related, Unrelated) X Probe Category (Warning, Indication) X Prime Format (Picture, Word).

No results were found to be significant. A mean of 95% of the cases were correct.

Table 4. Mean RT in all the experimental conditions of Experiment 3: 3x2x2: Prime-Probe Relatedness X Probe Category X Prime Format. Semantic Categorisation Task.

The Probe was always worded.

Prime Format

PICTURE WORD




579.59 89.28

585.94 110.72

569.57 92.2

581.82 94.49

RELATED 612.54 83.19

601.08 86.43

619.07 91.98

596.51 98.88

607.3 90.12


617.46 91.68

625.31 90.19

574.30 84.2

611.58 89.96

611.33 87.58

599.37 89.13

610.1 97.63

580.1 91.76

605.35 88.35

595.12 94.69

* Numbers in bold = Means, in ms; Ordinary numbers = Standard Deviations, in ms. * Significant main effects are shown in shaded areas.

Probe Category

WARNING INDICATION

610.71 92.6

589.73 90.44

Conclusions

Again, the Repetition Priming effect was found, and the Semantic Priming effect was not found. There were faster RT when the same sign was shown twice. These results are congruent with the previous research -Experiment 2- and partially congruent with those obtained in Experiment 1. Therefore, when the participants carried out the semantic


categorisation task with words, the semantic priming effect was not found. We can affirm that the unnatural categorisation established for traffic signs is based on weak semantic relations between its items, and that the items are in some cases more typical items in other categorisations.

Second, RT were faster for words conveying Indication than for words conveying warning. This result is congruent with those obtained in Experiment 1. Therefore, in this context, pictures seem to be the best way to convey warning messages, whereas words seem to be the best way to convey indication messages. We can argue again that there is an interaction between the category (warning-indication) and the presentation (worded-symbolic) of the stimuli. The fact that warning signs are usually symbolic whereas indication signs are usually worded may explain this finding. As these data indicate, it is advisable to follow this trend.

Third, Prime format produced a significant effect. Faster RT were found when the Prime format was a word than when it was a picture. When the Prime and the Probe were of the same type (word-word), faster RT were obtained than when they were of different types (picture-word). This result replicates the data found in the first experiment (using a different task) and in the second experiment (using a different Probe Format). Within the 'word-word' type, the fact that the meaning and format are the same increases the subjects' preparation and causes a decrease in RT. In the 'picture-word' types, equal meanings but different formats do not increase the subjects' preparation as much.

Since there is no significant interaction between the variables Prime-Probe Relatedness and Prime Format, the Repetition Priming effect is found both within types (word-word) and between types (picture-word). The Semantic Priming effect, however, is not found in either case. The strong word-word and weaker picture-picture semantic priming effects found by other authors within types (Sperber, McCauly, Ragain and Weil, 1979; Carr, McCauly, Sperber, Parmelee, 1982; Kroll and Potter, 1984) are not replicated in this research using traffic signs as stimuli either. The results obtained by other researchers (Durso and Johnson, 1979; Sperber et al., 1979; Scarbourg, Gerald and Cortese, 1979; Car et al., 1982) who found equivalent priming effects for both 'picture-word' and 'word-picture' types of stimuli when the participants performed tasks that involved semantic processing are not replicated either.

General Discussion

According to the results obtained, participants recognise a sign faster and more accurately when it is presented repeatedly. Yet, this priming effect is not found when two signs of the same category -indication or warning- are presented.

The repeated appearance of signs decreases the time needed to react to them. However, drivers do not react faster when two signs of the same category are presented consecutively. The fact that there is no semantic priming may be due to the fact that these indication and warning categories are not natural. They have only been used frequently for a few decades. The engineers who designed them have a clear theoretical idea or which items belong to each category, which format, colour and form they are expressed in, and what actions should be derived from their message. In practice, however, common users are not too familiar with the colours ascribed to such categories, or perhaps even ignore them. The categorisation of traffic signs, as happens with most everyday categories, is closer to the proposal of the prototypical


view or the probabilistic approach to concepts (Rosch and Mervis, 1975) than to the traditional idea of concepts. The same item may belong to more than one category with different levels of representativeness, depending on its family resemblance with the members of the category. Items which have the greatest number of attributes that define the category are the most typical items in it. These categories are said to be poorly defined because the boundaries between one category and another are fuzzy. Some items may belong to several categories, with different levels of representativeness in each of them. Within the category 'indication signs', there are items which belong to other categories as well, such as 'restaurant', 'picnic', 'camping', 'caravans', which may also belong to the category 'places of leisure'. Within warning signs, 'deer', 'cattle' and 'horses', for instance, are animals. To make matters worse, there are some categories other than those of warning and indication traffic signs that group some of its items. For example, 'train', a warning sign, and 'airport' and 'bus', indication signs, may be included in the category 'means of transport'.

It is important to be aware that other semantic relations may exist between the items which belong to the categories indication and warning signs and which are the object of this study.

Another interesting finding which was also somewhat unexpected is that RT for words that convey warning were significantly slower than for words conveying indication when the task was performed with words. The effect was the opposite when the task was performed with pictures. Warning signs are processed faster than indication signs if they are presented symbolically, whereas indication signs are processed faster when they are worded.

It might be expected that warning signs should always produce faster reactions because their message urges us to make certain decisions and /or take action to avoid the danger they warn about, whereas indication signs do not usually require a rapid response.

Our data show that Warning signs are more effective than Indication signs if they are shown symbolically but not when they are worded. This partly contradicts the previous literature, which stated the superiority of pictures (among others, Jacobs, Johnston and Cole, 1975; Ells and Dewar, 1979; Evans and Ginsburburg, 1985; Owsley and Sloane, 1987; Kline, Ghali, Kline and Brown, 1990; MacDonald and Hoffmann 1991; Kline, Kline, Fozard, Kosnik, Schieber and Sekuler, 1992; Kline and Fuchs, 1993) and the superiority of warning signs (among others, Shinar, Rockwell and Maleki, 1980).

This effect may be due to the fact that warning signs tend to be expressed symbolically in interurban settings whereas indication signs tend to be either worded or symbolic.

In any case, what is most important is that it is possible to extract applied conclusions from these data. Taking into account the interaction found between the information presented and its format, it would be advisable to continue the above-mentioned trend in presenting traffic signs.

Finally, an easily expected effect was found. When the stimuli are of the same type (word-word or picture-picture), RT are faster than when they are of a different type (picture-word or word-picture). Within types, the fact that the meaning and format are the same increases the subjects' preparation and causes a decrease in RT. Between types, equal meanings but different formats do not increase the subjects' preparation as much, hence the appropriateness of taking advantage of the repetition priming effect by maintaining the same stimulus type to achieve faster reactions.


Conclusions

As regards real word implementation, we may affirm that presenting two identical signs consecutively in the traffic environment is more useful than presenting signs with semantically related but not identical messages. The repetition priming effect should be taken advantage of when arranging traffic signs. The semantic priming effect only seems to occur in the specific case of warning signs when participants perform the lexical decision task. If the data are appreciated globally, the semantic priming effect is not found. This may be due to the fact that the categorisation of traffic signs is artificial. The items which belong to the categories indication or warning bear a great resemblance with each other and belong to other natural categories, with an even greater representativeness sometimes.

The data also demonstrate that the repetition priming effect is valid both with warning and indication traffic signs. Given the relevance of both categories, it is pivotal to apply this result to the real driving environment in order to provide information aimed at optimising drivers' perception of their environment and allowing them to anticipate the required manoeuvres. This can be achieved by pre-signalling the warning and indication signs that are considered most important. Such signs must be designed in a similar format. We insist that such format must be as identical as possible.

In fact, the concept of pre-signalling has already been used in some ways, to indicate motorway exits, for example. In the first sign, an arrow indicates the lane that the driver must choose to exit the motorway. It informs that it will happen 1 km from the first sign. The second sign informs the driver that the exit is 500 metres away. The third sign indicates that the exit is 100 metres away. The last sign indicates the destination in a frame with a different colour. Though this approach has showed to be useful because it helps the driver anticipate a future manoeuvre, our data allow us to take one step further. The repetition priming effect can be applied not only in these cases but is also useful in the case of an immediate manoeuvre. The effectiveness of a specific sign can be increased if the sign is shown repeatedly.

Presenting signs of the same category which are not identical does not seem to carry so many advantages as regards increasing the recognition speed and the accuracy in identifying the sign. This may be due to the fact that traffic signs have been categorised in an artificial way.

It is also necessary to draw practical conclusions from the surprising finding that challenges the classic idea about picture superiority. Pictures are better only at expressing information about danger, but they are counterproductive when the aim is to express indication messages. A worded presentation favours indication.

Immediate actions are derived from these questions. It is necessary to stress the importance of creating natural categories so that drivers can conceive the different signs or items as members of the same category. This need of creating natural categories may lead us for instance to establish a distinction between signs expressing obligation and prohibition. Though they are often used to convey similar messages, the attributes that define them are very different. Let us look at an example: The message of the obligation sign 'it is obligatory to turn right or go straight ahead' and the prohibition sign 'no left turn' express the same information, although with very different connotations. The obligation sign expresses the message in a positive way, whereas the prohibition sign uses negation. At the same time, both signs also express a different coercive function.


In the research work that we are currently planning, we expect to obtain semantic priming not between the traditional categories, but, following the theory of Phillip Johnson-Laird, between the mental models that they originate, with categories and stimuli which are really used in traffic. It is clearly possible to open a new line of research where many of these aspects are studied in depth.

In conclusion, it is necessary to find appropriate guidelines to divide or categorise the set of traffic signs so that they can express their message in a more effective way.

Authors Note

We would like to acknowledge the members of the AVRU (Applied Vision Research Unit) University of Derby, for their co-operation in running these experiments, as well as Alastair Gale, the director of the unit, for his valuable comments on its design and his co-operation in making this research possible.

References

Aberg, L. (1981). The human factor in game-vehicle accidents: a study of drivers' information acquisition, Studia Psychological Upsaliensa, 6, (Uppsala)

Avant, L.L., Thieman, A.A., Zang, A.L., and Hsu, S.Y. (1996). Memory codes for traffic sign information: Visual vs. meaning codes. En A.G. Gale, I.D. Brown, C.M. Haslegrave, S. Moorhead, S.P. Taylor (Eds.), Vision in Vehicles, V. Amsterdam: Elsevier.

Becker, C. A. (1985). What do we really know about semantic context effects during reading? In D. Besner, T. G. Waller and E.M. MacKinnon (Eds.) Reading research: Advances in theory and practice. Vol. 5. Toronto: Academic Press.

Becker, C.A. (1979). Semantic context and word frequency effects in visual word recognition. Journal of Experimental Psychology: Human, Perception and Performance, 5, 252-259.

Becker, C.A. (1980). Semantic context effects in visual word recognition: An analysis of semantic strategies. Memory and Cognition, 8, 493-512.

Becker, C.A., and Killion, T.H. (1977). Interaction of visual and cognitive effects in word recognition. Journal of Experimental Psychology: Human, Perception and Performance, 3, 389-401.

Besner, D., and Swan, M. (1982). Models of lexical access in visual word recognition. Quarterly Journal of Experimental Psychology, 34A, 313-325.

Bhise, V.D., and Rockwell, T.H. (1972). Development of a Methodology for evaluating road signs. Engineering Experiment Station, Final Report, EES 315B. The Ohio State University.

Bruyas, M.P. Le Breton, B., and Pauzie, A. (1996). Iconic Information Presentation: Pictograms understanding, study about ambiguity. In A.G. Gale, I.D. Brown, C.M. Haslegrave, S.P. Taylor (Eds.), Vision in Vehicles, V. Amsterdam: Elsevier.

Carr, T.H., McCauly, C., Sperber, R.D., and Parmelee, C.M. (1982). Words, pictures and priming: On semantic activation, conscious identification and the automaticity of


information processing. Journal of Experimental Psychology: Human perception and performance, 8, 757-777.

Castro, C. and Martos, F.J. (1995). Effect of the size, number and position of traffic signs mounted on the same post. Scientific Contributions to General Psychology, 13, 89-100.

Cole, B.L., and Hughes, P.K. (1990). Drivers don't search: They just notice. In D. Brogan (Ed.), Visual Search. London: Taylor & Francis.

Cole, B.L., and Jenkins, S.E. (1982). Conspicuity of traffic control devices. Australian Road Research, 12, 221-238.

Collins, A.M. and Loftus, E.F. (1975). A spreading-activation theory of semantic processing. Psychological Review, 82, 407-428.

Collins, A.M., and Quillian, M.R. (1969). Retrieval time from semantic memory. Journal of Worded Learning and Worded Behaviour, 8, 240-247.

Cooper, B.R. (1988). A comparison of different ways of increasing traffic sign conspicuity. Research Report, 157. Crowthorne: Transport and Road Research Laboratory.

Cooper, B.R. (1989). Comprenhension of traffic signs by drivers and non-drivers. Research Report, 167. Crowthorne: Transport and Road Research Laboratory.

Chapanis, A. (1994). Hazards associated with three signal word and four colours on warning signs. Ergonomics, 37, 265-275.

Chiarello, C., Senehi, J., and Nuding, S. (1987). Semantic priming with abstract and concrete words: A differential asymmetry may be post-lexical. Brain and Language, 31, 43-60.

Dannenbring, G.L., and Briand, K. (1982). Semantic priming and the Word repetition effect in a Lexical Decision task. Canadian Journal of Psychology, 36, 435-436.

De Groot, A.M.B. (1984). Primed Lexical Decision: Combined effects of proportion of related prime-target pairs and the stimulus-onset asynchrony of prime and target. Quarterly Journal of Experimental Psychology, 36A, 253-280.

Den Heyer, K., Goring, A., and Dannenbring, G.L. (1985). Semantic priming and word repetition: The two effects are additive. Journal of Memory and Language, 24, 600-716.

Drory, A., and Shinar, D. (1982). The effect of roadway environment and fatigue on sign perception. Journal of Safety Research, 21, 25-32.

Durgunoglu, A.Y. (1988). Repetition, semantic priming, and stimulus quality: Implications for the interactive-compensatory reading model. Journal of Experimental Psychology: Learning, Memory and Cognition, 14, 590-603.

Durso, F.T., and Johnson, M.K. (1979). Priming in naming and categorising repeated pictures and words. Journal of Experimental Psychology: Learning, Memory and Cognition, 5, 449-459.

Ells, J.G., and Dewar, R.E. (1979). Rapid comprehension of worded and symbolic traffic sign messages. Human Factors, 21(2), 161-168.

Evans, D.W., and Ginsburburg, A.P. (1985). Contrast sensitivity predicts age-related differences in highway-signs discriminability. Human Factors, 27, 637-642

Federal Highway Administration. (1989). The federal highway administration's action plan for older persons. Washington, DC: U.S. Department of Transportation.

Häkkinen, S. (1965). Perception of highway traffic signs. Report 1. Helsinki, TAIJA. Hall, R.D., MacDonald, M., and Rutley, K.S. (1991). An experiment to assess the reading

time of direction signs. In A.G. Gale, I.D. Brown, C.M. Haslegrave, S. Moorhead, S.P. Taylor (Eds.), Vision in Vehicles, V. Amsterdam: Elsevier.


Hoskovec, J., Stikar, J. and Raouf, A. (1974). The effectiveness of vertical traffic signs. Studia Psychologica, 16(3), 220-223.

Jacobs, R.J., Johnston, A.W., and Cole, B.L. (1975). The visibility of alphabetic and symbolic traffic signs. Australian Road Research, 5(7), 68-73.

Johansson, G., and Backlund, F. (1970). Drivers and road signs. Ergonomics, 13, 749-759. Johansson, G., and Rumar, K. (1966). Drivers and road signs. Ergonomics, 9, 57-62. Kintsch, W. (an1980). Semantic memory: A tutorial. In R.S. Nickerson (De), Attention and

performance, 7. Hillsdale, N.J.: Lawrence Erlbaum Associates Inc. Kline, D.W., Kline, T.J.B., Fozard, J.L., Kosnik, W., Schieber, F., and Sekuler, R. (1992).

Vision, aging, and driving: The problems of older drivers Journal of Gerontology: Psychological Sciences, 47, 27-34.

Kline, D.W., and Fuchs, P. (1993). The visibility of symbolic highway signs can be increased among drivers of all ages. Human Factors, 35(1), 25-34.

Kline, T.J.B., Ghali, L.M., Kline, D., and Brown, S. (1990). Visibility distance of highway signs among young, middle-aged, and older observers: Icons are better than text. Human Factors, 35(5), 609-619.

Kroll, J.F., and Potter, M.C. (1984). Recognising words, pictures and concepts: A comparison of lexical, object and reality decisions. Journal of Worded Learning and Worded Behaviour, 23, 39-66.

Kucera, H., and Francis, W.N. (1967). Computational analysis of present-day. American English. Providence: Brown University Press.

Lajunen, T., Hakkarainen, P., and Summala, H. (1996). The ergonomics of road signs: explicit and embedded speed limits. Ergonomics, 39(8), 1069-1083.

Lees, J., and Farman, M. (1970). An Investigation of the Design of Performance of Traffic Control Devices. Visible Language, 4, 7-38.

Lerner, N. (1994). Giving the older driver enough perception-reaction time. Experimental Aging Research, 20, 25-33.

Loo, R. (1978). Individual differences and the perception of traffic signs. Human Factors, 20(1), 65-74.

Lorch, R.F., Balota, D., and Stamm, E. (1986). Locus of inhibition effects in the priming of Lexical Decision: Pre- o post-lexical access? Memory and Cognition, 14, 95-103.

Luoma, J. (1986). The acquisition of visual information by the driver: Interaction of relevant and irrelevant information. Report 32. Helsinky: The Central Organisation for Traffic Safety.

Lupker, S.J. (1984). Semantic priming without association: A second look. Journal of Worded Learning and Worded Behaviour, 23, 709-733.

MacDonald, W.A. and Hoffmann, E.R. (1991). Drivers' awareness of traffic sign information. Ergonomics, 34(5), 585-612.

Martinez, R. (1988). Nads Current Events studies. Freightliner to donate truck cab, giving boost to simulator safety. Web page address: Http://www.nhtsa.dot.gov:80/people/ perform/nads/curevnts.htm

McClelland, J.L., Rummelhart, D.E., and the PDP Research Group (1986). Parallel distributed processing. Explorations in the microstructure of cognition. Vol. 2: Psychological and biological models. Cambridge, MA: MIT press.


McClelland, J.L., and Rumelhart, D.E. (1981). An interactive model of context effects in letter perception: Part 1. An account of basic findings. Psychological Review, 88, 375-405.

McCloskey, M. E., and Glucksberg, S. (1978). Natural categories: Well defined o fuzzy sets? Memory and Cognition, 6, 462-472.

Meyer, D.E., and Schavaneldt, R.W. (1971). Priming in recognising pairs of words: Evidence of a dependence between retrieval operations. Journal of Experimental Psychology, 90, 227-234.

Milosevic, S., and Gajic, R. (1986). Presentation factors and driver characteristics affecting road-sign registration. Ergonomics, 29, 325-335.

Moore, R.L., and Christie, A.W. (1963). Research on Traffic Signs. Road Research Laboratory. Proceedings of the Engineering for Traffic Conference.

Neely, J.H. (1991). Semantic priming effects in visual word recognition: A selective review of current findings theories. En D. Besner and G.W. Humphreys (Eds.), Basic processes in reading: Visual recognition of writing words. Hillsdale, N.J.: Lawrence Erlbaum Associates.

Owsley, C.E., and Sloane, M.E. (1987). Contrast sensitivity, acuity, and the perception of “real word” targets. British Journal of Ophthalmology, 71, 791-796.

Rockwell, T.H. (1972). Skills, judgement, and information acquisition in driving. In T.W. Forbes (De.), Human factors in highway research. New York: Willey.

Rosch, E. (1973). Natural categories. Cognitive Psychology, 4, 348-349. Rosch, E., and Mervis, C.B. (1975). Family resemblances: Studies in the internal structure of

categories. Cognitive Psychology, 7, 573-605. Sanderson, J.E. (1974). Driver recall of roadside signs. Traffic Research Report, 1. Ministry

of Transport. New Zealand: Wellington. Schneider, W. (1988). Micro Experimental Laboratory: An integrated system for IMB PC

compatibles. Behaviour Research Methods, Instruments and Computers, 20, 206-271. Schreuder, R., Flores d'Arcais, G.B., and Glazenborg, G. (1984). Effects of perceptual

conceptual similarity in semantic priming. Psychological Research, 45, 339-354. Shinar, D., Rockwell, T.H., and Malecki, J.A. (1980). The effects of changes in driver

perception on rural curve negotiation. Ergonomics, 23, 263-275. Shinar, D., and Drory, A. (1983). Sign registration in daytime and night-time driving. Human

Factors, 25, 117-122. Smith, E.E., Shoben E.J., and Ripps, L.J. (1974). Structure and process in semantic memory:

A featural model for semantic decision. Psychological Review, 81, 214-241. Snow, N., and Neely, J. H. (1987, Noviembre). Reduction of semantic priming from inclusion

of physically or nominally related prime-target pairs. Paper presented at the meeting of the Psychonomics Society, Seattle, WA.

Sperber, R.D., McCauly, C., Ragain, R.D., and Weil, C.M. (1979). Semantic priming effects and word processing. Memory and Cognition, 7, 339-345.

Sperber, R.D., McCauly, C., Ragain, R.D., and Weil, C.M. (1979). Semantic priming effects and word processing. Memory and Cognition, 7, 339-345.

Sremec, B. (1973). Perception of a traffic sign and driving speed. In First International Conference on Driver Behaviour, Zurich.

Summala, H., and Hietamäki, J. (1984). Drivers' immediate responses to traffic signs. Ergonomics, 27, 205-216.


Warren, C.E.J., and Morton, J. (1986). The effects of priming on picture recognition. British Journal of Psychology, 73, 117-130.

Young, S.L. (1991). Increasing the noticeability of Warnings: Effects of pictorial, color, signal icon and border. Proceedings of the Human Factor Society 35th Annual Meeting.

01 Castro PC

Documents

symbolic signs

symbolic traffic signs

recognition of traffic

repetition priming effects

presignalling traffic

spain priming effects

semantic relationships

worded format