Goldrick, Matthew. (in press). Utilizing psychological realism to advance phonological theory.

Hello All,

I just finished reading a draft of Matt Goldrick’s chapter from the upcoming Handbook of Phonological Theory (2nd Edition). I enjoyed it and found it helpful in the way it covers the relationship between theoretical work on generative grammar and psycholinguistic work. So, I wrote a short summary, which I’m posting


Goldrick essentially reviews the role of phonotactics in psycholinguistic literature, but takes as a starting point the term “psychological reality” as it was used by Sapir (1933) to refer to the cognitive status of a grammar. Goldrick argues that it is vital for linguists to approach their research with at least some understanding of psychological reality—how things happen in real time, for instance—and that theories of grammar can only be improved by consideration of related data from the psycholinguistic literature.

As an example, Goldrick discusses the division between pre-lexical processing and lexical processing in the speech perception literature. Pre-lexical processing refers to a cognitive function which takes in fine-grained acoustic information (something like the signal sent along the auditory nerve) and spits out a pre-lexical but phonologically detailed representation. Lexical processing then takes this representation that has been passed to it and finds the corresponding entry in the mental lexicon . The term ‘function’ is used consistently to refer to a theoretical mapping of inputs (say, the signal from the auditory nerve) to outputs (phonemic representations). Following Marr (1982) and Smolensky (2006), Goldrick contrasts these levels of description with a higher algorithmic level—which details how a function is computed—and a lower neural level—which explains how the brain acheives a function. Of course, description at all three levels is necessary.

Goldrick works through evidence that categorical and gradient phonotactics influence both pre-lexical and lexical processing stages withing the larger cognitive task of single word recognition. As an example, identification tasks show listeners erroneously hear ill-formed sequences as well-formed ones; discrimination tasks show listeners have difficulty keeping separate words with ill-formed sequences and well-formed words that contain the likely repair strategies for the ill-formed words. Importantly, identification and discrimination errors do not always lead to real words, so they are arguably pre-lexical processing effects. More broadly, the reviewed psycholinguistic literature supports the existence of phonotactic representations apart from lexical ones, and it appears the representations are actively engaged in multiple cognitive functions, including both pre-lexical and lexical processing.

As something of a cautionary tale to linguistics, Goldrick talks about what can be gleaned from studies of wordlikeness judgments. First, he points out that the cognitive mechanisms employed in a judgment task are poorly defined (as Rob Fiorentino would say, it’s a very offline task), so it’s difficult to say what in the task reflects grammar (a mapping between surface and underlying forms) and what reflects other cognitive functions. We know, for example, that lexical neighborhood affects influence judgments (Bailey & Hahn, 2001). We also know from Luce and Vitevitch’s work (see refs below) that having real works in other tasks with nonwords increases the effects of lexical neighborhoods, and recently Shademan (2006, 2007) has shown that including real words in a judgment task does the same thing. Albright (2009) argues that the distribution of phonotactic probabilities within a nonword set also influences the relative roles of lexical and phonotactic effects on judgments. Finally, Goldrick notes that judgments may be the result of prior processes, such as perceptual effects that warp the percept. For example, Dupoux, Kakehi, Hirose, Pallier, and Mehler (1999) showed that Japanese listeners “repair” illegal consonant clusters by inserting an epenthetic vowel (cf. Berent et al., 2008, in PNAS).

For linguists, all the literature above means that claims to study competence apart from performance are not tenable, at least if our data are from wordlikeness tasks. We can’t study competence from these tasks because we know that the judgments are influenced by extra-grammatical factors. Therefore, Goldrick’s initial goal of emphasizing the importance of psychological reality within the study of linguistics holds. Beyond that, Goldrick offers several steps for future research, including some ideas for the study of the interaction of phonotactic and lexical knowledge.

Representative References

Albright, Adam (2009). Feature-based generalisation as a source of gradient acceptability. Phonology 26: 9-41.

Bailey, Todd M. and Ulrike Hahn (2001). Determinants of wordlikeness: Phonotactics or lexical neighborhoods? Journal of Memory and Language 44: 568-591.

Berent, Iris, Tracy Lennertz, Jongho Jun, Miguel A. Moreno, & Paul Smolensky (2008). Language universals in human brains. PNAS 105: 5321-5325.

Dupoux, Emmanuel, Kazuhiko Kakehi, Yuki Hirose, Christophe Pallier, and Jacque Mehler (1999). Epenthetic vowels in Japanese: A perceptual illusion? Journal of Experimental Psychology: Human Perception and Performance 25:1568-1578.

Marr, David (1982). Vision. San Francisco: W. H. Freeman and Company.

Sapir, Edward (1933). La Réalité psychologique des phonèmes. Journal de Psychologie Normale et Pathologuique 30: 247-265. English translation reprinted in David G. Mandelbaum (ed.) (1949), Selected Writings of Edward Sapir in Language, Culture and Personality 46-60. Berkeley, CA: University of California Press.

Shademan, Shabnam (2006). Is phonotactic knowledge grammatical knowledge? In Donald Baumer, David Montero, and Michael Scanlon (eds.) Proceedings of the 25th West Coast Conference on Formal Linguistics 371-379. Somerville, MA: Cascadilla Press.

Smolensky, Paul (2006). Computational levels and integrated connectionist/symbolic explanation. In Paul Smolensky and Géraldine Legendre The Harmonic Mind: From Neural Computation to Optimality-Theoretic Grammar (Vol. 2, Linguistic and Philosophical Implications) 503-592. Cambridge, MA: MIT Press.

Vitevitch, Michael S. (2003). The influence of sublexical and lexical representations in the processing of spoken words in English. Clinical Linguistics & Phonetics 17: 487-499.

Vitevitch, Michael S., Paul A. Luce (1999). Probabilistic phonotactics and neighborhood density in spoken word recognition. Journal of Memory and Language 40: 374-408.


2 thoughts on “Goldrick, Matthew. (in press). Utilizing psychological realism to advance phonological theory.

  1. Kyle Gorman

    Thank you for this interesting summary. Since I don’t have access to this yet, it may be addressed in the text, but what specific theoretical assumptions give rise to the idea that wordlikeness judgements tap into grammar? This is worth being explicit about.

  2. Matt Goldrick

    Hi Kyle,

    Sorry for the long reply time; I just noticed this comment (thanks Eric for the blog re-design!)

    First off, you can download the paper here:

    I think your question is really interesting. I don’t think there is one specific set of assumptions that lead you to this conclusion. I think that there’s a wide array of assumptions about the functional locus (or loci) of grammar within the cognitive system that would lead to the prediction that wordlikeness judgments would be influenced by grammatical wellformedness distinctions.

    The indeterminacy regarding what theoretical assumptions generate this prediction reflects the simple fact that any behavior reflects the complex interaction of multiple cognitive processes. If ‘grammar’ is reflected in the structure of any one of these processes–or more than one–wordlikeness judgments will be influenced (in part) by grammatical wellformedness distinctions.

    The real difficult problem–as I try to discuss in the chapter–is the inverse inference. Given some pattern of wordlikness judgments, how do we relate it back to grammar? The complex interactions between distinct cognitive processes make this a difficult proposition. Given many possible sources, how do we attribute some behavior to a specific source?

    I believe that the solution involves two important steps. First, you must make explicit your assumptions about the nature of the cognitive processes that generate any behavior. Second, I think that these assumptions have to be consistent with what is know regarding the nature of the cognitive system. As I detail in the chapter, I think that current work in wordlikeness judgments makes some highly suspect assumptions–both given what is know about wordlikeness judgments specifically as well as judgment and decision making processes in language processing more generally.

Leave a Reply

Your email address will not be published.