The Blending Hypothesis

Comments welcome
George W. Grace
University of Hawaii

Reflections on the evolution of human language:

2. The Emergence of Analytic Processing:

2. A Role for Blends?

In his important book on the evolution of language (Deacon 1997), Terrence Deacon identifies the use of symbols--"symbolic reference"--as the distinguishing feature of human language. The previous forms of reference--iconic and indexical--involve holistic signals that are associated with particular recurring situations. A conspicuous difference of symbolic reference is that it involves the composing of ad hoc signals to refer to novel situations.

The central problem in the evolution of human language is how the transition to symbols came about. In fact, the question arises with regard to two kinds of symbols. First, there are the conventionalized symbols such as words. Deacon points out that these have meaning of a new kind--what has been called "sense" as opposed to "reference". The sense of a word may be defined only in terms of other symbols--consider dictionaries and thesauruses for example. By contrast, each iconic or indexical signal stands alone--it doesn't depend on any larger system.

Second, there are ad hoc symbols such as sentences. These can be composed to fit the occasion by combining conventional symbols according to a system of grammatical rules. Such ad hoc symbols are then accessible to analytic processing by whoever hears them.

Thus, there appear to be two questions: How did signals like words, whose meanings are defined in terms of other words, evolve? And how did the patterns (or rules) that permit the construction of ad hoc signals like sentences evolve?

In Grace 2004 I discussed the phenomenon of phonaesthesia as providing a possible clue to the second question. Here I want to propose what has been called "contamination" or "blending" as another possible clue.

Blends

"Brunch" and "smog" are two convenient examples of what such blends look like. Blending of this sort is frequently used in the advertising and marketing world--especially in naming new products. However, blends are also often produced accidentally. Charles Hockett (1958: 433) describes a case where a father (presumably Hockett himself), hesitating between "shout" and "yell", actually called out to his noisy children, "Don't shell so loud". An appendix to Speech errors as linguistic evidence (Fromkin 1973) lists 65 examples of such accidental blends that Victoria Fromkin had collected.

Blends have often been proposed as likely sources of lexical innovations. This idea has been particularly associated with Edgar Sturtevant (1917/61). It's hard to identify with certainty the words that have originated from such accidental blends, but in many languages there are words that have suddenly appeared with no known source that are suspected of resulted from such slips of the tongue. Hockett (ibid.) lists bash, clash, flare, glimmer, and smash as words that "are believed to have originated in blends". (He also includes a not-yet-fully-accepted word that one frequently hears today--it doesn't seem to have an established spelling yet, but seems to be most frequently spelled "squush"). Moreover, there are numerous other English words with no known etymology that have been suggested at one time or another as probable products of blending.

The blending hypothesis is, in fact, very parallel to the phonaesthesia hypothesis proposed in Grace 2004. It likewise presupposes a stage of development in which the ancestral species had acquired a considerable inventory of signals that permitted them to discriminate and refer to an increasing number of different situations.

The Hypothesis

The blending hypothesis goes like this: The users of those signals also occasionally made the equivalent of slips of the tongue, and some of these were blends analogous to those of today. That is, sometimes an individual confronted with a particular situation was faced with a momentary uncertainly as to what to call it--as to which of two signals would best represent it. And occasionally s/he would come out with a blend of the two.

Occasionally such a blend caught on and became established as part of the inventory of signals. Such blends would no doubt have evoked their sources (one or both components of the blend) in the minds of some individuals. This in itself would have suggested the notion, however unconscious, that the form of a signal might offer some clue as to meaning.

The consequences of such blending were largely the same as those I proposed for phonaesthesia. The blends also provided a hint of the possibility of signs being motivated. And if some signs might be motivated, then their meanings should be partially or entirely deducible from their form. In short, the concept of analytic processing was suggested. Once the concepts of motivated signs and analytic processing became sufficiently well established, it was only a relatively small step to the idea that it should be possible to compose such analyzable signs deliberately.

Some individuals could then undertake to compose occasional neologisms to refer to previously unnamed situations. I've suggested phonaesthemes as the most likely candidates for such recombination--as the most likely forerunners of morphemes (and words)(1). However, it also seems possible that the success of some accidental blends might have suggested the idea of producing deliberate blends. In fact, it's possible to imagine an accumulation of blends being a source of some phonaesthemes; it's also possible to imagine conscious or unconscious awareness of phonaesthemes as constituting a favorable condition for the occurrence of accidental blends.

Whatever the precise derivation of the neologisms, occasional ones were subsequently picked up and used by others--some to such an extent that they became part of the recognized inventory.

To complete the hypothesis: The success of some neologisms led to the production of more. As the production increased, particular elements (whether phonaesthemes or not) came to be recognized as particularly apt for recombination. At the same time, some particular patterns of combination began to take shape--for example, certain elements came to be used in initial position in the combinations, certain others in final position. And finally, to quote from Grace 2004:

"This would have been followed by something like the following sequence: as time passed the possibilities of such recombination came increasingly to be exploited. As this happened, parts suitable for recombination were increasingly isolated and given conventional recognition as potential elements for combination--i.e., approaching or achieving the status of morphemes.

"Finally, the elements available for combination and the conventions governing their combination reached the point that they could be called a grammar. Likewise, the frequency of innovation increased to the point that a large proportion of utterances had something ad hoc about them. At this point the analytic processing strategy had taken its place as a full-fledged partner of the holistic."

Conclusions

The phonaesthesia and blending hypotheses constitute an attempt to show how a process of gradual evolution might have led from a system of holistic signals to one in which ad hoc symbols like sentences can be composed. In other words they propose an answer to the second of the two questions posed at the outset. The first, how symbols like words--signs that have "senses" defined in terms of other words--evolved would lead us into the whole question of the linguistic construction of reality (on which see my book of that title--Grace 1987).

I've described this evolution as if it were solely a cultural evolution--a gradual expansion in the repertoire of signals occasioned largely by a gradual growth of flexibility. I assume this expansion to have been favored by an expansion in the uses found for these signals. There have been various suggestions as to what these selectively-favored uses might have been such as Dunbar's (1996, on which see also Grace 2003) social bonding or "gossip" hypothesis, or Miller's (2000) sexual choice hypothesis. I won't try to choose among them here.

Although I've only talked in terms of the cultural (or behavioral) aspects of the evolution, I assume that some biological evolution must have been required to permit this cultural evolution to continue. However, I would not be competent to speculate about its nature or timing. I can only suggest that there was a gradual selection in favor of those speakers who proved most fit at exercising the new options. Presumably at some point in this development the biological evolution that led to the language capacity shared by all normal contemporary humans had been completed, but I won't try to speculate about where in the proposed sequence this point might have come.

NOTE

1. It has been pointed out (see for example, Blust 1988: 68) that the distribution of phonaesthemes is limited to semantically related forms. Therefore, a considerable evolution would have been required before any phonaestheme could have assumed the role typical of morphemes and words. However, I don't think that is a major problem for this hypothesis when the size of the available time scale is taken into account. It does, of course, require that the phonaestheme over time became sufficiently firmly associated with the meaning of the signals in which it appeared that it was able to act as an independent agent--a sufficient carrier of that meaning by itself--so that it was able to attach that meaning to other combinations in which it participated.