CRL Newsletter

April 1994

Vol. 8, No. 2

Abstract Better Than Concrete: Implications for the Psychological and Neural Representation of Concrete Concepts

Sarah D. Breedin1, Eleanor M. Saffran1,2, and H. Branch Coslett1

Temple University

1 Center for Cognitive Neuroscience, Department of Neurology
2 Department of Speech-Language-Hearing


Note for this issue: Two figures are referred to at the end of this paper which could not be included in the text only e-mail version. If you would like a copy of the figures we would be happy to send it to you via surface mail.

Abstract

People are better at identifying and remembering concrete than abstract words (the concreteness effect). We present data from a patient who shows the opposite pattern: superior performance on abstract words across a range of tasks. These data challenge theories that differentiate between these word classes in essentially quantitative terms. DM's pattern of semantic loss together with data from imaging (SPECT/MRI co- registration) studies, indicates that concrete concepts depend on the integrity of perceptual attributes that are represented in neural systems in the inferior temporal lobe(s).

Paper

Concrete words, typified by object terms such as book and elephant, are easier to learn, comprehend, and remember than abstract words such as destiny and hope. This advantage, known as the concreteness effect1,2 is generally taken to reflect an additional code for processing or component of semantic representation that makes concrete words easier to access and remember1,2. Thus, concrete word meanings are thought to depend on two codes (symbolic and sensorimotor) whereas abstract word meanings rely only on one (symbolic)1.

Concepts that are multiply represented should be more resistant to loss, and, indeed, there is ample evidence that brain damage can serve to magnify the concreteness effect3. The opposite pattern, an advantage for abstract words, would be difficult to explain on current theories. While observations consistent with such a pattern have occasionally been reported4,5, the abstract advantage has not been systematically studied, nor have its cognitive and neural bases been explored.

In this paper we present a comprehensive investigation of a patient who demonstrates reversal of the concreteness effect, including neuro-imaging data that link this pattern to dysfunction of a specific region of the brain. These data provide the basis for an account of the concreteness effect that can also accommodate its occasional reversal by brain damage.

The patient, DM, has a language disturbance that conforms to the profile recently labeled "semantic dementia" 6, a progressive disorder characterized by fluent speech with severe anomia, poor single word comprehension, and a reading deficit (surface dyslexia) defined by an inability to read irregular words (e.g., yacht)7. In contrast, syntactic abilities are reported to be preserved, as are perceptual and visuo-spatial abilities6. Insofar as it can be tested, episodic (day-to-day) memory also appears relatively spared6. The word retrieval and comprehension deficits are attributed to erosion of the knowledge base (semantic memory) that supports language. Evidence of temporal lobe atrophy, particularly on the left, has been found in all of the cases reported thus far6. DM shares of the reported characteristics of these patients, differing from them only in that his knowledge of abstract concepts appears relatively preserved.

DM is a 56 year-old, right-handed professional with a master's degree who first exhibited cognitive impairments four and a half years ago when he began experiencing difficulty remembering names and appointments. Neurologic examination except for mental status was normal. IQ as measured by the WAIS-R was 86, which, given his educational level, clearly represents a decline in intellectual function. DM's most striking deficits on neuropsychological testing occurred on picture naming8 (11 of 60 correct) and word comprehension tests9 (below first percentile for normal adults).

Non-verbal tests of object knowledge and recognition abilities also revealed significant deficits. On the Pyramids and Palm Trees test10, which uses pictorial stimuli to assess knowledge of object properties, DM obtained a score of 60% correct where the mean for normal controls was 98.5%. Object recognition was assessed by asking him to decide whether line drawings depicted objects or nonobjects11, the latter created by combining parts of real objects (e.g., a cat's body with an elephant's head). DM was 78% correct whereas the mean for control subjects was 94% correct. He was also impaired on a task that required him to select one of four colors to fill in a line drawing, choosing incorrectly for 18 of the 28 objects (e.g., purple for rhinoceros, grey for carrot)12.

Further assessment of DM's knowledge of word meanings included a task in which he was asked to provide definitions for an auditorily presented list of concrete and abstract words (n=386) that were matched for frequency. Four naive raters rated each definition on a scale from 1 ("bad") to 7 ("good"); those given a mean rating of less than 5 were treated as incorrect. DM showed a reverse concreteness effect, that is, he was better at defining abstract words (59%) than concrete words (46%; X2(187, 194)=6.42, p<.025). Qualitatively, DM's definitions of concrete words were notable for the lack of perceptual descriptions. Thus, in the three examples that follow, DM does not describe the appearance of a ram, the fluidity or color of ink, or the shape, color or flavor of a plum:

ram - a little animal
ink - something that covers
plum - something you insert

In contrast to his concrete word definitions, DM's definitions for abstract words tended to be articulate and to the point:

try - try is to endeavor to accomplish something
opinion - your concept or perspective
measurement - something to evaluate the size of something, whether it's big or small in size

The same pattern emerged on other tasks that compared knowledge of concrete and abstract concepts (see summary in Table 1). These tasks included synonymy judgment tasks (concrete-abstract; noun-verb13) in which DM was asked to indicate which of three words (e.g., loyalty, obsession, allegiance) was least related to the others in meaning. The abstract advantage also emerged on a lexical decision task, where DM showed a marginally significant (t(1)=1.89, p<.06) 51 msec. advantage for abstract over concrete words, a result opposite to that of control subjects.

Table 1. Summary of DM's percent correct on tests of abstract and concrete comprehension.

_______________________________________________________
DM Controls (n=5)
Mean Range
_______________________________________________________
Shallice/McGill Word/Picture Matching
Abstract (n=30) 53 89 (87-93)
Concrete (n=30) 37 97 (93-100)
Difference 16 -8

Concrete/Abstract Synonymy Judgments
Abstract (n=26) 85 ceiling
Concrete (n=26) 58 ceiling
Difference 27*

Noun/Verb Synonymy Judgments
Verbs (n=16) 94 ceiling
Nouns (n=16) 63 ceiling
Difference 31*
_______________________________________________________
*p<.05

We have shown that DM demonstrates an advantage for abstract words across a range of tasks. As noted earlier, this pattern presents a challenge for theories of semantic representation predicated on the more typical finding of superiority for concrete words. If loss of brain tissue can reverse this effect, representations for abstract and concrete concepts are likely to differ in more than the additional features 14 or capacity to be imaged1 that have been posited for the latter. Concrete terms refer to objects in the world which are known primarily through the senses, while the meanings of abstract terms are acquired within a linguistic context and are shaped by that context (compare, for example, a mental process and a political process ) to an extent that is not true of concrete words (the properties of black and white horses are in all other respects identical). Such differences are likely to have implications for the format and content of semantic representations. Assuming that there are such differences, the concreteness effect could be reversed by the loss of information essential to the definitions of concrete words but not abstract ones; likely candidates are perceptual attributes that contribute importantly to the definition of object terms and critically to distinctions among them (consider lion and tiger, for example). We speculate that damage to DM's semantic network disproportionately involved such attributes.

DM's performance on a task that required him to distinguish among exemplars within semantic categories was consistent with this hypothesis. The task employed the synonymy judgment paradigm to assess his ability to distinguish among exemplars within categories that might be expected to differ in the relative importance of perceptual attributes15. For example, while tools and items of furniture are defined by their functions as well as their shapes, exemplars from the categories animals and insects are largely distinguished by perceptual features such as color and form16. DM was at chance on animals, insects, and musical instruments (see Table 2) and showed better preserved knowledge of body parts, furniture, occupations, and tools17.

Table 2. Percent correct for DM and control subjects on the semantic category synonymy task.

__________________________________________________________
Controls
__________________
Category DM Mean Range
___________________________________________________________
Animals (n=20) 30 98 (90-100)
Body Parts (n=18) 78 98 (94-100)
Furniture (n=14) 86 98 (93-100)
Insects (n=10) 30 98 (92-100) Musical Instr. (n=12) 33 95 (83-100)
Occupations (n=24) 71 98 (92-100)
Tools (n=21) 62 95 (81-100)

In a second experiment, we probed more directly for the differential loss of perceptual features by comparing DM's knowledge of perceptual and non-perceptual attributes of living and non-living things. We selected 39 living ( e.g., pear, ostrich) and non-living (e.g., anchor, vase) items that were matched for word frequency, as well as a set of perceptual (e.g., hard, soft) and nonperceptual attributes (e.g., safe, dangerous) that we had determined that DM understood. For each of the object terms we developed yes/no questions using two of the attributes (e.g., Does a lion have four legs? Is a lion safe?). DM was significantly worse at answering the perceptual questions (69%) than the non-perceptual questions (86%; X2(85, 71)=5.94, p<.02). Although he proved to be slightly better at answering questions about living (81%) than non-living things (72%), this difference was not significant (X2(78, 78)=1.73, .05<p<.3) 18.

Thus, in addition to documenting an advantage for abstract words across a range of tasks, our studies of DM provide a clear indication of the source of this effect: disproportionate loss of the perceptual descriptors that weigh heavily in the representation of concrete concepts. The present results also point to a gap in the psychological literature on semantic representation, which has little to say about the manner in which abstract concepts are represented. We will not attempt to address this issue here, other than to reiterate the point that, relative to concrete words, abstract words are more dependent on linguistic context both for acquisition and interpretation. Just as the representations for object terms incorporate information derived from the senses, it is likely that representations of abstract words are linked to grammatical operations that are essential to using them19. We note, in this regard, that DM's syntactic abilities were remarkably preserved20.

In addition to behavioral investigations of DM's deficit, neuro-imaging studies were carried out to define the anatomical basis of his impairment. MRI scan was normal. SPECT scan using HM-PAO (administered at rest with eyes closed) demonstrated decreased uptake in the inferior temporal and fusiform gyri bilaterally, especially on the left. Additional analyses in which MRI and SPECT datasets were co-registered were also performed.21 Mean intensity scores for regions of interest (ROIs, defined on the basis of precise anatomic landmarks visualized on the MRI) and the whole brain were calculated. As indicated in Figure 1, the ROI to Whole Brain ratio was lowest for the anterior portion of the inferior temporal and fusiform gyri (operationally defined as the anterior 5 cm. of the temporal lobe); scores for the posterior inferior temporal and fusiform gyri and the middle and superior temporal gyri did not differ from other frontal and parietal cortical ROIs. The score for the left anterior inferior temporal and fusiform gyri was approximately 10% lower than that of the right, an asymmetry which exceeded that of any other ROI.

Figure 2 depicts the ROI/Whole Brain Mean Intensity Scores for the right and left inferior temporal and fusiform gyri as a function of distance from the temporal tip. The perfusion deficit was maximal 2-3 cms. from the temporal tip where the Mean Intensity Scores on the left were approximately 50-60% of the value of the posterior portions of the inferior temporal gyri. Similar localized perfusion deficits were not observed in analyses of the middle and superior temporal gyri or in normal controls.

DM's neuroimaging studies demonstrate relatively circumscribed regions of abnormality involving anterior portions of the inferior temporal and fusiform gyri bilaterally but asymmetrically, with greater involvement on the left. Although limited, data for other patients with evidence of a reverse concreteness effect also demonstrate damage to one or both temporal lobes5. The localization data are of particular interest in light of evidence from humans and monkeys that implicates the inferior temporal lobe in object recognition22 and language23. PET studies in humans, for example, reveal that the inferior temporal lobe (area 20), is activated in object recognition tasks, particularly on the left24. On the basis of evidence that visual object recognition mechanisms are localized more posteriorly in humans than in monkey, presumably as a result of displacement by the development of phylogenetically newer cortical areas24, we speculate that anterior regions of inferior temporal cortex are involved in networks at the interface between perception and language. This contention is consistent with evidence that DM's impairment with concrete words is attributable, at least in part, to a loss of perceptual information.

On the account developed here, one might expect to observe reverse concreteness effects in other patients with lesions involving the anterior inferior temporal lobe (e.g., in cases of temporal lobectomy or herpes simplex encephalitis), yet reports of this pattern are rare. This may reflect the fact that the tests required to demonstrate the effect are seldom administered. Another possibility is that this phenomenon, like a number of other striking clinical syndromes, such as prosopagnosia, may typically be associated with bilateral temporal lobe lesions25. Finally, it may be that the reverse concreteness effect emerges only when extensive loss of perceptual components is coupled with some degree of impairment to other components of semantic representation. In this context, one should note that, although disproportionately impaired with concrete words, DM exhibited a deficit with abstract words as well.

References and Notes

1. A. Paivio, Can. J. of Psych., 45, 255 (1991); D. Kieras, Psych. Bul., 85, 532 (1978).

2. P.J. Schwanenflugel, in The Psychology of Word Meaning, P.J. Schwanenflugel, Ed. (Lawrence Erlbaum, Hillsdale, NJ, 1991), chap. 9.

3. M. Coltheart, K.E. Patterson, J.C. Marshall, Eds., Deep Dyslexia (Routledge and Kegan Paul, London, 1980); E.M. Saffran and N. Martin, in Neuropsychological Impairments of Short-Term Memory, G. Vallar and T. Shallice, Eds., (Cambridge University Press, Cambridge, 1990), chap. 6.

4. E.K. Warrington, Quar. J. of Exp. Psych., 27, 635 (1975).

5. E. K. Warrington, Brit. J. Psy. 72, 175 (1981); E. K. Warrington and T. Shallice, Brain 107, 829 (1984).

6. J. S. Snowden, P. J. Goulding, D. Neary, Behav. Neurol. 2, 167 (1989); J. R. Hodges, K. Paterson, S. Oxbury, E. Funnell, Brain 115, 1783 (1992).

7. K. E. Patterson, J. C. Marshall, M. Coltheart, Eds., Surface Dyslexia: Neuropsychological and Cognitive Studies of Phonological Reading (Lawrence Erlbaum Assoc, London, 1985).

8. E. Kaplan, H. Goodglass, S. Weintraub, The Boston Naming Test (Lea and Ferbinger, Phila., PA, 1983).

9. L. M. Dunn and L. M. Dunn, Manual for Forms L and M of the Peabody Picture Vocabulary Test - Revised (American Guidance Service, Circle Pines, MN, 1981).

10. D. Howard and K. Patterson, The Pyramids and Palm Trees Test (Thames Valley Test Co., Suffolk, VA, 1992).

11. M. J. Riddoch and G. W. Humphreys, Cog. Neuropsy. 4, 131 (1987).

12. Note that DM's performance cannot be attributed to impaired color perception, because his color perception was normal as measured by the Farnsworth D-15 panel.

13. We compared DM's knowledge of nouns and verbs because one dimension along which they are thought to differ is concreteness - nouns are thought to be more concrete than verbs; D.A. Allport and E. Funnell, Philos. Trans. Royal Soc. of London 295, 397 (1981).

14. D.C. Plaut and T. Shallice, Cog. Neuropsciy. 10, 377 (1993).

15. M. J. Farah and J. L. McClelland, J. Exp. Psy: Gen. 120, 339 (1991).

16. E. K. Warrington and R. M. McCarthy, Brain 110, 1273 (1987).

17. While frequency played some role in his performance (frequency correlated with DM's performance, r=.38, p<.0001), which was notably poor on some categories with low mean frequency values ( e.g., insects and musical instruments), it clearly was not the only determinant of his success on this task. Thus, although animals and tools were equivalent in frequency, DM performed significantly worse on animals than tools (X2(20,21)=4.2, p<.05).

18. While this finding is surprising, note that we made no effort to control for the difficulty of probes in the living and nonliving categories.

19. A. R. Schmauder, thesis, University of Massachusetts (1992).

20. S.D. Breedin and E.M. Saffran, in preparation.

21. H. H. Holcomb et al., J. Comp. Assist. Tomog. (in press); H. H. Holcomb et al., Arch. of Gen. Psy. (in press).

22. L.G. Ungerleider and M. Mishkin, in Analysis of Visual Behavior, D.G. Goodgale and R.J.W. Mansfield, Eds., (MIT Press, MA, 1982), pp. 549-586.

23. T. H. Burnstine et al., Neurology , 40, 966 (1990).

24. J. V. Haxby et al., Proc. Nat. Acad. Sci. 88, 1621 (1991).

25. A.R. Damasio, H. Damasio, and G.W. Van Hoesen, Neurology, 32, 331 (1982).

Address correspondence to Dr. Sarah Breedin, Center for Cognitive Neuroscience, Dept. of Neurology, Temple University School of Medicine, 3401 North Broad St., Philadelphia, PA 19140

This research was supported by NIH grants R01-DC00191 to the second author and R01-AG08870 to the third author.

Unsubscribe

This mailing list is no longer active. If you wish to unsubscribe from the CRL Newsletter mailing list, enter your name and e-mail address in the form below.

Your name:

Your e-mail address: