AN fMRI STUDY OF REGULAR AND IRREGULAR INFLECTION IN GERMAN

Carrie Campbell1, Alan Beretta1, Thomas Carr2, Jie Huang3,
Yue Cao3
Departments ofLinguistics1, Psychology2, and Radiology3
Michigan State University,
East Lansing, Michigan, USA

INTRODUCTION

Symbol-manipulating and associative accounts of the nature of cognitive processing have both focused on inflectional morphology. These conflicting accounts are in general agreement that irregular inflection is accomplished by associative memory. The conflict arises with regard to regular inflection. The associative account proposed by certain connectionist modelers (e.g., Rumelhart & McClelland 1986, Hare et al 1995, Plunkett & Juola 1999) posits that regular inflection is accomplished by the same associative memory process as irregular inflection. By contrast, the symbol-manipulating account (advocated by, e.g., Pinker 1999, Clahsen 1999, Marcus 2001) claims that regular inflection is achieved by a rule which applies in default circumstances (novel words, names, borrowings, and so forth). A great deal of evidence has been amassed that supports the notion of a default regular inflection but, at the same time, associative models have been devised which their proponents believe are clear advances over earlier models.

Recently, researchers have turned to neuroimaging to ask the brain how many processes it thinks it has: one, consistent with the single-route associative hypothesis, or two, which would be compatible with the dual-route symbol-manipulating hypothesis. Two PET studies have been carried out (Jaeger et al 1996, Indefrey et al 1997) and one fMRI study (Bergida et al 1998). These studies all report findings which they interpret as conducive to the dual-route account. However, their findings must be treated with caution. First, all three studies used block designs, which have been criticized for being susceptible to differential priming and strategy effects between blocks of trials -- that is, between regulars and irregulars (e.g., Seidenberg & Hoeffner 1998). Second, two of the studies (Jaeger et al, Bergida et al) used English, which may not be ideal for this kind of inquiry since it has two inherent confounds: (i) the regular inflection is also the most frequent form; (ii) the regular involves adding an affix to the stem, while the irregular involves no affix but instead a variety of stem changes.
 
The present study tries to overcome the problems of these previous imaging experiments. We employ an event-related fMRI design which permits random presentation of stimuli. We examine German which lacks the confounds associated with English: regulars are not the majority forms, and segmentable affixes are common to both regulars and irregulars. 

METHODS AND MATERIALS

Subjects: Eight right-handed native speakers of German participated in the study (4 female, 4 male, mean age 28.5 years, range 23-45, SD=6.65).
 
Materials: Word lists were developed for nouns and verbs. The lists contained regular and irregular words matched for frequency using the CELEX database (Baayen et al 1995). The noun pluralization paradigm included 12 regular and 12 irregular items, while the verb past participle paradigm included 12 regular items and 12 each of two different irregular classes of verbs. Only low frequency words were chosen. Words in the noun list all showed a plural frequency no greater than 20 per million and in the verb list none had a participle frequency greater than 254 per million.

Language protocol: Subjects performed tasks which involved (i) producing noun plurals and, (ii) producing the past participles of verbs. In the noun task, items from the word list were visually presented to the subjects using an event-related paradigm to randomly mix regular and irregular words. Stimulus words were displayed sequentially, one every 14 seconds for 1 second each, separated by fixation periods. Subjects were told to silently generate the plural of the word. The verb task was identical in design and subjects were asked to silently generate the past participle of the word. The order of presentation of the noun and verb trials was counterbalanced across subjects. The presentation of the words was synchronized with MRI data acquisition.

fMRI: T2*-weighted images of 19-21 sagittal sections were acquired on a GE 1..5 T clinical scanner using a Gradient Echo EPI sequence (FOV = 24 cm, TE/TR = 50/2000 ms, flip angle 90º, matrix size 64x64, slice thickness = 7 mm). After the language paradigms were completed 19-21 T1-weighted spin echo images were collected at the same anatomical locations, with TE/TR = 14/500 ms , flip angle 90º, FOV = 24 cm, matrix 256x192, and slice thickness/spacing = 7.0/0.0mm. A set of 3D spoiled gradient echo T1-weighted MR images of the whole brain were collected, TE/TR = 3.3/8 ms, flip angle 30º, slice-thickness/spacing = 1.2/0.0 mm, 124 slices, matrix size 256x192, and FOV = 24 cm.

Data pre-processing: Images were assessed and corrected for possible in-plane translation and rotation of the head (Cao et al 1993). The signal intensity time course was corrected for possible slow baseline drifts by fitting 0, 1st, and 2nd order polynomials to the time series of each voxel.

Statistical analysis of activation: Time courses of images were cross-correlated with a gamma reference function to yield an activation map (estimated p<0.01) (Bandettini et al, 1993, Boynton et al, 1996, and Dale et al, 1997). In order to explore effects related to regularity alone and not merely to category (noun-verb), noun and verb images were collapsed to obtain an average regular and irregular image for each subject. Subtractions of these images were then performed, creating maps showing the voxels only activated by regular but not irregular words (R-I), irregular but not regular words (I-R), and those voxels activated in both paradigms (Common). Activated voxels were counted and localized in cortical regions.

RESULTS AND DISCUSSION

An omnibus ANOVA was performed on the voxel activation data (before subtraction) considering three variables as possible sources of variation: hemisphere (left or right), region (anterior or posterior), and regularity (regular or irregular). A main effect for regularity was found, with more voxels activated for irregulars than for regulars, F (1, 62) = 24.42, p < 0.001 (see Figure 1). Regularity did not interact with any other variables. Here we find a broad distinction between regular and irregular activation where single-route connectionist arguments predict none. This pattern is consistent with findings by Jaeger et al. (1996) and Indefrey et al. (1997) and corresponds to longer RTs often observed for low-frequency irregulars (e.g. Seidenberg 1992). Jaeger et al. in fact demonstrate a correlation between long RTs and extent of neural activation. On a dual process view, this could be interpreted as showing higher processing costs for memory retrieval tasks than for symbolic manipulation which is blind to the specifics of individual tokens or their frequencies. Operating on variables (verb and noun stems), the regular rule can be seen as less demanding of neural resources than associative processes which are sensitive both to frequency and token specificity. 

In order to identify the particular regions that exhibit the greatest distinction between regulars and irregulars, eight major regions were identified in each hemisphere, based on general functional similarities and distinctions and, particular to the present study, commonalities and distinctions in patterns of activation. Frontal regions included Broca’s area, the precentral gyrus, the anterior cingulate, and the remaining portions of the prefrontal cortex (i.e., not including Broca’s area). Posterior regions included the postcentral gyrus, the insula, the posterior temporal lobe (treated as one unit), and the areas of the supramarginal gyrus, the angular gyrus, and the superior parietal lobule combined as one unit (SMG/AG/SPL). Regions showing a significant effect for regularity were the right precentral gyrus (p < 0.05), the left prefrontal cortex (p < 0.01), the right SMG/AG/SPL (p < 0.01), and the left insula (p < 0.05). In all cases irregular activation was more significantly widespread than regular.

The ANOVA also revealed a trend toward a greater activation in the left hemisphere than in the right, F (1, 62) = 2.63, p = 0.11 (see Figure 1). While this trend was not significant, lateralization did turn out to be significant when the subtraction maps showing the areas activated selectively in irregular and regular processing were compared. While both the common map and the R-I subtraction showed significantly more widespread activation in the left hemisphere (p < 0.01 and p < 0.001, respectively), the I-R subtraction showed no hemispheric difference in activation. If we assume, following a dual-route account, that the regular forms utilize a linguistic, rule-based process and that irregulars do not, then it is reasonable that they would show a characteristic linguistic left hemisphere dominance. Purely memory based processes, like the uniquely irregular data in the I-R subtraction, are associated with no such hemisphere preference. This differentiation in lateralization is inconsistent with a single-route account but follows naturally from a dual-route account.

To further characterize activation specific to regular and irregular processing, Tables 1 and 2 show anatomical areas activated in the I-R and R-I subtraction conditions including the relative stabilities of activation across subjects. It can be seen that in each anatomical area examined irregulars produce greater activation than regulars (apart from the left precentral gyrus), compatible with the finding that there is more extensive irregular activation in the (larger) major regions discussed above. Table 1 & Table 2

CONCLUSIONS

The findings reported in the present study are clearly preliminary. Further analysis will determine if differences exist between regular and irregular nouns and verbs and, at a finer level of analysis, between different classes of irregular verbs. These analyses may yield a more precise understanding of the nature of inflection. Nonetheless, even these preliminary findings tell us in gross terms that the brain behaves differently for regulars than for irregulars. To summarize, we find: (1) irregulars show greater activation than regulars both overall and, where significant differences were found, within each of the eight major regions we investigated and (2) regulars show greater lateralization to the left hemisphere than do irregulars. We interpret these broad findings as most consistent with a dual-route, symbol-manipulating account of inflection.

REFERENCES

1. Baayen, H. et al. The CELEX lexical database. Ling. Data Consortium, U. Penn. (1995).
2. Bandettini, P. et al. Magn. Reson. Med. 30, 161. (1993).
3. Bergida, R. et al. Paper presented at CNS, San Francisco. (1998).
4. Boynton, G. et al. J. Neurosci. 16, 4207. (1996).
5. Y. Cao, et al., J. Magn. Reson. Med., 3, 869 (1993).
6. Clahsen, H. Behavioral and Brain Sciences. (1999).
7. Dale, A. et al. Human Brain Mapping 5, 329. (1997).
8. Hare, M. et al. Language and Cognitive Processes. (1995).
9. Indefrey, P. et al. NeuroImage, 5, S548. (1997).
10. Jaeger, J. et al. Language, 72, 451. (1996).
11. Marcus, G. The algebraic mind. MIT Press. (2001).
12. Pinker, S. Words and rules. Basic Books. (1999).
13. Plunkett, K., Juola, P. Cognitive Science, 23, 463. (1999).
14. Roland, P. et al. Human Brain Mapping, 5, 222. (1997).
15. Rumelhart, D., McClelland, J. Parallel distributed processing. MIT Press. (1986).
16. Seidenberg, M. In S. Davis, ed. Connectionism, theory and practice. OUP. (1992).
17. Seidenberg, M., Hoeffner, J. Language, 74, 104. (1998).