Spatial working memory and gender differences in Science.

One reason for the lack of female participation in science could be due to cognitive differences between males and females. The present study measured verbal and spatial working memory for 15 males and 48 females. Males were found to have both a larger verbal memory and a larger spatial memory. Participants then read texts that either presented the information in both the text and diagram, or in only the text or only the diagram. Recall and question answering data found that males comprehended the material better than females. It was also found that information from the text was remembered better than information from the diagram. The results were explained in terms of working memory span and comprehension.

**********

Although females have made great strides in recent years in catching up to males in science participation, science is still a male dominated discipline. In 1994, 74% of the students taking the Physics Advanced Placement portion of the SAT were male. We see the same discrepancy in education-42% of the two-year degrees in the physical sciences go to females, 32.6% of Bachelor degrees, 29% of Masters degrees, and 22% of Doctorates in the physical science are awarded to females (U.S. National Center for Educational Statistics, 1993). This gender difference increases when we look at who is actually working in the scientific fields. Females only comprise 27.3% of the natural scientists, and 31.9% of the chemists (U.S. Bureau of Labor, 1995). The overall picture that emerges from these figures is that males outperform females on tests of scientific achievement, are more likely to pursue a degree is science, and are more likely to actually work in the scientific fields.

At least two main explanations are offered as to why males outperform females on tests of scientific ability, and why males are more likely to pursue a career in science. One explanation is that science is stereotyped as a male activity, which causes girls to have lower expectations for success (Eccles, 1987). For example, Shinar (1975) has found that physicist is a highly male stereotyped occupation, and we know that even elementary school children can distinguish occupational stereotypes and base their occupational preferences upon these stereotypes (Tremaine & Schau, 1979). Achievement in science is bound to be less valued if one does not view science as a possible occupation.

The educational setting has also been used to explain why females do not perform as well as males in science. Textbooks present information in stereotyped fashion-portraying boys as active, resourceful, creative, and problem solvers, and girls as passive, helpless, and dull (Scott, 1981). The proportion of teachers in the scientific fields is also weighted in favor of males, providing females with fewer science role models. The classroom environment may also discourage females from participating and asking questions, especially in a science class that is primarily male (Kung, 1997). Staberg (1994) found that boys would boo when girls gave the wrong answer, and sigh when girls asked questions. Some girls reported that they hardly ever raised their hand because they "were afraid of saying something wrong" (p. 39).

Another way to explain gender differences in the sciences is to take a cognitive psychology approach. This approach starts by looking at the known cognitive differences between males and females. In the past, females outperformed males on verbal tasks, but these differences have been eliminated in recent years (Hyde & Linn, 1988). The same is not true for spatial abilities, though. Research has consistently found that males outperform females on spatial tasks, and this gender difference persists (Robinson, Abbot, Berninger, & Busse, 1996).

Next, one needs to examine the underlying processes of verbal and spatial ability. Both of these abilities are related to working memory capacity, which is the amount of material that can be attended to at any given moment. In contrast to earlier views, it has been recently suggested that working memory can be separated into at least two pools of resources: Verbal working memory and spatial working memory (Shah & Miyake, 1996). It also appears that different cognitive processes may be fueled by different pools of working memory resources (Daneman & Tardiff, 1987). Verbal working memory can be seen as either a phonological loop and a central executive, or as a processing component and a storage component (Baddeley, 1999). The verbal memory span (Daneman & Carpenter, 1980) was developed to assess the storage component of verbal spatial memory.

Spatial working memory can also be broken down further into several components. Shah and Miyake (1996) state that spatial working memory consists of a processing component and a storage component. They developed a spatial memory span test, similar to the verbal span test, to measure the storage component of spatial working memory. Dror and Kosslyn (1994) identify the components of spatial memory as: image generation, image maintenance, image inspection, and image transformation. Using several different spatial measures, Dror and Kosslyn went on to demonstrate age differences in the efficiency of each component of working memory. If age differences exist, it is also entirely possible that gender differences could also exist for the components of spatial working memory. Each of these models, then, recognizes that working memory is not a single entity, and that each working memory system is itself made up of different parts. The increased spatial ability found for males could be due to a greater spatial working memory capacity relative to females.

How can this gender difference in spatial working memory capacity explain differences in scientific achievement? It has been repeatedly shown that verbal working memory capacity predicts reading performance (Just & Carpenter, 1992). Among other things, working memory is used for establishing coherence between parts of the text, for generating inferences related to the text, and for anaphoric reference. Science textbooks contain more than just words, however. The text is often accompanied by diagrams, which help to explain and demonstrate the concepts discussed in the text. These pictures also guide the generation of inferences and assist in the construction of mental models for the text, which support the discovery of relationships left implicit in the text (Glenberg & Langston, 1992). These inferences would take place in spatial working memory. When comparing text alone to text plus diagrams, Glenberg and Langston found that the mental models for the text alone condition would represent the information in the order that it is presented in the text, which may or may not be correct. A diagram was needed to understand the correct spatial orientation of the material.

Many science textbooks involve diagrams where the movement of parts must be inferred to fully comprehend the concept that is discussed. For example, the diagram may illustrate the movement of a block down an incline plane or the compression of a spling. Hegarty and Just (1993) found that the text alone led to the formation of a mental model concerning a pulley system, but the reader needed a diagram to fully understand the movement of the pulleys. The diagram served as an external memory aid, freeing up some working memory resources for the comprehension of additional text material. These additional resources may have also helped the reader to make additional inferences about the system, which would lead to greater comprehension. After observing the eye-fixations of the readers, Hegarty and Just concluded"readers construct a mental model of a pulley system incrementally by integrating the information in a text and diagram" (p. 735). These findings demonstrate how verbal and spatial working memory work together to form an overall representation of a passage of text. They also demonstrate that mental models formed from text-alone presentations are often inadequate or incorrect.

Given the importance of diagrams to constructing mental models of text, the importance of both verbal working memory and spatial working memory becomes evident. Research has demonstrated that males have better spatial abilities than females, but gender differences in the spatial memory span have not yet been tested. If such a difference exists, it could explain the gender difference found in spatial abilities. In addition, the hypothesized advantage that males may have over females in terms of spatial working memory capacity may lead to better comprehension and integration of the diagrams that are presented with scientific text. The textbase (Kintsch & van Dik, 1978; Schmalhofer & Glavanov, 1986) would be the same for males and females, since this knowledge representation would utilize verbal working memory. The difference in comprehension would appear at the situation model level. During reading, the larger spatial working memory for males would allow them to process more of the diagram at one time, and they would be able to make a greater number of inferences relative to females. These inferences would then be used to build a more complete mental model of the material, which would also lead to better comprehension.

There is support for the hypothesized relationship between spatial ability and comprehension of scientific textbooks. Staberg (1994) observed females in a physics class, where one female student commented "I understand all the components in the electric motor ... but then I don't understand why it starts rotating ... one understands up to a certain point"(p. 40). Kliese and Over (1993) had males read texts discussing how various mechanisms operated. These texts contained diagrams that varied in the amount of information provided. Some of the diagrams only labeled the parts of the apparatus, while other diagrams labeled the parts and explained the steps in the operation. Kliese and Over found that the amount of additional information provided in the diagrams had no effect on the males' comprehension, suggesting that males were able to comprehend the parts-only diagrams well enough to make the necessary movement inferences for full understanding. When working with female readers Mayer (1989) found the opposite result. Comprehension was facilitated when the diagrams were more elaborate, thus reducing the number of inferences that the reader had to make. This finding suggests that the spatial ability, and possibly working memory size, of the females affected the amount of processing that occurred with the diagram. These two studies show that males were better able to comprehend the diagrams that accompany passages of scientific text. These differences in comprehension presumably occurred because of differences in spatial working memory capacity.

One last issue to address involves the hypothesized gender difference in spatial working memory size. If it is true that males show a larger spatial working memory, on average, than females, is there a biological difference that can account for this finding? The Levy (1976) hypothesis concerning brain lateralization may offer an explanation. The Levy hypothesis states that spatial performance is optimized when it is strongly lateralized in one hemisphere. With strong lateralization verbal processing occurs only in the left hemisphere, and spatial processing is limited to the right hemisphere. When lateralization is weaker, the two hemispheres will compete when performing a task that requires both verbal and spatial processing. This competition leads to impaired performance, especially for spatial processing. Levy concludes that females tend to be less lateralized, which leads to an increase in verbal performance relative to spatial performance because more cortical space is devoted to verbal functions. This increase in verbal ability comes at the expense of spatial ability, as the verbal processing "crowds out" the spatial processing. Other research suggests that lateralization is not complete until adolescence, which is when gender differences in spatial ability usually appear.

Another explanation for spatial working memory differences is that males engage in activities that are more spatial in nature. Greenfield and Lauber (as cited in Halpern, 1986) believe that many leisure-time activities that attract a higher percentage of males, such as video games and billiards, involve spatial processing and repeated practice in these activities improve spatial abilities. If practice can affect spatial processing, then spatial skills training should also improve spatial ability, and this improvement as been found for embedded figures tests (Connor, Schackman, & Serbin, 1978) and mental rotation tasks (Vandenberg, 1975). It is still unclear as to the way in which these biological and experiential factors interact to produce the gender differences in spatial ability.

The present study tested the hypothesis that the gender differences found tbr scientific achievement are due, at least in part, to the way the material is presented in textbooks. It was hypothesized that too much of the time information is only presented in diagrams, and not explicitly presented in the text. Those who have large spatial working memories can process the diagram effectively and extract this information, whereas those with smaller spatial working memories miss this information, and comprehension suffers. The present study first measured the verbal memory span and the spatial memory span of all participants to see if gender differences in memory span could be found. The second phase of the study involved the reading and comprehension of scientific text. Three short expository passages were presented with accompanying diagrams. In some of the passages important concepts were only presented in the diagrams (the Short passages), whereas other passages presented this material in both the diagrams and the text (the Long passages). Participants were then tested on recall and question answering measures to see if comprehension was affected by the way the material was presented.

If the spatial working memory span is the underlying mechanism for the gender difference in spatial ability, males would have a larger spatial memory span than females, with no difference between genders for verbal memory span. Furthermore, if the Levy (1976) hypothesis is correct, then the spatial memory span and the verbal memory span would be correlated in females, but not in males. If spatial ability does predict how well information from both the text and a diagram are processed and integrated, then males should show no comprehension difference between the short passages (information presented in either text or diagram, but not both) and the long passages (all information in diagram is also presented in text), whereas females would show better recall and question-answering for the long passages.

Method

Participants

Sixty-three students (15 males, 48 females) from two midwestern universities participated in the present study for course credit. Due to an error in administering some of the recall tests protocols from only 37 participants (9 males, 28 females) were used for the recall data.

Materials

Verbal working memory span was assessed using the memory span test developed by Daneman and Carpenter (1980). In this test, participants read a set of sentences out loud, one sentence at a time, and remember the last word of each sentence. After reading the final sentence in the set, they are asked to recall all of the last words. The sets start out with two sentences, and increase in size by one sentence after every five trials, up to a maximum set size of six sentences. Verbal working memory span is the set size where three of the five the participant can recall at least 3 of the 5 sets of words.

Spatial working memory span was assessed using a modified version of the spatial memory span test developed by Shah and Miyake (1996). In the present experiment the stimuli were presented on a series of index cards, instead of on a computer as in the Shah and Miyake study. Participants saw capital letters that were rotated from an upright position. The letters were presented in either a normal manner or as a mirror image. The participant had to state whether the letter was a normal image or mirror image, and had to remember how far the letter was rotated. After seeing the last letter in each set, the participant had to recall the amount of rotation of each letter in the set. The spatial memory span was defined as the set size where the participant could correctly recall the rotation of all letters in at least three of the five sets.

The expository passages used in the present study were taken from Encyclopedia Americana, and described the operation of an electron microscope, an internal-combustion engine, and a refrigeration unit. Each passage was presented with a diagram that showed the spatial layout of the object as well as different steps in the operation. Some of these passages presented important concepts concerning layout and operation only in the diagram (the Short passages), and others presented this information in the text as well as in the diagram (the Long passages).

Procedure

In the first session each participant was tested individually on the verbal working memory span test, and then on the spatial working span test. In the second session the participants were given the three passages and instructed to study the material as they would for an exam. Each participant received either 1 Short and 2 Long passages, or 2 Short and 1 Long passage, and was given unlimited time to read all three of the texts. After the participants were finished reading, they were asked to recall as much of each passage as they could. After they were finished with the recall test they were presented with 6 true/false questions for each passage. The true/false questions asked about spatial relationships between parts of the object, the operation of the object, and about the order that parts moved as the object operated.

Results

The first analysis involved the verbal and spatial working memory spans found during the first session. Verbal working memory span was calculated as the set size where the participant could recall the last words in three out of the five sets. If the participant was only able to correctly recall two of the five sets a value of 0.5 was added to the previous set size. The spatial working memory span was calculated in the same way as the verbal working memory span, only using the letter position recall to determine span size. The means and standard deviations for the verbal and spatial memory spans are presented in Table 1.

Males were found to have both a larger verbal working memory span, t(61) = 2.26, p [less than or equal to] .05, and a larger spatial memory span, t(61) = 2.62, p [less than or equal to] .05. For males, the two spans were independent, r = .08. The two were slightly correlated for females, r = .22.

Three recall scores were then calculated for each reader: a score from the long passages (Long), a score from the text part of the short passages (Short Text), and a score from the diagram part of the short passages (Short Diagram). The recall means are presented in Table 2.

The recall scores were analyzed using a 2(Gender) x 3(Source: Long, Short Text, or Short Diagram) mixed-design ANOVA with repeated measures on the last factor. A marginally significant difference was found between males and females, F(1, 35) = 3.13, p = .08. The source of the information did affect recall. F(2, 70) = 39.82, p [less than or equal to] .05, with the Short Diagram information (M = .08) being recalled less than either the Short Text information (M = .24) and the Long information. The Gender x Source interaction was marginally significant, F(2, 70) = 2.51, p = .09. This interaction occurred because the males recalled the short text slightly better than the long text, whereas there was no difference between text types for females. One must remember that these marginally significant effects occurred with a tow number of participants in each group.

The true/false scores were submitted to a 2(Gender) x 2(Text: long or short) x 2(Source: text or diagram) mixed-design ANOVA with repeated measures on the last two factors. The means for the true/false scores are presented in Table 2. There was a main effect for Gender, F(1, 61) = 8.17, p [less than or equal to] .05, as males made more correct responses than females (.79 vs .67). There was also a main effect for Source, F(1, 61) = 16.61, p [less than or equal to] .05, with information coming from the text (M = .78) leading to better performance than information coming from the diagram (M = .63). The Text x Source interaction was significant, F(1, 61) = 10.10, p [less than or equal to] .05, and showed that there was no difference between information that came from the text or the picture in the long passages (.73 vs. .65), but information from the text of the short passages (M = .83) led to better performance than information found in the diagram of the short passages (M = .59). Gender differences were also found as males also showed better comprehension for short diagram information, t(61) = 2.20, p [less than or equal to] .05, and for long diagram information, t(61) = 3.27, p [less than or equal to] .05.

Discussion

The present study did find the hypothesized gender differences in spatial working memory span. As predicted, males demonstrated a larger spatial span than females. This finding explains past research showing that males outperform females on tests of spatial ability. Shah and Miyake (1996) believe that spatial memory span is a good predictor of spatial ability as their spatial working memory span measure correlated highly with other measures of spatial performance. In the present study an unexpected finding occurred for verbal working memory, with males showing a larger span size than females. It is unclear as to why this finding occurred.

Spatial and verbal memory spans were not correlated in males, which may support the idea that these two abilities are strongly lateralized (Levy, 1976). In females the two spans were slightly correlated in a positive way, and not negative as predicted. At first glance it seems that this finding contradicts the Levy hypothesis. However, it may be females were using a verbal type of processing on the spatial span test.

In terms of text comprehension, the finding were mixed concerning gender and format of the text. The general findings were that presenting the information in both the text and a diagram had no effect for recall, but did aid in answering true/false questions. Gender differences were found for question answering concerning information from the diagram but not for information from the text.

There were no gender differences found for recall of the material, but the source of the information did matter. Participants recalled information when it was presented in the text, but not when it was presented in the diagram. Judging from the recall means, it appears that participants were unable to incorporate information from the diagram into their final memory representation. Another possible explanation for the low diagram recall could involve the instructions given the participants. The instructions told the participants to write down as much as they could remember from the passage, as well as any inferences that they made. It is possible that the participants misunderstood these instructions, and only wrote down ideas that were mentioned in the text.

Males did outperform females on the true/false questions when the question involved information contained in the diagram. This finding supported the idea that the larger spatial working memory found for males would lead to better comprehension of the diagram. In addition, this difference occurred with both the short passages and the long passages. The difference between males and females for comprehension of the short passage diagrams was expected, as the smaller spatial working memory for females affected how well females integrated the information from both sources. It appears that females could not incorporate any information from the diagram into memory, as their responding on the true/false questions was just slightly above chance.

The difference between males and females with the long passage diagrams was unexpected. It was hypothesized that if the diagram information was also explicitly stated in the text then the females would also learn this material. Their performance on these questions did improve relative to short diagram questions, but it was still below the males' performance. The males' comprehension may have been facilitated because this information was presented to them in both a verbal and spatial form, so it was encoded in multiple ways (Paivio, 1991). The females only encoded this information in a verbal way, and the length of the text pushed their verbal working memory capacity to the limit.

In general, the true/false data support the finding of the recall task that the females could not comprehend the information in the diagrams. Performance was greater for the text part of the short version than for the long version, probably because there was less overall information to represent in memory. The difference between males and females for both of the text measures was not significant, which supports the hypothesis that there is no difference in verbal ability between the genders. However, males were getting additional information from the diagram that the females were missing.

The overall conclusions of this study support the original hypotheses that gender differences in working memory size lead to gender differences in comprehension of texts. When important information is contained in the diagram but not in the text, this informing is missed by females, but is comprehended to a certain extent by males. When the text reiterates what is found in the diagram females comprehension of this material improves to a certain degree. Males may have outperformed females in this condition because they have more experience incorporating information from the two different formats. The diagram is present for the females to use as a memory aid, but past experiences and frustrations may have taught them to ignore the diagram and only concentrate on the text.

The suggestions which the present study offers to reducing gender differences in science is to look at the materials that are used, and the styles of learning that are stressed. Science textbooks need to make sure that important concepts are explicitly mentioned in the text. and not just left for the reader to infer from the diagram. In addition, we need to stress the importance of using diagrams as external memory aids, and teach ways of incorporating information from both sources into a final memory representation, With some modifications in our textbooks and instructional methods, we can assure that the next generation will all enter 21st century science together.

Table 1 Means and Standard Deviations
for Working Memory Span

          Verbal        Spatial
Gender    Span          Span

Males     2.86 (0.60)   2.50 (1.13)
Females   2.51 (0.69)   1.80 (0.87)

Table 2
Means and standard deviations for the recall
and question answering data

Measure                    Males       Females

Long text recall           .25 (.13)   .18 (.12)
Short text recall          .30 (.18)   .19 (.10)
Short diagram recall       .07 (.12)   .06 (.10)

Long text true/false       .77 (.31)   .71 (.31)
Long diagram true/false    .82 (.21)   .59 (.24)
Short text true/false      .88 (.19)   .81 (.20)
Short diagram true/false   .68 (.19)   .56 (.19)

References

Baddeley, A. D. (1999) Essentials of Human Memory. East Sussex, England: Psychology Press, Ltd.

Daneman, M., & Carpenter, P. A. (1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19, 450-466.

Dror, I. E., & Kosslyn, S. M. (1994). Mental imagery and aging. Psychology and Aging, 9, 90-102.

Eccles, J. (1987). Gender roles and achievement patterns: An expectancy value perspective. In J. M. Reinisch, L.A. Rosenblum, & S. A. Sanders (Eds.), Masculinity/Femininity. New York: Oxford University Press.

Glenberg, A. M., & Langston, W. E. (1992). Comprehension of illustrated text: Pictures help to build mental models. Journal of Memory and Language, 31, 129-151.

Hegarty, M., & Just, M. A. (1993). Constructing mental models of machines from text and diagrams. Journal of Memory and Language, 32, 717-742.

Hyde, J. S., & Linn, M. C. (1988). Gender differences in verbal ability: A meta-analysis. Psychological Bulletin, 104, 53-69.

Just, M. A., & Carpenter, P. A. (1992). A capacity theory of comprehension: Individual differences in working memory. Psychological Review. 99. 122-149.

Kintsch, W., & van Dijk, T.A. (1978). Toward a model of text comprehension and production. Psychological Review, 85, 363-394.

Kung, J. (1997). A student's view. The Physics Teacher, 35, 118-119.

Levy, J. (1976). Cerebral lateralization and spatial ability. Behavior Genetics, 6, 171-188.

Robinson. N. M., Abbott, R. D., Berninger, V. W., & Busse, J. (1996). The structure of abilities in math-precocious young children: Gender similarities and differences. Journal of Educational Psychology, 88, 341-352.

Schmalhofer, F., & Glavanov, D. (1986). Three components of understanding a programmer's manual: Verbatim, propositional, and situational representations. Journal of Memory and Language, 25, 279-294.

Scott, K. P. (1981,April). Whatever happened to Jane and Dick? Sexism in texts reexamined. Peabody Journal of Education. 135-140.

Shah, P., & Miyake, A. (1996). The separability of working memory resources for spatial thinking and language processing: An individual differences approach. Journal of Experimental Psychology: General, 125, 4-27.

U.S. Bureau of Labor. (1995). Employment and Earnings.

U.S. National Center for Education Statistics. (1993). Digest of Educational Statistics.

John F. Geiger, Department of Psychlogy and Human Ecology, Cameron University. Robert M. Litwiller, Palmer College of Chiropractic

Correspondence concerning this article should be addressed to Direct all correspondence to John F. Geiger, Department of Psychology and Human Ecology, Cameron University. 2800 W Gore Blvd., Lawton, OK 73505. Email: [e-mail].