The meaning of spatially unified English compound words differs from the meaning of its constituent words when these constituents appear as free lexemes. For such morphologically complex words, one central question is whether these words have a lexical representation of their own or if they are only lexically represented by their morphological constituents. Two general models have emerged to address this question. According to a lexicalised model such compound words have an independent entry in the mental lexicon. In contrast, the predominant model of compound word recognition, morphological decomposition, proposes that each lexeme retains a distinct function during word recognition.

To determine whether compounds are morphologically decomposed, we selected English compound words in which we controlled the overall word frequency, but orthogonally varied the frequencies of the individual lexemes (low/low, low/high, high/low, high/high). If compounds are decomposed into their constituent lexemes, then corresponding frequency effects should emerge for one or both lexemes. To examine whether the locus of such effects, if any, is driven by the location of overall compound meaning, we selected for each condition compounds in which the overall meaning was essentially associated with the beginning lexeme ("headed") or the ending lexeme ("tailed"). If this is the case, a bigger frequency effect for the beginning lexeme should emerge for the "headed" compounds. The opposite should be true of the "tailed" compounds, with a bigger effect emerging for the ending lexeme. The data from a lexical decision and a naming task demonstrated a robust frequency effect for both first and second lexeme, supporting a morphological decomposition model. The data also showed that the frequency effects were mainly influenced by the location of the meaningful lexeme, with a larger frequency effect in the beginning lexeme for headed compounds and a larger effect in the ending lexeme for tailed compounds. This also indicated that the interaction between meaning and frequency occurred after the compound word form was accessed.

We are currently working on an eye-movement experiment where the same compounds are embedded in sentences in order to examine lexeme and meaning effects in a more natural reading environment. The eye-tracking task should give a more accurate time-locking of lexeme and meaning effects, since it allows analyses of immediate processing (e.g., first fixation duration) as well as delayed processing (e.g., gaze durations, second fixations and/or spillover). For example if the interaction between meaning and frequency accrues after the meaning of the whole compound word was accessed, we would expect an interaction in the gaze or second fixation durations on the compound word but not in the first fixation durations.

In addition, we will also present data from lexical decision, naming, and eye-tracking tasks in which we compare English transparent compound words, where the overall meaning is equally split between both lexemes, and opaque compound words, where the overall meaning is not related to the meaning of the individual constituents.