Representations & Computations in the Spelling System
CM: Evidence for Feedback in the Spelling System
This study developed an argument for grapheme-to-lexeme feedback in the cognitive spelling system, based on the impaired spelling performance of dysgraphic patient CM. The argument relates two features of CM’s spelling. First, letters from prior spelling responses intrude into subsequent responses at rates far greater than expected by chance. This letter persistence effect arises at a level of abstract grapheme representations, and apparently results from abnormal persistence of activation. Second, CM makes many formal lexical errors (e.g., carpet → compute). Analyses revealed that a large proportion of these errors are “true” lexical errors originating in lexical selection, rather than “chance” lexical errors that happen by chance to take the form of words. Additional analyses demonstrated that CM’s true lexical errors exhibit the letter persistence effect. This pattern of results can be understood only within a functional architecture in which activation from the grapheme level feeds back to the lexeme level, thereby influencing lexical selection.
McCloskey, M., Macaruso, P., & Rapp, B. (2006). Grapheme-to-lexeme feedback in the spelling system: Evidence from a dysgraphic patient. Cognitive Neuropsychology, 23, 278-307.
McCloskey, M., Badecker, W., Goodman-Schulman, R. A., & Aliminosa, D. (1994). The structure of graphemic representations in spelling: Evidence from a case of acquired dysgraphia. Cognitive Neuropsychology, 11, 341-392.
Neural Network Simulations of the Spelling System
Manny Vindiola is working on a project examining a Competitive Queuing simulation (Houghton, Glasspool, and Shallice,1994) of the spelling serial-order mechanism. The theory of competitive queuing been proposed as a general hypothesis about the mechanism underlying serial behavior (i.e., behavior in which a series of responses is generated one at a time in a specific order). Houghton et al. demonstrated that a competitive queuing simulation could be used to generate serial production of letters in spelling. These researchers also argued that a "damaged" version of the simulation produced spelling errors similar to those of brain-damaged patients with deficits affecting serial-order processes in spelling. The first part of this study has focused on the limitations of the simulation by examining the error patterns produced by the "damaged" simulation. This work discovered that although the simulation was capable of producing many of the common error patterns associated with damage to the serial-order mechanism, the actual spelling errors made by the simulation were markedly different from those produced by patients. As an example, the patients show a strong tendency to produce spelling errors in which they substitute an incorrect letter in their response (e.g., straight → skraight ). An investigation of the distribution of errors revealed that when subjects made substitution errors, the letter that was produced was often a letter not in the word (e.g., length → lendth). On the other hand, the simulation only substituted letters from within the word. Also, when the simulation made a substitution error, the letter substituted in was always the first letter in the word (e.g., length → lenglh). In contrast, when patients substitute a letter from within the word, these letters come from many different positions in a word. These results demonstrate that simply observing that the simulation is capable of making substitution errors is not enough; care must be taken to ensure the properties of these errors are similar to those produced by patients. The second part of this study focuses on investigating the formal properties of the simulation that caused the specific errors to be produced. This ongoing work will aid in understanding why the simulation does not produce the same errors patients do.
A second project has recently been initiated which grew out of the one just described. It is concerned with creating a novel serial-order simulation for spelling that matches the error patterns patients make in relevant ways, and allows for a formal analyses of its performance. The development of this simulation is guided by principles and techniques described by Paul Smolensky and Geraldine Legendre in their recent book The Harmonic Mind. In particular, tensor product representations will be used as a method for instantiating and manipulating symbolic constructs in a computational simulation system. This approach allows for both a high-level symbolic description of the functions being computed as well as a low-level formal analysis of the operations being performed to give rise to the symbolic phenomena. Once this simulation is developed it will be used as a means of investigating several open questions regarding the orthographic representations and computational processes used in spelling.
Houghton, G., Glasspool, D.W., and Shallice, T., (1994). Spelling and serial recall: Insights from a competitive queueing model. In G.D.A. Brown and N.C. Ellis (Eds.), Handbook of spelling: Theory, process and intervention. Wiley: Chichester.
Smolensky, Paul & Legendre, Géraldine. 2006. The Harmonic Mind: From Neural Computation To Optimality-Theoretic Grammar Vol. 1: Cognitive Architecture. MIT Press.